SpaceX plans to send an uncrewed Dragon 2 capsule to Mars in 2018, targeting human missions on the Mars Colonial Transporter by 2024 with arrival in 2025. (17:06)
Tesla booked approximately 400,000 Model 3 reservations without paid marketing and moved its total production timeline forward by two years, targeting 500,000 vehicles delivered in 2018 and roughly 1,000,000 by 2020. (23:28)
Musk established OpenAI as a nonprofit to prevent digital superintelligence from becoming dangerously concentrated in corporate hands, and proposed a direct brain-computer neural lace interface to overcome human input-output bandwidth bottlenecks. (40:41)
Musk considers autonomous vehicle engineering fundamentally solved, predicting Tesla will achieve autonomy superior to human drivers within two years, while regulatory validation will require billions of fleet miles. (77:56)
Detailed notes
## Background
Published on June 2, 2016, on the YouTube channel On with Kara Swisher, this video features a full-length interview from Code Conference 2016 where SpaceX and Tesla CEO Elon Musk sits down with Recode's Kara Swisher and The Verge's Walt Mossberg. According to the video description, the discussion covers Musk's plans to send a rocket to Mars in 2018 with human colonists arriving by 2025, escalating competition in electric vehicles and autonomy from companies like Apple and Google, and his drive to democratize artificial intelligence.
## TL;DR
Elon Musk outlines his technical and philosophical roadmaps across SpaceX, Tesla, and artificial intelligence, beginning with the physics and economics of reusable orbital rocketry, the Falcon Heavy, and planned Mars missions starting with an uncrewed Dragon 2 flight in 2018. He details Tesla's production acceleration for the mass-market Model 3, demystifies battery chemistry, and explains why he views Apple as a more direct future automotive rival than Google. Musk also issues warnings about concentrated digital superintelligence, advocating for broadly distributed AI through OpenAI and proposing neural lace brain-computer interfaces to prevent human obsolescence. Across wide-ranging audience questions, he further addresses Hyperloop's engineering trade-offs, simulation theory probability, fully autonomous driving timelines, and direct democracy governance models for Martian colonies.
## Atmosphere & style
The session carries an intellectually charged, dryly humorous, and occasionally apocalyptic atmosphere as mundane business queries quickly escalate into existential territory. Swisher and Mossberg maintain an energetic, no-nonsense journalistic banter, prodding Musk on past delivery delays, manufacturing bottlenecks, and competitive pressures to keep his sweeping visions tethered to commercial reality. Musk responds with his hallmark halting, first-principles cadence, seamlessly toggling between granular industrial supply-chain logistics and startling philosophical musings on hot-tub simulation debates, neural bandwidth limits, and avoiding a future where humans become pampered house cats to digital superintelligence.
## Bulletin
- [17:06](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1026s) SpaceX plans to send an uncrewed Dragon 2 capsule to Mars in 2018, targeting human missions on the Mars Colonial Transporter by 2024 with arrival in 2025.
- [23:28](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1408s) Tesla booked approximately 400,000 Model 3 reservations without paid marketing and moved its total production timeline forward by two years, targeting 500,000 vehicles delivered in 2018 and roughly 1,000,000 by 2020.
- [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) Musk established OpenAI as a nonprofit to prevent digital superintelligence from becoming dangerously concentrated in corporate hands, and proposed a direct brain-computer neural lace interface to overcome human input-output bandwidth bottlenecks.
- [01:17:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4676s) Musk considers autonomous vehicle engineering fundamentally solved, predicting Tesla will achieve autonomy superior to human drivers within two years, while regulatory validation will require billions of fleet miles.
## Key takeaways
- [01:06](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=66s) Orbital mechanics requires tremendous horizontal velocity (roughly 17,000 mph) rather than simple vertical altitude, which dictates ocean drone ship booster landings on high-energy ballistic trajectories.
- [09:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=545s) First-stage booster reusability is economically critical because the first stage represents roughly 70% ($30 million to $35 million) of total launch vehicle hardware cost.
- [10:08](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=608s) Propulsive vertical landing with engines is the only viable multi-planetary recovery method because wings and runways are useless in a vacuum or on the Moon and Mars.
- [11:40](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=700s) Becoming a multi-planetary species is not just an insurance policy or escape hatch from Earth, but a philosophical imperative to extend human consciousness and inspire humanity.
- [01:06:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3960s) SpaceX functions as a baseline transport infrastructure provider for Mars, analogous to the Union Pacific Railroad, enabling independent commerce and settlement to emerge.
- [26:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1565s) Elon Musk views Apple as a far more direct manufacturing competitor to Tesla than Google, which operates primarily as an autonomous software licensor.
- [32:49](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1969s) Battery energy density is no longer the binding bottleneck for electric vehicle adoption; mass-market scale depends entirely on lowering the cost per kilowatt-hour through manufacturing iteration.
- [01:08:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4119s) To prevent production bottlenecks experienced with Model S and Model X, Model 3 was re-engineered specifically for manufacturability and scheduled for a hard design freeze.
- [56:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3389s) Without high-bandwidth neural interfaces, biological humans face intellectual obsolescence, potentially living catered, subordinate lives akin to house cats relative to artificial superintelligence.
- [57:53](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3473s) A direct neural lace interface could provide a third digital cognitive layer above the human limbic system and neocortex, theoretically installed endovascularly via the bloodstream.
- [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) Given rapid historical progress in computing from Pong to photorealistic simulations, Musk argues the statistical probability that humanity inhabits base reality is only one in billions.
- [01:21:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4916s) Governance of future Martian colonies should rely on direct democracy rather than representative legislatures, with higher thresholds to enact laws than to repeal them and mandatory sunset clauses.
## Detailed walkthrough
### [00:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=0s) Late Arrival and Landing Gear Trouble
Elon Musk arrives late to the stage, apologizing to Kara Swisher and Walt Mossberg by explaining that his plane had to fly the entire way with its landing gear deployed. A cockpit warning light signaled an issue, and his pilot cautioned that if they retracted the gear, it might not come back down again. Swisher jokes about the irony of the situation, noting that even Musk cannot escape mechanical landing problems. Musk quips in response, “I have landing gear problems of various kinds.” Swisher thanks him for attending, ribbing him that he was probably drunk when he originally made the promise to come on the show, before steering the conversation toward the dramatic progress SpaceX has made over the preceding years.
[01:06](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=66s) Orbital Mechanics: Space vs. Orbit
Swisher asks Musk to reflect on SpaceX's achievements, particularly landing rockets after many critics assumed it would be impossible. Musk confesses that there were numerous instances where he personally doubted they would succeed. He highlights SpaceX's milestone achievement: landing an orbit-class booster stage both back on solid ground at Cape Canaveral and on an autonomous drone ship stationed out in the ocean.
Addressing common confusion over why SpaceX bothers landing on a drone ship at sea rather than returning to land, Musk explains that orbital mechanics require substantial horizontal velocity rather than merely climbing high. He clarifies the common misconception between simply reaching "space" and achieving "orbit," comparing going to space to crossing into international waters 100 miles offshore, whereas entering orbit is equivalent to circumnavigating the entire globe. In orbit, outward radial acceleration counteracts the inward pull of gravity, producing net zero gravity. He illustrates this with the International Space Station: while it appears to float motionless, it is actually circling Earth at roughly 17,000 miles per hour—more than 25 times faster than a bullet fired from a.45-caliber handgun.
```mermaid
flowchart LR
accTitle: Difference Between Reaching Space and Reaching Orbit
accDescr: Diagram illustrating vertical ascent to space versus the high horizontal velocity required for sustained orbit.
subgraph SpaceOnly["Suborbital Space (~100 km)"]
A["Vertical Lift-off"] --> B["Crosses 100 km boundary"]
B --> C["Gravity still ~95-98%\nFalls straight back"]
end
subgraph OrbitalFlight["Orbital Trajectory (17,000+ mph)"]
D["High Horizontal Velocity"] --> E["Radial Acceleration Balances Gravity"]
E --> F["Sustained Free-Fall Orbit\nCircumnavigating Earth"]
end
```
- At the nominal 100-kilometer boundary of space, gravity is only a few percent weaker than on Earth's surface.
- Escape velocity, Musk stresses, governs spaceflight rather than altitude: “There's no such thing as an escape altitude there's only escape velocity.”
- He compares the Earth's gravity well to a four-dimensional coin funnel: to climb out, an object must spin horizontally at extreme speed.
[04:13](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=253s) The Physics of the Ocean Drone Ship Landing
Musk details why physics dictates landing the booster on an ocean barge instead of flying it back home. Launch vehicles fly straight up solely to clear the dense lower atmosphere—which acts "thick as molasses" at extreme velocities—before immediately pitching horizontally. By the time stage separation occurs roughly 100 kilometers downrange, the booster is moving away from the launch pad at up to Mach 10 (ten times the speed of sound).
Because a rocket in the vacuum of space has no air to bank or push against, reversing direction demands burning fuel equivalent to—or double—the kinetic energy it took to accelerate in the first place: canceling the forward velocity and accelerating back. Rockets cannot carry enough fuel mass to do this on high-energy missions. Continuing along the booster’s natural ballistic arc onto a pre-positioned drone ship—held in place via GPS coordinates down to a one-meter tolerance—is the only physically feasible option.
Musk walks through the booster's descent flight profile:
- **Vacuum phase:** Small cold-gas nitrogen thrusters orient the rocket's attitude where air control surfaces are useless.
- **Atmospheric entry:** Four deployable "waffle-shaped" titanium grid fins guide pitch, yaw, and roll across hypersonic, supersonic, transonic, and subsonic regimes.
- **Terminal landing burn:** The rocket ignites three of its nine Merlin engines to aggressively decelerate, cutting down to a single engine for fine control moments before touchdown.
[09:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=545s) The Economics of Rocket Reusability
Swisher cuts to the heart of the matter, observing that executing this precision maneuver is exceptionally difficult and asks why SpaceX is so fixated on it. Musk explains that the first-stage booster represents about 70 percent of the total cost of the rocket, equating to roughly $30 million to $35 million per launch vehicle.
To convey the financial absurdity of discarding boosters in the ocean, Musk shares the analogy he regularly pitches to SpaceX engineers: “I tell my team it's like imagine there was a pallet of cash that was falling through the atmosphere and it was going to burn up and smash into tiny pieces. Would you try to save it?” Swisher and Mossberg agree that saving it makes complete sense. Musk laments the wastefulness of traditional expendable rocketry, saying he finds it tragic when multimillion-dollar precision hardware smashes into tiny fragments at the bottom of the ocean. Mossberg then begins to ask why fifty years of aerospace history elapsed before anyone attempted this approach.
[10:08](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=608s) Why Wings Don't Work in Space: Propulsive Landing
Walt Mossberg opens by pressing Elon Musk on why governments, the Russians, and legacy space programs never managed to save their spent rockets instead of letting them crash like a discarded pallet of cash. Musk points out that while the Space Shuttle was an attempt at reusability, a rocket booster had never returned to its launch site after an orbital mission until SpaceX did it, let alone touched down vertically on an autonomous drone ship out at sea. When Mossberg notes that traditional rockets never attempted to fly like airplanes, Musk pushes back sharply against the aerospace instinct to make spacecraft look like planes: “The plane thing is not a good idea in my view.”
Musk explains from first principles that every transportation medium dictates its own vehicle architecture. While wings make intuitive sense for atmospheric flight, they become dead weight in space where there is no air to generate lift. Furthermore, designing a winged vehicle assumes asphalt infrastructure that simply does not exist anywhere else in the solar system. “If you want to go somewhere other than Earth, there's also no runways,” Musk notes, highlighting that propulsive landing is the only viable method for landing on the Moon, Mars, or beyond. When Mossberg points out that the Space Shuttle was essentially built like a bulbous glider, Musk deadpans that its aesthetic served a political rather than an aerodynamic purpose: it looked cool and appealing to Congress.
[11:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=699s) Jeff Bezos, Blue Origin, and Becoming a Multi-Planet Species
Mossberg brings up Jeff Bezos, who had appeared on the Code Conference stage the previous evening. When asked about his rivalry with Musk, Bezos had described them as "like-minded" in their broad ambition to lay the basic infrastructure for space travel through reusable rocketry. Musk agrees that they share common ground regarding humanity's long-term trajectory: both believe humanity must become a spacefaring civilization that ventures out among the stars. However, Musk draws a distinct line between merely surviving planetary threats and actively building an inspiring future across multiple worlds.
Addressing Kara Swisher’s interjection about planetary destruction from asteroids, Musk clarifies that his drive is not simply about fleeing an impending catastrophe or remaining a single-planet species that just shifts locations. Instead, he views becoming a multi-planet species as an essential philosophical imperative to extend human consciousness throughout the solar system and eventually to other star systems. For Musk, this vision provides an indispensable emotional spark for humanity: “Life can't be just about solving problems. Like there have to be things that are inspiring and exciting that make you glad to be alive.”
[13:32](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=812s) Fleet Reusability, Launch Backlog, and the Falcon Heavy
Turning to SpaceX's operational pipeline, Musk outlines the company's aggressive immediate and medium-term milestones following their first successful booster recoveries. SpaceX currently has four recovered first-stage boosters in its hangar. Despite experiencing brutal aerodynamic heating and forces upon atmospheric re-entry, the stages remain in remarkably good condition. Musk reveals plans to re-fly the first recovered orbital booster toward the end of the summer, roughly two to three months away, which will prove the viability of orbital-class reuse.
Musk also shares SpaceX's heavy-lift ambitions, targeting the inaugural demonstration flight of Falcon Heavy by the end of 2016. When launched, Falcon Heavy will become the world’s most powerful operational rocket by more than a factor of two, producing roughly 5 million pounds of thrust at liftoff—approximately two-thirds the power of the Saturn V. Musk relishes the historical resonance of launching Falcon Heavy from Launch Complex 39A at Kennedy Space Center, the exact same pad used by Apollo 11. Addressing past schedule slips, Musk notes that the first flight is a demonstration without commercial satellites, as paying customers are booked for operational flights the following year.
- **Launch split:** Roughly 25% of SpaceX launches are dedicated to NASA resupply missions; the remaining 75% serve commercial clients, including broadcasting, communications satellites, and foreign scientific payloads.
- **Flight cadence:** SpaceX is clearing a significant backlog caused by a six-month launch hold from an anomaly the previous year, ramping up cadence to a launch every two to four weeks—a pace exceeding the national launch rates of Russia, Europe, or China.
- **Key commercial contracts:** In addition to standard geostationary satellites, SpaceX is slated to deploy the next-generation Iridium constellation, launching 60 to 70 satellites to provide global broadband coverage.
[17:06](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1026s) Dragon 2, the 2018 Mars Mission, and "The Martian"
Looking beyond low-Earth orbit, Musk details the roadmap for Dragon 2 (Crew Dragon), which is scheduled to fly the following year. Designed to transport up to seven astronauts to the International Space Station, Dragon 2 is equipped with integrated SuperDraco thrusters for propulsive touchdown. Musk emphasizes that this capability makes Dragon 2 far more than a space station taxi: it functions as a general-purpose scientific delivery lander capable of touching down anywhere in the solar system. He drops a major operational target: “We're going to send one to Mars in 2018.”
Swisher and Mossberg immediately question whether Musk plans to be aboard. Musk laughs off the idea, recalling his famous quip about wanting to die on Mars, just not on impact. He explains why Dragon 2 is entirely unsuitable for human Mars transit: with an interior volume comparable to a large SUV, spending the grueling six-month flight inside would be miserable. More crucially, the mission is strictly one-way because the spacecraft lacks the propulsion or fuel infrastructure to launch itself back from the Martian surface. When Swisher jokes that the lack of a return ticket should appear in the fine print like side effects in a pharmaceutical ad, the conversation shifts to Ridley Scott's film *The Martian*. Musk praises the movie for being about 80% scientifically accurate, though he chuckles at Matt Damon's cinematic survival tactics:
> “Well, I mean, I don't think you can sort of just take off from Mars on an unguided rocket really and and then prick your finger on a spacesuit and navigate to a spaceship.”
[20:04](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1204s) Mars Architecture and the Human Launch Timeline
Musk reveals that he plans to present SpaceX's complete architecture for Mars colonization later in September at the International Astronautical Congress (IAC), the space industry’s premier global conference. He stresses that true colonization hinges on industrial-scale logistical capacity: transporting large numbers of passengers and ultimately millions of tons of cargo to build an expanding, self-sustaining city rather than merely planting flags. When Kara Swisher asks why he hasn't traveled to space himself, Musk points out that suborbital flight is trivial compared to orbit, suggesting he might go to orbit in four to five years. Swisher jokes that she would prefer he test the journey first before sending off colonizers, prompting Musk to outline SpaceX's upcoming launch cadence pegged to the 26-month orbital rendezvous windows between Earth and Mars.
Starting with an uncrewed mission in 2018 and continuing through the 2020 window, SpaceX plans to establish an unbroken cargo pipeline that settlers can rely on. If testing proceeds on schedule, Musk estimates human flights could lift off on the Mars Colonial Transporter system by 2024 and touch down on the Martian surface in 2025. When Walt Mossberg jokes about whether this timeline is more reliable than United Airlines' schedule, Musk admits there are real uncertainties. Pressed by Mossberg on whether the launch vehicle will dwarf the Saturn V, Musk plays coy, refusing to give specific dimensions before the September conference beyond confirming it will be "very big." He also firmly rejects the premise that his endeavor is about fleeing Earth. “Why would we abandon Earth? It's really nice here,” Musk insists, dismissing claims that he has advocated for Mars as a pure "Plan B" escape hatch.
```mermaid
flowchart LR
accTitle: SpaceX Mars Exploration and Colonization Cadence
accDescr: A timeline tracing SpaceX Mars mission milestones from 2018 cargo flights to the targeted 2024 human launch and 2025 landing.
subgraph Cargo ["Uncrewed Pipeline (Every 26 Months)"]
M18["2018 Mission<br/>First Cargo Flight"]
M20["2020 Mission<br/>Secondary Cargo Flight"]
M18 --> M20
end
subgraph Crewed ["Crewed Colonization System"]
M24["2024 Launch<br/>Mars Colonial Transporter"]
M25["2025 Arrival<br/>First Humans Land on Mars"]
M24 --> M25
end
M20 -.->|"Validates Architecture"| M24
```
- SpaceX plans a Mars mission during every 26-month orbital alignment window starting in 2018.
- Target human departure date is planned for 2024, aiming for an arrival on Mars in 2025.
- Musk anticipates personally traveling to orbit in four to five years.
[23:28](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1408s) Accelerating Tesla and the Proliferation of Electric Vehicles
The discussion shifts from space back to Earth, focusing on the surge of capital and new entrants pouring into the autonomous and electric vehicle markets. Musk jokes about the flood of announcements, remarking that he is waiting for his own mother to unveil an autonomous EV startup. On a serious note, he points out that four or five Chinese-funded EV startups have already raised over a billion dollars each in the United States alone, while legacy giants like Volkswagen are committing to massive battery factory initiatives. Far from feeling threatened, Musk reiterates that Tesla's overarching mission is to accelerate the transition to sustainable energy, which is why the company open-sourced its patent portfolio and welcomes competitive momentum across the industry.
To maintain its lead and drive the clean transport shift, Tesla has pulled its operational schedules forward, accelerating the production ramp of the mass-market Model 3 by two full years. Musk lays out an aggressive roadmap targeting total vehicle production of 500,000 units by approximately 2018, scaling toward a million cars per year by 2020. Alongside manufacturing scale, Musk highlights software and autonomous capability as essential requirements for the immediate future of the automotive market. As he puts it: “It's going to be odd to have a car without autonomy in the future.”
- Four to five Chinese-backed electric vehicle startups are operating in the US with funding exceeding $1 billion each.
- Tesla's revised Model 3 schedule pulls volume targets forward by two years.
- Musk targets approximately 500,000 total vehicle deliveries in 2018 and roughly 1,000,000 units annually by 2020.
[26:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1565s) Assessing Big Tech Rivals: Google Licensing vs. Direct Competition from Apple
When Swisher and Mossberg ask Musk to evaluate Silicon Valley peers entering the autonomous vehicle market, Musk draws a sharp distinction between Alphabet and Apple. He praises Google for demonstrating the raw technical potential of autonomous driving over years of public road tests, but he notes that Google is fundamentally a software and services company rather than an automaker. Pointing to Google's partnership with Fiat Chrysler, Musk concludes that Google will likely act as a platform licensor to existing manufacturers rather than build an end-to-end car company. Consequently, Tesla views Google's automotive partners as potential rivals, but does not see Google as a direct manufacturing competitor.
Apple, on the other hand, represents a direct challenge. Referring to Apple’s aggressive talent acquisition across the automotive and battery engineering sectors, Musk notes, “Yeah that that'll be more direct.” While he welcomes their entry and hopes the project succeeds, Musk argues that Apple moved too slowly, noting they should have initiated their project much earlier and are unlikely to achieve volume vehicle production before 2020. He frames this delay not as a fatal blow, but as a notable missed opportunity. Even so, Musk cautions that the sheer scale of global auto manufacturing—where the largest player commands no more than roughly 10 percent of the market—means no single competitor will wipe out the rest overnight:
> “The car industry is very big so it's not as though there's you know one company to the exclusion of others.”
Musk elaborates that outsiders routinely underestimate the physical and logistical reality of building cars. The final assembly plant is merely the tip of the iceberg; beneath it lies a sprawling industrial infrastructure of tier-two, tier-three, and tier-four suppliers that involves an order of magnitude more complexity and physical scale than casual observers realize.
[29:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1769s) Traditional Automakers and the Massive Footprint of the Gigafactory
Turning to legacy automakers such as General Motors and Mercedes-Benz, Musk is candid about their progress in electrification. While the incumbent manufacturers continue to announce electric initiatives and develop prototypes, Musk asserts that none of them have yet cracked the formula for a truly compelling, top-tier electric vehicle. “I don't think anyone's any of the car companies as far as I've made a really great electric car,” Musk remarks, challenging the interviewers to identify an existing incumbent model that meets that bar. While he acknowledges they will iterate and eventually build great electric cars over time, he sees an execution gap in what has hit dealer lots so far.
The conversation turns to the physical infrastructure backing Tesla’s ambitious vehicle output: the Gigafactory in Nevada. Mossberg notes that portions of the facility are already operational while construction continues. Musk confirms that once fully constructed, the Gigafactory will possess the largest building footprint of any structure in the world, eclipsing all other factories and commercial buildings globally. When Mossberg pushes back on Musk’s apparent obsession with record-breaking superlatives—from the largest rockets to the largest buildings—Musk counters that excessive scale is an unavoidable physical consequence of their production goals. “It's not scale for scale's sake,” Musk explains. “It's just like if you say, 'Well, we want to accomplish these goals, then then you kind of have to be make a big thing.'”
[30:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1854s) The Gigafactory and Battery Chemistry Realities
Kara Swisher introduces Tesla’s massive manufacturing facility, the Gigafactory, noting its enormous scale and inquiring about an upcoming event. Elon Musk explains that while the factory has already been operational for some time, they plan to hold a celebration soon, joking that it looks like an “alien dreadnought.” When Swisher asks if the factory is producing the exact same lithium-ion batteries found in smartphones, Musk emphatically rejects the comparison and dismantles common misconceptions regarding energy storage. He points out that “about you can believe about 1% of what you read” regarding battery breakthroughs, noting that the umbrella label “lithium-ion” encompasses a vast spectrum of chemistries with radical differences in power density, energy density, and cycle life.
Musk explains that cell terminology is fundamentally misleading because lithium accounts for merely about 2% of the cell mass: “the lithium is actually 2% of the cell mass. So so like it's like the salt in the salad.” Rather than calling them lithium-ion batteries, Musk suggests “our battery should be called nickel graphite cuz it's mostly nickel and graphite.” Specifically, the cathode composition is nickel-cobalt-aluminum, while the anode contains graphite layered with silicon oxide.
```mermaid
flowchart LR
accTitle: Tesla Battery Cell Mass Composition
accDescr: Diagram contrasting the small lithium fraction with the dominant nickel and graphite elements in a Tesla cell.
subgraph Cell["Tesla Cell Composition"]
direction TB
Cathode["Cathode & Anode Bulk (~98% Mass)<br/>Nickel-Cobalt-Aluminum & Graphite / Silicon Oxide"]
Lithium["Active Salt (~2% Mass)<br/>Lithium Ions ('Salt in the salad')"]
end
Lithium -.->|"Enables ion transfer"| Cathode
```
[32:49](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1969s) Moore’s Law vs. Battery Physics: Density, Range, and Cost
Walt Mossberg observes that battery improvements seem painfully sluggish when contrasted with integrated circuits governed by Moore's law, questioning whether Tesla has engineered a major leap in energy yield per volume. Musk clarifies that battery energy density improves steadily at roughly 5% compound growth per year, which adds up to significant gains over time. He dispels the notion that Tesla simply strung together off-the-shelf consumer laptop cells, explaining that automotive applications require distinct cylindrical cell internals that are further specialized at the Gigafactory.
Crucially, Musk emphasizes that raw energy density is no longer the critical bottleneck for consumer electric vehicles. A contemporary Model S already offers roughly 300 miles of range—rarely exceeded in single driving stretches—and Tesla could readily construct a 400-mile car today. Instead, the true hurdle is driving down the dollar cost per unit of energy to unlock mass-market affordability. Musk outlines that this reduction depends on two parallel tracks: constant design iteration (much like the evolutionary history of smartphones) and aggressive economies of scale.
- **Energy Density Growth:** Progresses at roughly 5% annually, compounding significantly over successive design generations.
- **Form Factor & Optimization:** While maintaining a cylindrical geometry, cell internals at the Gigafactory are purpose-built for automotive duty cycles rather than consumer electronics.
- **Range vs. Economics:** Tesla already delivers ~300-mile ranges, making cost per kilowatt-hour the primary operational challenge rather than vehicle distance limits.
[35:50](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2150s) Model 3 Demand and Mass-Market Economics
Swisher shifts the conversation to the extraordinary demand for the Model 3, confirming that Tesla booked approximately 400,000 advance reservations despite running no traditional advertising campaigns, guerrilla marketing, or high-budget spectacles beyond a modest webcast before 1,000 attendees. When Swisher suggests the overwhelming response is fueled by the personal mystique and promises of Musk himself, Musk deflects, insisting that Tesla’s 15,000-person team deserves the credit for systematically building credibility: first with the Roadster, followed by the Model S (hailed by *Consumer Reports* as the best car tested), and the Model X, which, despite early "teething issues," had become "sublime." Consumers simply extrapolated that Tesla's cheaper offering would maintain the same excellence.
Addressing the economics of the Model 3's $35,000 base sticker price, Musk explains that electric vehicles are fundamentally cheaper to run than internal combustion vehicles:
- **Maintenance Reductions:** Electric powertrains eliminate regular oil changes.
- **Braking Longevity:** Regenerative braking absorbs vehicle momentum, allowing the physical brakes to last the lifetime of the car and reducing regular service to just tire replacements.
- **Parity with Conventional Cars:** The average gasoline car in the U.S. transaction price sits at roughly $32,000 once options are added, placing the $35,000 Model 3 in direct competition.
Mossberg notes that the average selling price (ASP) of the Model 3 will likely climb higher due to option packages. Musk agrees that realized transaction prices will skew higher, but stresses that the base model without any add-ons is designed to be a complete, high-quality vehicle on its own. When Mossberg asks whether full autonomous driving capability will be included on the $35,000 model, Musk hedges, teasing a dedicated Tesla event later in the year before offering a cryptic hint: “Let me just say that we're going to do the obvious thing.”
[40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) The Tech Drive for AI and the founding of OpenAI
Kara Swisher and Walt Mossberg turn the discussion to artificial intelligence, noting that earlier speakers at the conference—including Jeff Bezos, Sundar Pichai, Mark Fields from Ford, and representatives from Microsoft and Facebook—have highlighted a massive industry push toward intelligent assistants. The prevailing Silicon Valley narrative suggests that Siri and Microsoft’s tools will become dramatically smarter, Google will outpace the competition, and technological progress will ultimately help far more than it harms. Elon Musk stakes out a much more cautious view, observing that a comprehensive answer requires deep explanation because not all AI futures are benign.
Musk explains that if humanity builds a digital superintelligence that far exceeds human capabilities in every domain, ensuring that it remains benign is paramount. To counter the risks of corporate consolidation, he co-founded OpenAI as a 501(c)(3) nonprofit. He highlights its governance structure as a critical safeguard: unlike traditional corporations, it has no fiduciary obligation to maximize profits or commercialize proprietary advances. Unlike many slow-moving nonprofits, however, Musk notes that OpenAI operates with intense urgency and has gathered extraordinary technical talent.
```mermaid
flowchart TD
accTitle: OpenAI Governance and Strategy
accDescr: Chart contrasting corporate AI incentives with OpenAI's nonprofit democratization approach.
subgraph Corporate["For-Profit AI Development"]
A["Fiduciary Duty to Generate Profit"] --> B["Proprietary Superintelligence"]
B --> C["Concentration of Power"]
end
subgraph OpenAIModel["OpenAI Non-Profit Model"]
D["501(c)(3) Nonprofit Governance"] --> E["Democratize AI Technology"]
E --> F["Broadly Distributed Capabilities"]
end
C -.->|"Countered by"| F
```
The underlying mission is the democratization of AI power. Musk invokes Lord Acton’s famous philosophy:
> “Freedom consists of the distribution of power and despotism in its concentration.” — Elon Musk
Musk emphasizes that concentrating advanced AI inside a tiny handful of hands creates severe hazards of despotism, whether the despot is the software itself or the humans pulling the levers.
[44:21](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2661s) Probability Streams and the "Only One" Competitor
When Swisher asks if the envisioned despot is the computer or the humans controlling it, Musk clarifies that it would likely be the people controlling it, at least initially. Asked directly if he fears any specific tech giants currently pivoting toward AI, Musk drops a cryptic warning: “I won’t name a name, but there is only one.” Mossberg jokingly checks whether that company is busy building a competing electric car, to which Musk simply doubles down: “There’s only one.”
Musk rejects the framing that this effort is about commercial competition. Rather than fighting over market share in an arena of mutual destruction, he describes OpenAI as an effort to nudge the branching probability streams of human destiny toward positive outcomes. In his mental model, the future is a branching tree of probabilistic streams that converge and collapse onto concrete events before branching again. OpenAI’s function is to maximize the odds that benign timelines prevail.
Addressing the counterargument that open-sourcing AI research arms malicious actors, Musk contends that distributing power widely remains the superior safety strategy:
> “If instead the AI power is broadly distributed and to the degree that we can link AI power to each individual’s will... if somebody did try to do something really terrible, well, then the collective will of others could overcome that bad actor.” — Elon Musk
- **Proprietary Risk**: A single proprietary AI that is a million times smarter than anything else, controlled by a small clique, poses an existential risk that distributed systems can actively check.
- **Democratic Parallels**: Quoting Winston Churchill’s maxim that democracy is the worst form of government except for all the others, Musk argues distributed power is far safer than concentrated control.
[47:16](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2836s) The 2016 Presidential Election and Institutional Checks
Swisher pivots the conversation toward the contentious 2016 U.S. presidential election, noting Musk’s status as a naturalized citizen and asking for his perspective. Musk expresses reluctance to dive in:
> “I don’t think this is the finest moment in our democracy.” — Elon Musk
Musk remains skeptical that his individual involvement would alter the outcome, maintaining that he would step in only if he believed he could make a decisive difference. He draws reassurance from the constitutional design of American government:
> “I’m glad that the framers of the Constitution saw fit to ensure that the president was someone who was captain of a large ship with a small rudder.” — Elon Musk
Musk elaborates that structural constraints naturally restrict the damage or good any chief executive can achieve. If a president had a magic button to instantly improve the economy, they would press it at the speed of light, but no such mechanism exists. When Mossberg raises the president’s authority as commander-in-chief to initiate nuclear strikes, Musk maintains that institutional guardrails and military leadership would prevent arbitrary, unilateral nuclear strikes without consulting Congress.
[50:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3000s) Clarifying Involvement in the Hyperloop
The interview shifts to the Hyperloop transportation concept. Musk clarifies his actual involvement, acknowledging public confusion around his role. He explains that he originally proposed the concept several years prior after shooting his mouth off, only to realize his initial engineering concept was flawed and would not work in practice.
Through iterative design and physics calculations, Musk developed an updated proposal where the physics reliably hung together. Because his executive duties running Tesla and SpaceX fully consume his time, he chose to publish the technical design openly as a white paper, inviting outside entrepreneurs and engineering teams to build and commercialize the technology.
[50:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3054s) Hyperloop Philosophy, Tube Economics, and the Student Pod Competition
Musk clarifies that the primary value of the Hyperloop white paper he published was to spur bold new thinking about transportation systems, rather than settling for incremental upgrades like high-speed rail that fail to match technology Japan built in the 1980s. Particularly in California, he believes innovators should strive to invent something entirely new and dramatically superior. Although he stays neutral and has not invested in any of the third-party commercial startups pursuing Hyperloop—wishing them moral support without playing favorites—SpaceX actively engages by hosting an open student pod competition to encourage fresh engineering approaches.
```mermaid
flowchart TD
accTitle: Hyperloop Optimization Trade-off
accDescr: Diagram contrasting high-mileage tube cost reduction with vehicle pod mechanics.
subgraph TubeInfrastructure["800-Mile Tube System (Two Directions)"]
A["Low-Cost Vacuum Tube"]
B["No Complex Track Actuation"]
end
subgraph PodMechanics["Pod Architecture Options"]
C["Air Bearings / Air Skis<br/>(Compressor air removes drag & suspends pod)"]
D["Electromagnetic Suspension<br/>(High tube track complexity & cost)"]
end
C -->|"Minimizes infrastructure cost"| A
D -.->|"Drives up infrastructure cost"| B
```
Musk explains the engineering trade-off at the heart of Hyperloop design: for an 800-mile round-trip route, the tube represents almost the entire economic cost, while the vehicle pods are comparatively cheap. Consequently, optimizing the system requires keeping the tube as simple and inexpensive as possible. While the winning student team used electromagnetic suspension, Musk's original concept proposed air bearings ("air skis") fed by the nose compressor, redirecting air buildup to eliminate drag while providing lift even at supersonic speeds up to Mach 1.1 without burdening the tube itself with expensive active elements. If commercial startups fail to make Hyperloop a reality, Musk hints he might build a demonstration system himself:
> “I don't want to sort of front run them, you know, and say here's this free idea and then meanwhile I go and do it myself, you know, that wouldn't be nice.”
- Japan achieved bullet train speeds decades ago; Musk questions the point of building high-speed rail systems that cannot exceed 1980s benchmarks.
- Air bearings operate effectively across high speeds and have been demonstrated up to Mach 1.1.
- In Hyperloop route economics, tube expenditure dominates total capital outlay, making tube simplicity the single most critical parameter.
---
[55:02](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3302s) The Tech Landscape and the Talent Mismatch
When asked to evaluate the overall state of the technology industry, Musk rejects the notion that society is entering a low-innovation lull. He points to astonishing advancements in artificial intelligence and breathtaking breakthroughs in genetics as proof of broad momentum. However, he diagnoses an inefficient distribution of top-tier engineering and entrepreneurial talent, arguing that too many gifted founders are concentrated within internet products and consumer web applications when their capabilities are urgently needed in foundational physical industries.
> “I think that there's probably a few too many talented entrepreneurs in kind of the internet space and and I think their talent actually would be better served in some other industries.”
---
[56:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3389s) The Threat of Human Obsolescence: The "House Cat" Scenario
Looking deeper into artificial intelligence, Musk cautions that assuming any steady rate of progress in machine learning, biological humanity will inevitably be left far behind. Even under the best-case, non-dystopian outcome where superintelligent AI remains completely benign and cares for humanity, humans would become functionally equivalent to household pets due to the enormous intellectual disparity.
> “Like the if you assume any rate of advancement in AI, we will be left behind by a lot.”
Musk and the interviewers acknowledge this benign endpoint: humans living comfortable, fully catered lives under benevolent artificial stewards, effectively reduced to the status of a pet or house cat. When asked if that outcome is acceptable, Musk dryly responds:
> “I mean I don't love the idea of being a house cat.”
---
[57:53](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3473s) Neural Lace and Overcoming the Input-Output Bottleneck
Musk argues that avoiding domestic-pet status requires humans to merge symbiotically with machine intelligence. Humans are already rudimentary cyborgs because of digital extensions like smartphones, computers, email, and social networks, which grant individual citizens capabilities far exceeding the tools available to the President of the United States two decades ago. The critical bottleneck holding humanity back is input-output (I/O) bandwidth. While the brain's visual interface accepts massive streams of data rapidly, human output is bottlenecked by the agonizingly slow speed of tapping two thumbs onto glass smartphone touchscreens.
```mermaid
flowchart LR
accTitle: Human-Digital Symbiosis via Neural Lace
accDescr: Flow diagram showing biological brain layers connecting through an I/O interface to digital intelligence.
subgraph DigitalLayer["Digital Interface"]
NL["Neural Lace<br/>(Direct cortical I/O)"]
AI["Digital Intelligence / Cloud"]
NL <--> AI
end
Cortex <-->|"High-bandwidth direct interface"| NL
```
To solve this, Musk proposes developing a direct cortical interface called a "neural lace"—a third digital cognitive layer that sits atop the biological cortex and limbic system:
> “Mostly effectively merging in a symbiotic way with digital intelligence revolves around eliminating the IO constraint.”
Crucially, Musk notes this would not resemble wearable devices like Google Glass, nor would it strictly require open-skull brain surgery. Because neurons consume massive amounts of energy and depend on high blood volume, the human circulatory system provides a natural, pre-existing highway to every neuron. An interface could theoretically be introduced endovascularly through blood vessels—such as via the jugular vein—threading into the cranial vasculature to interface directly with cortical neurons without invasive skull trephination.
[01:01:40](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3700s) The Urgency of Neural Interfaces and Life Outside Work
Wrapping up his thoughts on artificial intelligence, Elon Musk returns to the idea of creating a high-bandwidth direct neural interface to link human consciousness with our digital extensions. Without such a link, humans risk becoming effectively pets in comparison to superintelligent systems. Walt Mossberg asks if Musk plans to explore building this technology himself. Musk hesitates, admitting he does not necessarily want to take on another project, but feels the existential weight of the problem: “Somebody should do it, and I mean, if somebody doesn't do it, then I think I should probably do it.” The explicit goal, he emphasizes, is to prevent superintelligent AI from becoming an external power concentrated under the control of a small group of people, creating an entity so vastly superior in intellect that humanity would be treated like house cats before a god.
Kara Swisher remarks wryly on the ominous tone of the conversation: “Well, this has been really cheerful. Thank you.” She points out that they started the evening talking about asteroid strikes and have now moved to artificial intelligence subjugation. Musk notes that while asteroid collisions are a genuine existential threat, they represent a low-probability catastrophe across human-relevant time scales, whereas artificial superintelligence demands immediate urgency.
Looking to break the heavy mood, Swisher asks what "Elon the house cat" actually does for fun. Musk replies that he leads a relatively normal off-hours life: he watches movies and spends time playing video games with his children.
```mermaid
flowchart TD
accTitle: Neural Interface vs Concentrated AI Risk
accDescr: Flowchart illustrating Musk's argument for personal neural interfaces over concentrated external AI.
subgraph Threat["Existential AI Risk"]
A["External Superintelligent AI"] -->|"Concentrated in hands of a few"| B["Godlike Cognitive Disparity"]
B --> C["Humans Reduced to House Cats"]
end
subgraph Solution["Musk's Proposed Defense"]
D["High-Bandwidth Neural Interface"] -->|"Direct link to digital self"| E["Democratized Cognitive Bandwidth"]
E --> F["Symbiosis with AI"]
end
```
---
[01:03:11](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3791s) Work Triage, Sinusoidal Stress, and Family Life
Opening the floor to audience questions, an attendee asks Musk how he handles the extreme stress of his ambitions and maintains any semblance of everyday work-life balance. Musk explains that his operational life functions in a constant state of "work triage mode." He notes that high-growth companies exist in a sinusoidal state of ups and downs, and managing both SpaceX and Tesla is generally manageable as long as their respective crisis waves do not peak at the exact same moment. When both companies hit major emergencies simultaneously, the strain becomes severe. However, he reflects that at the present moment, things are "motoring along okay," he has the relevant operational context loaded in his head for both businesses, and he can clearly see a viable path to successful outcomes.
Musk emphasizes that family time remains his primary anchor, noting, “Of anything in my life, I would say kids by far make me the happiest.” He jokes that 99 percent of the time children bring pure happiness, even if that remaining 1 percent can be trying. Because children are often engrossed in their own worlds and do not need hours of continuous conversation, he likes to sit in the same room with them, answering work emails while remaining immediately available whenever they want to talk. Outside of playing video games together and traveling, he mentions taking them earlier that week to the newly opened Wizarding World of Harry Potter at Universal Studios Hollywood, praising the creators and recommending the butterbeer to the audience.
- **Stress Management:** Operates in "triage mode," relying on the assumption that SpaceX and Tesla crisis cycles will remain out of phase.
- **Parenting Routine:** Balances intense work demands by sharing physical space with his children, alternating between answering emails and engaging with them.
- **Recent Outing:** Visited the new Harry Potter land at Universal Studios Hollywood on Monday, raving about the park design and the butterbeer.
---
[01:06:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3960s) SpaceX as a Transportation Platform for Mars Entrepreneurs
Evan Burns, founder of Odyssey, asks whether SpaceX intends to launch its own internal consumer business lines—such as satellite-based global Wi-Fi—or if the vision is to serve as an infrastructure platform for outside companies to build upon. Musk clarifies that SpaceX's overarching financial imperative is gathering earth-based revenue related to rockets and spacecraft to fund the immense capital required to develop a Mars transport system.
Once that transportation architecture exists, Musk envisions SpaceX functioning much like the Union Pacific Railroad did during the expansion of the American West. When the transcontinental railroad was built, skeptics dismissed it as pointless because few people lived in California; in time, the transportation backbone fostered a thriving economy. Similarly, SpaceX aims to establish the baseline transport link so independent entrepreneurs can build the Martian economy from scratch:
> “What SpaceX is trying to do is establish a transport link and then try to create a fertile environment for entrepreneurs on Mars to flourish.”
Everything from iron ore refineries to local pizza parlors will need to be developed on Mars. When asked how long a pizza delivery from Earth would take, Musk jokes about it arriving cold, but notes that SpaceX can realistically achieve Earth-to-Mars transit times under three months, with the long-term potential to cut travel down to under a month, even though energy requirements increase exponentially as travel times shrink.
---
[01:08:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4119s) Production Delays and Designing the Model 3 for Manufacturability
Nilay Patel of *The Verge* steps up to question Musk on his history of aggressive, missed deadlines. Pointing out that Falcon Heavy was initially promised for 2012, the Model X faced substantial delays, and the upcoming mass-market Model 3 faces immense expectations with hundreds of thousands of customer deposits already collected, Patel asks what Tesla will do differently to hit its manufacturing timelines for mainstream consumers.
Musk explains that Tesla's central strategic shift for the Model 3 is prioritizing design for manufacturing (DFM) from day one. He admits that when Tesla built the Model S, it was the company's first time engineering an entire car from scratch—unlike the original Roadster, where Lotus provided the glider chassis and Tesla supplied the powertrain. Consequently, the team focused exclusively on designing an exceptional vehicle without understanding the nuances of factory production, resulting in a car that is exceptionally difficult to build. They exacerbated the problem with the Model X by taking the complex Model S platform and piling on intricate, "whiz-bang" technologies.
With the Model 3, Musk insists Tesla is taking the polar opposite approach. The engineering team is committed to designing a simpler, highly appealing car where every single component is optimized specifically to be manufactured at high volume, at an affordable price point, and within the target delivery schedule. He points out that an automobile is an intensely integrated system consisting of several thousand unique parts, leaving production schedules vulnerable to supply chain bottlenecks.
> “The biggest thing is just designing the car for manufacturing.”
[01:10:44](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4244s) Supply Chain Disasters, Vertical Integration, and Model 3 Design Freeze
Elon Musk breaks down the brutal mechanics of automotive manufacturing, explaining that when factoring in tier-two and tier-three suppliers, Tesla relies on several thousand individual companies. Production is dictated by the weakest link: “So, things move as fast as the least lucky and least competent supplier.” He lists a string of bizarre, real-world calamities that have hit their suppliers—factories burning down, earthquakes, tsunamis, massive hailstorms, tornadoes, sunken cargo ships, and even a shootout at the Mexican border that delayed trunk carpet shipments because US border patrol impounded the truck due to bullet holes. That single border incident shut down Tesla's production line for several days.
To mitigate this, Tesla aims to expand its optionality and vertical integration, building internal capabilities so that if an external supplier is unwilling or unable to fulfill orders, Tesla can immediately manufacture the part internally. Addressing the Model 3 production timeline, Musk shares that almost the entire design is finalized and the team is targeting complete "pencils down" within about six weeks. Any cool new design ideas will be pushed into subsequent versions. Musk defends his timeline forecasts against accusations of setting unrealistic targets: “It is actually the schedule I think is true. It's it's not some fake schedule that I don't think is true.” Walt Mossberg asks if Tesla plans to hold a reveal event in six weeks to announce autonomous driving features as part of the pencils-down plan, but Musk clarifies that no event is scheduled for that timeframe.
- Thousands of tier-two and tier-three suppliers are involved in Tesla’s production pipeline.
- Supply interruptions have included natural disasters, sinkings, and a border shootout involving trunk carpet that shut down assembly lines for days.
- Model 3 design reaches complete "pencils down" in roughly six weeks, with further design changes deferred to versions two and three.
- Major supplier partners have committed and remain on board for the Model 3 ramp.
```mermaid
flowchart LR
accTitle: Tesla Supply Chain Mitigation Flow
accDescr: Diagram illustrating how Tesla mitigates third-party supplier disruptions by building rapid internal manufacturing capabilities.
subgraph Disruption["Supply Disruptions"]
A["External Suppliers (Tiers 1-3)"]
B["Natural Disasters / Border Incidents"]
A --> B
end
subgraph TeslaStrategy["Tesla Response"]
C{"Supplier Fails to Deliver?"}
D["Rapid Internal Manufacturing"]
E["Uninterrupted Assembly Line"]
B --> C
C -->|"Yes"| D
D --> E
C -->|"No"| E
end
```
---
[01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) The Simulation Argument and Base Reality Odds
An audience member named Josh raises the philosophical proposition that an advanced civilization would eventually create indistinguishable reality simulations, asking Musk whether he believes we are living in one. Musk laughs and confesses that he has contemplated the topic constantly—to the point where he and his brother instituted a strict rule banning any AI or simulation talk whenever they are in a hot tub because “that really kills the magic.”
Musk then lays out the logical foundation of his belief: forty years ago, video games consisted merely of *Pong*—two rectangles and a dot. Today, civilization possesses photorealistic, 3D multiplayer simulations played by millions simultaneously, with virtual and augmented reality advancing rapidly. Even if the rate of technical progress were to drop by a factor of 1,000, games will become indistinguishable from reality within 10,000 years, which is a blink on the evolutionary timescale. Because billions of computers or set-top boxes could run these worlds, Musk argues that “the odds that we're in base reality is one in billions.” Furthermore, Musk notes humanity should hope this is a simulation: “Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options.”
- The jump from *Pong* (1970s) to modern photorealistic simulations occurred within only 40 years.
- Even if technology advances 1,000 times slower than current speeds, indistinguishable simulations will emerge well within 10,000 years.
- Musk assigns a "one in billions" probability that humanity is currently living in base reality.
- The alternative hypothesis—that humanity never reaches the simulation threshold—implies civilization will be wiped out by an existential catastrophe before getting there.
---
[01:17:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4676s) The Timeline for Autonomous Driving and Regulatory Reality
Jeetu Patel from Box asks two questions: when fully autonomous vehicles will become reality across city and traffic conditions, and when human driving will become illegal or prohibitively expensive. Musk replies assertively: “I really would consider autonomous driving to be basically a solved problem.” He clarifies that dense traffic environments like Beijing are actually straightforward because cars move slowly, can easily brake, and simply keep their distance without risk of striking pedestrians. Similarly, divided highways without pedestrians are relatively simple, and Musk points out that Tesla’s Model S and Model X can already navigate highways more safely than a human driver.
The trickiest environment, according to Musk, is navigating urban streets at speeds between 30 and 40 mph. In response to Patel's follow-up about when an elderly family member could sit unattended in a driverless vehicle, Musk projects that Tesla is less than two years away from achieving complete vehicle autonomy that surpasses human safety. However, he cautions that government regulators will require at least an additional year—and potentially longer depending on the jurisdiction—because officials will demand billions of miles of statistical real-world data proving that autonomous driving significantly reduces collisions before granting unconditional approval.
- Musk categorizes autonomous driving as a technically solved engineering problem.
- Dense urban stop-and-go traffic and controlled-access highways are the easiest scenarios for autonomous systems to master.
- The hardest edge case for vehicle autonomy is 30 to 40 mph travel in complex urban settings.
- Musk predicts complete technological autonomy—surpassing human safety—in under two years.
- Regulatory approval is expected to lag technology by at least an extra year to collect billions of fleet miles demonstrating statistical safety advantages.
[01:20:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4841s) Autonomous Safety Thresholds and the Right to Drive
Musk elaborates on the benchmarks required before autonomous driving achieves widespread regulatory approval. He emphasizes that autonomous vehicles must prove themselves to be dramatically safer than human operators—at least twice as good, and potentially five to ten times better in terms of safety metrics. Furthermore, regulators will demand a statistically relevant dataset to validate these claims, meaning billions of miles driven across widely differing road conditions and edge-case situations. From a timeline perspective, Musk estimates that full autonomy is less than two years away technically, but will likely take around three years to clear the regulatory hurdles.
When asked whether driving might one day become illegal for human beings, Musk rejects the idea of heavy-handed government prohibitions, pointing out that because society functions as a democracy, any such rule would depend entirely on what the public decides. Personally, he is philosophically opposed to banning human driving, stating: “Um like I'm in favor of freedom. Um and and not restricting what people do.” Instead of an outright ban, Musk suggests that licensing requirements could simply become far more rigorous over time, requiring human drivers to demonstrate a significantly higher level of skill before being granted permission to drive manually.
- Autonomous systems must prove to be 2x to 10x safer than human drivers across billions of miles on varied roads.
- Projected timeline: Less than two years for technical viability, roughly three years for regulatory clearance.
- Musk advocates for higher driving test standards rather than outlawing human driving.
---
[01:21:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4916s) Governance, Direct Democracy, and Martian Law
For the final question of the session, an audience member asks Musk about his vision for founding human society on Mars—specifically how he imagines cultural unification, systems of government, and the rule of law being established in the colony's early days. Musk opens with a deadpan joke: “Well, I think I'll just declare king of Mars and I'm going to go.” After the crowd laughs, he pivots to a serious structural proposal, stating that the most appropriate form of government on Mars would be a direct democracy rather than a representative one, where citizens vote directly on individual issues. In his view, direct voting drastically curtails the threat of political corruption that routinely plagues representative systems.
Musk also highlights the dangerous "inertia of laws," arguing that human legal codes tend to accumulate indefinitely because laws have infinite lifespans unless deliberately repealed. To prevent regulatory ossification on Mars, Musk outlines a framework where creating a law requires a higher bar than striking one down. He proposes requiring a 60% supermajority to enact a law, while allowing a 40% minority vote to repeal it at any time. In addition, Musk insists that every single piece of legislation should include an automatic sunset clause: “If it's not good enough to be voted back in, maybe it shouldn't be there.” Kara Swisher and Walt Mossberg bring the interview to a close, with Swisher jokingly telling him, “You'll be a good king,” before thanking him to warm applause from the audience.
```mermaid
flowchart TD
accTitle: Martian Legislative Framework Proposed by Elon Musk
accDescr: Flowchart illustrating the voting thresholds and automatic sunset provisions for passing and repealing laws in a direct democracy on Mars.
subgraph Enactment["Enacting New Legislation"]
A["Proposed Martian Law"] -->|"Direct public vote"| B{"Approval threshold"}
B -->|">= 60% support"| C["Law Enacted with Built-in Sunset Clause"]
B -->|"< 60% support"| D["Proposal Rejected"]
end
subgraph Repeal["Repealing or Expiring Legislation"]
C -->|"Direct public vote"| E{"> 40% vote to repeal?"}
E -->|"Yes"| F["Law Removed"]
E -->|"No"| G["Remains Active until Sunset"]
G -->|"Sunset deadline reached"| H{"Voted back in by >= 60%?"}
H -->|"Yes"| C
H -->|"No"| F
end
```
- **Martian Government Structure:** Direct democracy where citizens vote directly on issues rather than through representatives, minimizing systemic corruption.
- **Asymmetric Voting Thresholds:** 60% voter threshold required to pass a law; 40% threshold required to repeal a law at any point.
- **Mandatory Sunset Clauses:** All laws expire automatically unless explicitly renewed through a fresh vote.
## Notable quotes
> [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options. — Elon Musk
> [11:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=699s) Life can't be just about solving problems. Like there have to be things that are inspiring and exciting that make you glad to be alive. — Elon Musk
> [01:17:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4676s) I really would consider autonomous driving to be basically a solved problem. — Elon Musk
> [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) Freedom consists of the distribution of power and despotism in its concentration. — Elon Musk
> [23:28](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1408s) It's going to be odd to have a car without autonomy in the future. — Elon Musk
> [01:10:44](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4244s) So, things move as fast as the least lucky and least competent supplier. — Elon Musk
> [47:16](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2836s) I’m glad that the framers of the Constitution saw fit to ensure that the president was someone who was captain of a large ship with a small rudder. — Elon Musk
## Mentioned
Alphabet, Apollo 11, Apple, Blue Origin, Box, Cape Canaveral, Consumer Reports, Dragon 2, Elon Musk, Evan Burns, Facebook, Falcon Heavy, Fiat Chrysler, Ford, General Motors, Gigafactory, Google, Google Glass, Hyperloop, International Astronautical Congress, International Space Station, Iridium, Jeetu Patel, Jeff Bezos, Kara Swisher, Launch Complex 39A, Lord Acton, Lotus, Mark Fields, Mars Colonial Transporter, Matt Damon, Mercedes-Benz, Merlin engines, Microsoft, Model 3, Model S, Model X, NASA, Nilay Patel, Odyssey, OpenAI, Pong, Ridley Scott, Roadster, Saturn V, Siri, Space Shuttle, SpaceX, Sundar Pichai, SuperDraco, Tesla, The Martian, The Verge, U.S. Border Patrol, Union Pacific Railroad, Universal Studios Hollywood, Volkswagen, Walt Mossberg, Winston Churchill, Wizarding World of Harry Potter
## Video description
SpaceX CEO Elon Musk talks with Recode's Kara Swisher and The Verge's Walt Mossberg about his plans to send a one-way rocket to Mars in 2018. He estimates colonists could start arriving on the Red Planet by 2025. Musk also talks about the proliferation of electric vehicle initiatives that compete with his other company, Tesla, and why autonomous cars will become the norm. He says he doesn't see Google as a competitor, but that "Apple will be more direct." Plus: Why Musk wants more people to have access to the power of artificial intelligence.
## Background
该视频由 YouTube 频道“On with Kara Swisher”于 2016 年 6 月 2 日发布，记录了在 Code Conference 2016 上对埃隆·马斯克（Elon Musk）的独家专访。视频简介指出，马斯克与 Recode 的卡拉·斯威舍（Kara Swisher）及 The Verge 的沃尔特·莫斯伯格（Walt Mossberg）就 2018 年发射火星飞船、2025 年运送首批移民、电动汽车行业竞争格局（包括苹果与谷歌的动向）以及人工智能的权力普及等议题展开了深入探讨。
## TL;DR
在这场长达近一个半小时的高密度对话中，埃隆·马斯克从 SpaceX 攻克一子级海上垂直回收的轨道动力学和工程经济学谈起，重申了打造多行星物种的愿景，并给出了 2018 年火星货运与 2024 年载人任务的时间表。在地面交通领域，他拆解了特斯拉超级工厂的物理逻辑、电池降本路径以及 Model 3 的易制造性设计，同时预测自动驾驶技术将在两年内成熟。此外，马斯克深入阐述了对中心化超级人工智能的生存担忧，提出了通过开源技术、脑机“神经织网（Neural Lace）”解除人机 I/O 瓶颈以抗衡风险的方案，并分享了他对“世界是虚拟模拟”的统计推导以及未来火星直接民主政体的立法构想。
## Atmosphere & style
整场对话充满了硅谷顶级科技峰会的机智交锋与技术硬核感，斯威舍与莫斯伯格犀利、幽默且频频追问细节，而马斯克则在松弛自嘲与第一性原理的技术深潜之间自如切换。现场气氛既有探讨飞机起落架故障、墨西哥边境枪战丢零件以及“泡澡禁聊模拟假说”等荒诞轶事的哄堂大笑，也有直面超级人工智能危机、人类沦为“宠物猫”乃至文明存亡等终极哲学命题时的凝重与紧迫。观众能够深切感受到一种技术理想主义与冷酷物理现实相交织的张力，仿佛在见证一位工程狂人当场向世界推演宏大的星际未来。
## Bulletin
- [20:04](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1204s) SpaceX 计划从 2018 年起利用每 26 个月一次的地火窗口常态化发射货运任务，并争取在 2024 年发射载人飞船、2025 年让首批人类抵达火星。
- [01:17:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4676s) 马斯克认为自动驾驶在技术层面基本已获解决，预计 2 年内技术成熟，但要在复杂路况累积数十亿英里数据以获得监管审批仍需约 3 年。
- [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) 为防止数字超级智能被单一商业实体或少数人垄断，马斯克与团队将 OpenAI 设为非营利组织，以去中心化和开源方式赋予全人类集体抗衡的能力。
- [56:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3389s) 马斯克提出“神经织网”（Neural Lace）构想，主张通过静脉血管网络植入电极直连大脑皮层，打破人类与数字智能之间严重的输出带宽瓶颈。
## Key takeaways
- [01:33](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=93s) 轨道动力学核心在于极高的水平速度而非垂直高度，海上无人驳船接回火箭是顺应抛物线弹道与节省返航燃料的物理最优解。
- [08:53](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=533s) 第一级助推器占据整枚火箭约 70% 的制造成本（约 3000 万至 3500 万美元），实现垂直回收复用是彻底重塑航天经济性的关键驱动力。
- [10:08](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=608s) 传统航天飞机的机翼构型在无大气的月球和火星完全失效，推进式垂直着陆才是通用于地外星体探测与殖民的普适技术方案。
- [11:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=699s) 人类成为多行星物种并非因为要放弃地球，而是为了拓展生命在宇宙中的生存空间，为人类文明提供超越单纯解决日常问题的宏大远景。
- [30:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1854s) 特斯拉电池包中锂的质量占比仅约 2%（类似沙拉里的盐），真正决定性能与成本的是正负极材料以及依托设计迭代和规模效应带来的降本。
- [26:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1565s) 在汽车产业竞争中，马斯克将 Google 视为技术授权方而非直接主机厂，而 Apple 的整车制造项目将更直接地与特斯拉交锋，但其量产最早预计在 2020 年。
- [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) 自由在于权力的分散而专制在于权力的集中，OpenAI 选择非营利治理旨在消除股东受信义务，避免 AI 超级智能沦为单极垄断力量。
- [50:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3054s) 超级高铁（Hyperloop）的系统经济学决定了必须优先压缩双向几百英里管道的建造成本，因此免改动管道的气垫轴承方案优于需要全线铺轨的电磁悬浮方案。
- [56:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3389s) 人类当前是极度“输出受限”的生物，双眼接收信息的带宽极高，但拇指打字或语言输出极慢，神经接口是防止人类沦为 AI“家猫”的根本共生路径。
- [01:06:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3960s) SpaceX 的终极定位类似于当年的联合太平洋铁路，通过开辟低成本地火运输通道来激活火星本地的重工业、矿产与商业生态圈。
- [01:08:40](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4120s) Model 3 吸取了 Model S 与 Model X 的试错教训，从研发之初就采取面向制造的设计（DFM），大幅简化结构以保证大批量稳定生产。
- [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) 基于视频游戏 40 年来从像素点到写实 3D 的指数级演进，马斯克推导得出人类目前身处“基础现实”的统计学概率仅有十亿分之一。
- [01:22:03](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4923s) 马斯克为火星构想了直接民主政体，并提出废除法律门槛应低于制定门槛且必须附带日落条款，以彻底对抗地球社会普遍存在的立法惯性与官僚僵化。
## Detailed walkthrough
### [00:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=0s) 起落架故障插曲与航天成就回顾
采访开场，卡拉·斯威舍（Kara Swisher）对埃隆·马斯克（Elon Musk）安全抵达表示欣慰。马斯克向全场致歉并解释自己迟到的原因：他乘坐的飞机起飞时出现了起落架警告灯故障，飞行员警告称如果将起落架收起，降落时可能就无法再次放下，因此飞机全程顶着放下状态的起落架低速飞抵会场。面对主持人的打趣，马斯克自嘲道：“是的，我确实总遇到各种各样的起落架问题。”斯威舍随后感谢马斯克信守诺言出席大会，并开玩笑说他当初答应受邀时大概喝醉了。
斯威舍指出，在过去几年里，马斯克旗下的两家公司都取得了令人瞩目的实质性突破，尤其是在外界普遍唱衰的情况下成功实现了火箭回收。马斯克坦言，甚至连他自己最初也不认为能够全部成功。在他看来，近期取得的最重大里程碑，莫过于成功着陆一枚轨道级火箭的第一级助推器——不仅实现了陆地回收（返回卡纳维拉尔角），还成功在海上的无人驳船（无人驾驶浮动平台）上完成垂直降落。
### [01:33](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=93s) 轨道动力学：为什么必须去海上接火箭
马斯克表示，要理解为什么 SpaceX 非要把火箭降落在海上驳船这种看似极其不便的地方，公众需要接受一点轨道动力学常识的科普。许多人混淆了“进入太空”与“进入轨道”的概念。马斯克做了一个形象的比喻：进入太空就像是从海岸向外航行 100 英里离开领海，此时技术上你确实进入了太平洋（太空边缘）；但进入轨道则相当于驾驶船只高速环绕地球航行。
```mermaid
flowchart TD
accTitle: 垂直高度与轨道运行原理对比
accDescr: 对比单纯进入太空与维持环绕轨道的物理机制
subgraph SubSpace["亚轨道 / 仅进入太空"]
A["垂直爬升到 100 km"] --> B["仍受近地重力拉扯"]
B --> C["横向速度不足，直接坠回"]
end
subgraph SubOrbit["入轨 / 环绕地球"]
D["水平加速至约 17,000 mph"] --> E["向外径向加速度平衡地球引力"]
E --> F["形成失重 / 持续绕地环行"]
end
```
马斯克强调，物体想要“升空并留在太空中”，核心在于平行于地球表面的水平速度，而非垂直高度：“在太空中根本不存在所谓的脱离高度，只有脱离速度。”国际空间站之所以看似静止地悬浮在轨道上，实际上是以每小时约 17,000 英里的惊人速度在飞掠地球；相比之下，.45口径手枪子弹的飞行速度略低于音速，空间站的运行速度比子弹还要快 25 倍以上。如果将四维时空中的引力井想象成一个投币漏斗玩具，只有给硬币施加极高的水平自转初速度，它才不会沿着漏斗内壁滑落到底部。
- 在公认的太空边界（约 100 公里卡门线高度），地球引力仅比地表微弱几个百分点，几乎毫无二致。
- 只有当飞船环绕地球的向外径向加速度与向内的地球引力完全抵消时，才会产生净零重力的失重状态。
- 火箭发射初期的垂直向上爬升，仅仅是为了迅速脱离像糖浆一样浓厚稠密的低空大气层；只要看火箭发射的长曝光照片就会发现，它很快就转为几乎水平的贴地飞行轨迹。
### [05:27](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=327s) 超音速分离与网格翼空气动力学控制
马斯克详细拆解了助推级分离时的严苛物理限制。当主助推器与第二级火箭分离时，它所处的位置距离发射场可能只有大约 100 公里，但其相对发射场的逃逸飞行速度已经高达 10 马赫（音速的 10 倍）。在没有空气可以借力反推的太空真空环境下，如果要强行将火箭掉头调回发射场，必须携带极其庞大的燃料与液氧来完全抵消这 10 马赫的前进速度，再反向加速飞回。马斯克解释说：“这并非为了省燃料钱，而是物理规律上根本行不通。”
因此，最优解是让助推器顺着既有的抛物线弹道继续向前飞行，在数百公里外的大海中，由一艘提前根据经纬度精确定位（误差在 1 米以内）的无人驳船在落点待命接应。在重返大气层的过程中，火箭展现了复杂的多阶段姿态控制机制：
```mermaid
flowchart LR
accTitle: 助推器再入大气层与受控着陆流程
accDescr: 助推器从真空姿态调整到气动控制再到着陆点火的全过程
subgraph Stage1["太空真空阶段"]
A["氮气推进器"] -->|"控制姿态与朝向"| B["开始再入"]
end
subgraph Stage2["大气层高速阶段"]
B --> C["展开 4 具钛合金网格翼"]
C -->|"全速域控制俯仰/偏航/滚转"| D["精确定位驳船"]
end
subgraph Stage3["末端降落阶段"]
D --> E["点火 3 台主发动机减速"]
E --> F["切至单台发动机精准触地"]
end
```
针对网格翼（Grid Fins）的设计，马斯克解释其外观酷似华夫饼，优势在于能够在高超音速、超音速、跨音速到亚音速的极其宽泛的速域区间内稳定工作，而常规飞机气动舵面很难兼顾如此悬殊的气动环境。在完成气动减速逼近驳船后，火箭首先启动 9 台发动机中的 3 台急速刹车，最后关停外侧两台、仅留 1 台发动机进行米级精度的末端平稳垂直降落。
### [08:53](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=533s) “天降钞票”：回收助推器的经济动机
面对主持人对这套极其复杂操作背后动机的询问，马斯克算了一笔账：一枚火箭中成本最高昂的部分就是第一级助推器，它占据了整枚火箭总成本的大约 70%，折合单枚造价在 3,000 万到 3,500 万美元左右。
为了让工程师和团队直观理解为什么必须死磕回收技术，马斯克向主持人分享了他经常激励团队的经典比喻：
> “我试着告诉我的团队，这就像是想象有一整托盘的现金正从大气层坠落下来，并且即将烧毁、砸得粉碎——你会不会想办法把它接住？”
马斯克强调，如果能完好无损地将助推器带回地面并重复使用，SpaceX 就不需要每次都耗资数千万美元从头制造一枚新火箭。看着耗费巨资打造的高精密火箭最终在深海粉身碎骨，在经济和工程角度都是极其悲剧的事情。主持人沃尔特·莫斯伯格（Walt Mossberg）随后提出疑问：人类探索太空已经有 50 年历史，为什么在马斯克与 SpaceX 出现之前，从未有人尝试做到这一点？围绕这一航天历史难题的讨论随即展开。
### [10:08](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=608s) 为什么不造带翅膀的航天飞机：推进式着陆的技术逻辑
主持人提到，不仅政府机构和俄罗斯从未尝试过回收一级火箭，杰夫·贝索斯（Jeff Bezos）的公司也做过尝试，并追问为何过去的常规火箭从未试过回收“这一大笔现金”。埃隆·马斯克解释称，航天飞机虽然做过部分回收尝试，但从未实现过轨道任务推进器返回发射场，更不用说在海上驳船上着陆。对于传统火箭为何不像飞机那样设计机翼，马斯克直言：“在我看来，做成飞机并不是个好主意。”
他从物理介质的本质出发剖析道：每一种运输工具的外形设计都必须适应其所处的介质。直觉上飞机造型似乎合情合理，但在太空中机翼毫无用处，因为“太空里根本没有空气”；更关键的是，如果人类的目的地是地球之外的天体，那里同样不存在跑道。正因如此，当年阿波罗登月时采取的是推进式垂直着陆（Propulsive landing）。至于当年的航天飞机为何造得像一架臃肿的客机，马斯克调侃道，那纯粹是因为外形看起来很酷，能够讨得美国国会的拨款青睐。
```mermaid
flowchart TD
accTitle: 航天器着陆方式对比
accDescr: 对比航天飞机机翼滑翔模式与SpaceX火箭推进式垂直着陆的技术逻辑差异
subgraph Shuttle["传统航天飞机路线"]
A["飞机气动布局 / 带有大机翼"] --> B["需要大气层提供升力"]
B --> C["必须依赖地面长跑道"]
C --> D["无法在无大气的月球/火星着陆"]
end
subgraph SpaceX["SpaceX 推进式垂直着陆路线"]
E["无翼箭体构型"] --> F["反推发动机控制减速"]
F --> G["适应海上驳船与陆地发射场"]
G --> H["通用于地球、月球与火星表面"]
end
```
### [11:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=699s) 与贝索斯的异同：多行星物种与星际文明的愿景
主持人谈及前一晚杰夫·贝索斯（Jeff Bezos）受访时的表态，贝索斯称他与马斯克在通用大方向上“志同道合”，都致力于打磨可重复使用火箭的基础设施，以支撑更宏大的远景。马斯克认可这种观点上的相似性，指出两人都深信人类必须成为跨越星际的航天文明，最终走向群星之间，这才是令人兴奋与鼓舞的未来。
马斯克进一步厘清了他对“多行星物种”的执念。这并不仅是为了规避小行星撞击地球等生存灾难，也不只是从一个星球整体搬迁到另一个星球的单一物种状态，而是真正建立起分布于多个星球上的文明体系，让人类与已知生命的足迹从地球拓展至整个太阳系乃至更遥远的恒星系。马斯克在阐述生活动力时说出了一句充满哲学意味的名言：“生活不能仅仅是为了解决一个又一个问题，还必须有那些令人振奋和激动的事情，让你发自内心地为活着而感到高兴。”
### [13:30](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=810s) SpaceX 近期排期：首度复飞二手火箭与重型猎鹰首秀
谈及实现火箭回收这一里程碑后的近期规划，马斯克透露，SpaceX 计划在两到三个月内（约在夏季末）首次复飞一枚已经成功回收的一级助推器。尽管助推器在重返大气层时经历了严酷的极端环境，但目前 SpaceX 已经封存了四枚回收的火箭级，其机械状态都相当优良。
更为震撼的节点计划在年底前到来——首次发射猎鹰重型运载火箭（Falcon Heavy）。马斯克介绍，这枚火箭起飞推力约为 500 万磅，相当于土星五号（Saturn V）推力的三分之二左右，其实力将超过当时现役所有运载火箭的两倍以上，成为全球运载能力最强悍的火箭。颇具历史象征意义的是，Falcon Heavy 将从肯尼迪航天中心发射阿波罗 11 号登月任务的同一个发射台（39A 发射台）升空发射。
- **复飞时间窗口**：计划在 2016 年夏季末（约 2-3 个月后）执行首次回收火箭的复飞任务。
- **箭体储备状态**：截至访谈时，SpaceX 库房中已有 4 枚成功着陆回收的一级火箭助推器。
- **Falcon Heavy 规格**：起飞推力达约 500 万磅，运载力超现役其他火箭两倍以上，共享阿波罗 11 号的历史发射工位。
- **客户与商业模式**：Falcon Heavy 首飞为纯演示飞行，不搭载商业载荷，首批商业客户发射定于次年；相比之下 Falcon 9 积压了大量订单，其中约 1/4 为 NASA 任务，3/4 为商业广播通讯卫星及各国科研任务。
- **发射频次与补进度**：受前一年发射失利导致的约 6 个月停飞期影响，公司正全力消化订单，计划将发射节奏拉高至每 2 到 4 周一次，这一频次不仅远超 NASA 历史水准，也超越了俄罗斯、欧洲和当时的中国航天。
- **大型组网项目**：后续还将承接新一代铱星系统（Iridium NEXT）的发射，该系统包含约 60 至 70 颗卫星，旨在提供全球宽带覆盖。
### [16:36](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=996s) 2018 年火星发射计划与《火星救援》的科学槽点
在规划下一阶段主力飞船时，马斯克表示 2017 年将发射第二代龙飞船（Dragon 2，即 Crew Dragon）。该飞船不仅具备推进式着陆能力、可向国际空间站运送多达 7 名宇航员，还被设计为通用的深空科研运载平台，能把载荷投送至太阳系的任何角落。紧接着，马斯克抛出了一个爆炸性计划：SpaceX 打算在 2018 年直接向火星发射一艘龙飞船。
针对主持人关于“你本人是否会随船前往”的追问，马斯克果断予以否认。他曾说过“死在火星上无所谓，只要别死在着陆撞击时就行”，但他澄清自己并没有某种急切的“火星寻死欲”，只是觉得既然出生在地球，死在火星算是个不错的人生归宿。他强调 2018 年的任务绝不推荐人类搭乘：Dragon 2 内部空间仅相当于一辆大型 SUV，前往火星单程就要约 6 个月，在 SUV 大小的密闭空间里生活半年不仅煎熬，而且最关键的是，2018 年的测试版本根本没有返回地球的能力。
主持人调侃称这就像处方药广告里不起眼的副作用说明一样，顺势聊起了电影《火星救援》（The Martian）。马斯克评价该片 80% 的科学设定是严谨的，但剧本强行串联了一系列极度小概率的极端事件。他边笑边吐槽了电影的高潮桥段：“我不认为你能坐着一枚毫无制导的改装火箭从火星起飞，然后戳破宇航服手指漏气，就这么一路精准导航飞到一艘飞船上去。”虽然物理上并非绝对不可能，但概率微乎其微——除非你是自带光环的马特·达蒙（Matt Damon）。
### [20:04](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1204s) 火星殖民架构、时间表与巨型火箭
埃隆·马斯克透露，他计划在同年 9 月于国际宇航大会（IAC，即全球顶级航天行业大会）上正式公布火星殖民的整体架构。他指出，殖民的核心要义在于能够将海量人口以及数以百万吨计的物资送往火星，唯有如此，才能在火星建立起一个能够自我维持并持续壮大的城市。当被问及为何自己至今尚未飞往太空时，马斯克回应称自己大概会在四到五年内进入地球轨道，但他特别澄清了概念：“太空和轨道是截然不同的两码事。”
关于奔赴火星的具体规划，SpaceX 的策略是从 2018 年起，利用大约每 26 个月出现一次的地球-火星轨道交会窗口，每一次都向火星发射任务，从而建立起外界可以依赖的常态化火星货运班次。马斯克给出了雄心勃勃的载人时刻表：“如果一切按计划推进，我们应该有能力在 2024 年发射载人任务，预计于 2025 年抵达。”当主持人打趣这是否比美联航的航班表更靠谱时，马斯克坦承其中确实存在变数。此外，针对主持人追问那枚用于火星殖民运输系统（MCT）的火箭究竟有多大、是否超越土星 5 号（Saturn V）甚至达到其两倍大时，马斯克守口如瓶，仅不断强调那将是一枚“极其庞大”（Very big）的火箭，一切细节都留待 9 月揭晓。
```mermaid
flowchart LR
accTitle: SpaceX火星时间线规划
accDescr: 展示从2016年架构发布到2025年人类抵达火星的关键节点
A["2016年9月 IAC<br/>公布火星殖民架构"] --> B["2018年 / 2020年<br/>启动常态化货运任务"]
B --> C["2024年窗口期<br/>首次发射载人飞船"]
C --> D["2025年<br/>首批人类抵达火星"]
```
- **火星发射窗口**：地火轨道交会周期约为每 26 个月一次（如 2018 年、2020 年、2022 年、2024 年）。
- **载人时间节点**：计划于 2024 年发射火星殖民运输系统，2025 年着陆火星。
- **运输规模要求**：不仅要运送人员，最终需要向火星运送数百万吨货物以构建生长型城市。
### [23:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1385s) 地球不必放弃：全球电动车浪潮与特斯拉产能加速
主持人提到马斯克过去是否说过“人类可能不得不放弃地球，因此需要 Plan B”，马斯克当即坚决否认，直言自己从未说过这种话：“我们为什么要放弃地球？这里棒极了。”随后话题转向地面业务，主持人问及当前自动驾驶与电动汽车领域的混战格局。马斯克笑称现在的行业发布会多到夸张：“我都在等我妈宣布成立一家自动驾驶电动汽车初创公司了，就像‘妈，你也来凑热闹？’”
玩笑之余，马斯克严肃剖析了行业激增的态势。他指出仅在美国就有四到五家背后有中国资本支持、融资金额达到 10 亿美元以上的电动汽车初创公司，大众汽车等传统巨头也宣布投建大型电池工厂。马斯克重申特斯拉对待同行的态度是全面欢迎的，特斯拉开源自身专利就是为了加速这一进程。为了推动可持续能源时代更快到来，特斯拉主动将 Model 3 的量产计划提前了整整两年。马斯克给出了明确的产能爬坡目标：“我们希望努力在 2018 年前后实现年产 50 万辆汽车的总目标，这是一个非常激进但可实现的时间表。”至于未来，他认为到 2020 年实现年产 100 万辆的路径也相当清晰，同时强调自动驾驶将是不可或缺的标配，未来拥有一辆没有自动驾驶功能的汽车将会显得极其怪异。
- **中国资本入局**：仅在美国本土就有 4 至 5 家由中国资本注资、融资金额达 10 亿美元以上的电动汽车初创企业。
- **传统车企动态**：大众集团宣布建设大型动力电池工厂。
- **特斯拉产能规划**：原计划提前两年，目标 2018 年达到年产 50 万辆，2020 年达到年产 100 万辆。
### [26:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1565s) 剖析对手：技术授权的 Google 与正面交锋的 Apple
当逐一拆解科技巨头跨界造车带来的竞争威胁时，马斯克对 Google 和 Apple 给出了完全不同的定性。他高度评价 Google 在展示自动驾驶潜力方面做出的出色贡献，但指出 Google 本质上不是一家整车制造厂，其商业路径大概率是将技术授权给菲亚特（Fiat）等传统主机厂，因此特斯拉未来面对的直接对手是获得授权的车企，而非 Google 自身。谈及 Apple 时，马斯克直言：“那将会更加直接。”
“在量产规模上，我认为他们最早也要到大约 2020 年才能实现。”针对 Apple 的汽车项目，马斯克根据行业招聘动向与公开迹象判断，虽然乐见其成并希望 Apple 能成功，但他认为 Apple 启动该项目的时机偏晚，错失了一定先机。不过，马斯克并不认为 Apple 晚到就会失败，因为汽车工业体量极其庞大，不存在赢家通吃的局面。全球主流汽车制造商多达十几家，哪怕是行业领头羊，其市场份额最高也不过 10% 左右。他特别指出，汽车工业的制造规模与供应链复杂度外人很难想象：“一旦你深入到二级、三级、四级供应商，其复杂度与规模大概还要再大上一个数量级。”对于现有传统汽车厂商的表现，马斯克直率地给出了犀利评价：“我认为目前为止没有任何一家汽车公司造出了一款真正优秀的电动汽车。”
- **Google 的定位**：技术提供商与技术授权方（如与菲亚特合作），不直接制造整车。
- **Apple 的时间窗口**：预计最早到 2020 年才可能实现规模化量产，涉足虽晚但仍有生存空间。
- **汽车行业格局**：全球重要车企约有十几家，单一公司最高市占率仅约 10%，无法通过一款产品垄断市场。
### [30:18](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1818s) 超级工厂：达成宏伟目标的物理必然
对话随后转向特斯拉正在内华达州建设的超级工厂（Gigafactory）。虽然该工厂在当时仅完工了一部分并已投入初期运营，但马斯克透露，一旦这座巨型工厂完全建成，它将成为全球占地面积最大的单体建筑——不仅超越所有工业制造厂房，而是超越全球任何类别的建筑物。
面对主持人关于特斯拉“不是要做最大的火箭，就是要造最大的建筑”的调侃，马斯克解释这绝非为了追求宏大而追求宏大，而是目标导向下的客观物理需求：“如果你说‘我们想要实现这些目标’，那么你就必须造出一个如此巨大的东西。”他强调，要支撑全球向可持续交通的大规模转型，动力电池的产能需求呈几何级数增长，除了以史无前例的工业规模进行集中生产之外别无他途。
### [30:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1854s) 超级工厂的“外星母舰”与锂电池的成分真相
卡拉·斯威舍和沃尔特·莫斯伯格将话题引向了特斯拉正在建造的超级工厂（Gigafactory）。这座庞大无比的工厂已经投入运营了一段时间，并准备举办一场庆祝派对，马斯克打趣地邀请主持人参加，并将这座超大尺度的自动化工厂形容为“简直就像一艘外星无畏舰（alien dreadnought），太疯狂了”。当莫斯伯格追问这里的电池与智能手机中的锂电池有何区别、特斯拉是否取得了颠覆性的电池技术突破时，马斯克直言外界充斥着对电池的误解与炒作，甚至调侃道：“市面上关于电池的说法废话太多了，你读到的东西大概只能相信 1%。”（“There's so much nonsense out there about batteries. Like about you can believe about 1% of what you read.”）
马斯克详细拆解了所谓“锂离子电池”的本质。他指出，“锂离子”涵盖的技术范围极广，不同化学配方之间的功率密度、能量密度和循环寿命可能存在天壤之别。真正决定电池性能的是正极（cathode）和负极（anode）材料，而锂本身在电芯质量中其实仅占约 2%，“就像沙拉里的盐一样”，无论在重量还是成本构成中都微乎其微。他解释说，大众因为名称以为锂是主体，但特斯拉的电池其实更应该被称为“镍石墨电池”——其正极主要是镍、钴、铝，负极则是带有氧化硅涂层的石墨。
```mermaid
flowchart LR
accTitle: 锂离子电池质量构成示意
accDescr: 展示特斯拉电芯中主要材料构成，镍与石墨占主导，锂仅占2%类似沙拉中的盐。
subgraph CellComposition["电芯内部核心构成"]
direction TB
C["正极材料 Cathode<br/>(主要为 镍、钴、铝)"]
A["负极材料 Anode<br/>(带氧化硅层的石墨)"]
L["活性锂离子 Lithium<br/>(仅占电芯总质量 ~2%，犹如'沙拉中的盐')"]
end
C --- L
L --- A
```
### [32:49](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1969s) 电池能量密度的复利增长与规模降本逻辑
莫斯伯格指出，消费电子产品用户习惯了芯片遵循摩尔定律高速演进，而电池技术在日常设备中似乎推进极慢，特斯拉把全部身家押注在电池上，究竟如何实现突破。马斯克解释，电池的能量密度每年大约稳健提升 5% 左右，这听起来虽然不起眼，但在复利效应的作用下，经过数年积累就会形成非常可观的跃升。他澄清说，很多人误以为特斯拉只是把笔记本电脑的电芯拼凑在一起就造出了一辆好车，但如果是这么简单，竞争对手早就蜂拥而至了；虽然特斯拉沿用了圆柱形封装规格，但电芯内部结构完全不同，超级工厂制造的更是针对汽车工况进行深度优化的产品。
马斯克强调，当前阻碍电动车普及的核心问题并不是能量密度或续航里程。目前 Model S 的续航已经能达到约 300 英里，车主极少有连续开 300 英里不停车的需求，而且特斯拉在技术上完全可以在当下造出 400 英里续航的车。电动车真正需要攻克的核心是降低单位能量的电池包成本，唯有如此才能实现亲民售价。他指出，降低成本并实现大众化主要依托两个维度：一是经历多次版本迭代的“设计迭代”（design iteration），二是“规模经济”（economies of scale）。正是这两者的结合，才使得智能手机从高门槛设备演变为人手一台的口袋超级计算机，汽车电池也必须走通这条工业化路径。
- 能量密度年化增速约为 5%，长期复利积累成效巨大
- 现有车型的 300 英里续航已能满足绝大多数出行需求，技术上具备制造 400 英里车型的能力
- 大众化量产的核心支柱：持续的产品设计迭代与庞大的生产规模效应
### [35:50](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2150s) 40万份订单的背后：从产品口碑到实际拥有成本
当斯威舍提到 Model 3 已经斩获约 40 万份预订订单时，她尖锐地指出，特斯拉几乎没花钱做营销，现场发布会也只有千余人，许多预订完全是基于对马斯克本人的崇拜与号召力，甚至调侃在硅谷只要马斯克上台读花生酱的配料表都会有人欢呼。马斯克对此并不赞同，表示自己不应揽下这份功劳：“我认为特斯拉卓越团队所做的事情，是全公司 15,000 名员工拼尽全力创造出极具吸引力的产品、造出伟大的汽车。”（“What Tesla's done with a phenomenal team is like 15,000 people at the company worked super hard to create compelling products, to create great cars.”）他回顾道，从初代 Roadster、被《消费者报告》（Consumer Reports）评为史上最佳汽车的 Model S，到经历了产能磨合与初期阵痛但日渐臻于完美的 Model X，消费者亲眼见证了特斯拉产品的实力，因而相信特斯拉同样能造出一款既优秀又平价的下一代轿车。
谈到 Model 3 定价 35,000 美元的平价属性，马斯克从汽车全生命周期使用成本（TCO）的角度算了一笔经济账：
- **能源成本差**：日常充电成本远低于汽油支出。
- **机械维护锐减**：无需更换机油；由于配备了动能回收制动系统（regenerative braking），机械刹车片的损耗极低，其寿命几乎与整车寿命相当。
- **常规保养极简**：日常用车基本上只需要定期更换轮胎。
马斯克指出，美国新车平均实际成交价在 32,000 美元左右，Model 3 已经极其贴近主流市场区间。面对莫斯伯格关于选装配置会拉高平均售价（ASP）的质疑，马斯克保证，即使消费者只花 35,000 美元购买不带任何额外选配的基础版，得到的依然会是一辆令人极为满意的优秀汽车。
### [39:35](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2375s) Model 3 的自动驾驶悬念与转向 AI 的担忧
莫斯伯格紧接着追问：售价 35,000 美元的 Model 3 是否会配备自动驾驶功能？马斯克对此没有直接亮出底牌，而是露出神秘微笑，透露特斯拉可能会在年底前后举办另一场专属发布会来详加说明。莫斯伯格和斯威舍现场施压，起哄让他直接在这里宣布重磅新闻，马斯克依然守口如瓶，只留下了一句耐人寻味的话：“我只能说，我们会做显而易见的选择。”（“Let me just say that we're going to do the obvious thing.”）
见马斯克在车辆配置细节上点到即止，斯威舍顺势将访谈转入下一个重要板块——人工智能（AI）。斯威舍提到大会现场关于 AI 的讨论非常热烈，而媒体常把马斯克描绘成一个“害怕机器人”的末日预言者，借此要求马斯克当场澄清他对人工智能技术的真实立场与核心忧虑究竟是什么。
### [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) 创立 OpenAI 与分散人工智能权力
Kara Swisher 指出，此前台上多位科技巨头（如 Jeff Bezos、Sundar Pichai、Mark Fields 以及微软和 Facebook 的代表）都对 AI 抱有极其乐观的态度，认为智能助理竞争将让生活变得更美好；然而 Elon Musk 却持有截然不同的审慎立场。Musk 坦承，要完整阐述其观点需要很长篇幅，但他核心的担忧在于 AI 的未来并非全然良善：“我认为公平地说，并非所有的 AI 未来都是良善的（not all AI futures are benign）。”一旦人类创造出在各方面都远超自身能力的数字超级智能（digital superintelligence），确保其怀有善意便至关重要。
为了防止 AI 技术被单一商业利益劫持，Musk 与其他人共同创立了 OpenAI。他特别强调其治理结构选择为 501(c)(3) 非营利组织的原因：这能避免公司背负为股东追求利润的受信义务（fiduciary duty）。虽然传统非营利组织往往缺乏紧迫感、技术开发缓慢，但 OpenAI 聚集了顶尖人才，并保持着极高的推进紧迫感。Musk 引用阿克顿勋爵的名言“权力导致腐败，绝对权力导致绝对腐败”，进一步引申出：“自由在于权力的分散，而专制在于权力的集中（freedom consists of the distribution of power and despotism in its concentration）。”他认为，AI 的终极危险在于这项技术被极少数人垄断，从而将人类带入被独裁者控制的世界。
```mermaid
flowchart TD
accTitle: OpenAI 权力分散与安全逻辑
accDescr: 展示集中式超级 AI 与分布式 AI 在抗风险能力上的逻辑对比架构
subgraph Centralized ["中心化专制风险"]
A["单一实体 / 专有代码"] --> B["独占式超级 AI<br/>（高出人类百万倍）"]
B --> C["少数人专制统治<br/>（无法被有效制衡）"]
end
subgraph Decentralized ["OpenAI 分布式安全架构"]
D["非营利治理<br/>501(c)(3)"] --> E["技术开源与权力普及"]
E --> F["个人专属 AI 代理广泛分布"]
F --> G["群体意志抗衡恶意行为者<br/>（Collective Will）"]
end
```
- 创立架构：OpenAI 设立为 501(c)(3) 非营利组织，旨在切断追求利润对底层技术的捆绑，实现 AI 技术的民主化（democratize AI power）。
- 奇点难以预测：谈及 AI 独裁者的具体形态，Musk 指出在奇点之下很难精准预见未来，但最现实的威胁是“控制计算机的人”成为统治者。
### [44:21](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2661s) 一家独大的隐患与概率论视角
当 Walt Mossberg 追问在当下竞逐 AI 的科技巨头中是否存在具体的担忧对象时，Musk 坦言：“我不会指名道姓，但确实只有一家（I won't name a name, but there is only one）。”在 Mossberg 开玩笑问及那家公司是否也在造车竞争时，Musk 再次确认他的担忧仅针对这一家公司。Musk 强调，这绝非单纯的商业竞争，因为“为相互毁灭而竞争毫无意义”。从长远来看，关键在于如何提升未来走向积极结局的概率。他将未来比作一组不断分叉、随后又坍缩至特定事件并再次展开的概率流（probability streams），而 OpenAI 的职责就是尽一切可能增加良善未来发生的几率。
面对 Swisher 关于“开源是否会被恶意行为者滥用”的质疑，Musk 提出了 AI 权力的制衡理论。他认为，如果 AI 超级智能仅由单一实体垄断，其智力高出其他事物上百万倍，人类将毫无招架之力；反之，若 AI 力量广泛赋权给每个人，让每个人都拥有自己的 AI 代理（AI agent），那么当某个恶意行为者试图作恶时，其他人的集体意志（collective will）便能够形成制衡并压制危险。Musk 借用丘吉尔关于民主的名言指出，分布式民主尽管不完美，却是防止专制的最优路径。
- 专有超级智能风险：最危险的情形是出现一个专有的、拥有自主意志或被少数几个人绝对掌控的超级 AI。
- 集体防御机制：抵御单点恶意滥用的最佳防线是让全人类共同掌握同等水平的技术，以群体合力对冲局部威胁。
### [47:16](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2836s) 2016 总统大选与宪政“大船小舵”论
话题转向当时的 2016 年美国总统大选，作为入籍公民的 Musk 对目前的政治局势感到无奈，直言：“我不认为这是我们民主制度的最佳时刻（I don't think this is the finest moment in our democracy）。”他表示自己目前选择置身事外，因为不确定仅凭个人影响力能对选举结果产生多大作用。
不过，Musk 对美国的制度设计表达了信心，他盛赞宪法制定者的高瞻远瞩：“我很庆幸宪法制定者确保了总统只是一艘大船上握着小舵的船长（being the US president is like being captain of a large ship with a small rudder）。”他解释道，由于这套制衡体系，任何一任总统实际上能带来的危害或好处都极其有限——就像若有某个能瞬间提振经济的按钮，每任总统都会以光速去按，但现实中并不存在这种魔法。面对主持人关于总统掌握核武发射权的质询，Musk 坚持认为制度和军方具备常识刹车机制，绝不会任由总统单方面、随意地发动先发制人的核打击。
- 权力的局限：总统职位受制于庞大的官僚体制与权力制衡，其改变国家航向的能力天然受限。
- 核武决策的防线：Musk 坚信军方在面临重大毁灭性指令时，会坚守原则并要求国会介入，绝不会盲目执行无端打击。
### [50:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3000s) Hyperloop 的提出与开源初衷
采访尾声，话题短暂转向超级高铁（Hyperloop）。Musk 澄清了外界对他参与程度的误解。他坦言自己几年前最初构想的一个版本后来被证明是错误的、在物理上行不通，但他当时已经对外放话称自己有了可行的方案。
在经过大量的推演和迭代后，Musk 终于拿出了一个在物理学上完全立得住脚（where the physics hangs together）的设计方案。考虑到自己同时管理特斯拉和 SpaceX 已经分身乏术，Musk 最终选择撰写并公开发布了白皮书，邀请全球有志之士免费采用这一构想进行落地研发。
- 物理可行性验证：Hyperloop 经历从最初错误想法到严密物理推导的迭代过程。
- 完全开源共享：Musk 将其作为开源构想发布，供全行业自由探索，自身并未将其作为全职商业项目来运营。
### [50:54](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3054s) 超级高铁（Hyperloop）的愿景与工程经济学
马斯克探讨了新型交通解决方案的价值，指出任何能够以更安全、成本更低、更便捷的方式将人们送达目的地的方案都极具意义。他直言，发表超级高铁（Hyperloop）白皮书最大的价值，在于打破现有的思维局限，激发人们去构想全新的交通体系，而不是把目光局限在修建普通的高铁上。他犀利地吐槽道：“那甚至都赶不上日本在80年代搞出来的水平，我真不明白那有什么意义。”他认为尤其在加利福尼亚州，人们理应去思考和发明远超现有水平的全新事物。
在面对是否投资了该领域相关公司的提问时，马斯克明确表示自己没有投资任何从事超级高铁研发的公司，一直努力保持中立态度，不偏袒某一家，只是向所有怀抱兴趣的团队提供道义上的支持，并真诚希望他们能够成功。目前他在超级高铁领域唯一参与的事务是举办大学生竞赛，旨在鼓励年轻人探索令人兴奋的新型交通架构。今年早些时候获胜的团队采用了一套与马斯克白皮书不同的悬浮机制——电磁悬浮，而马斯克原本提出的方案是利用压缩机将车头积聚的空气引向气垫滑板（Air Skis），在消除头部风阻的同时实现吊舱悬浮，该方案甚至已被证实可以在高达 1.1 马赫的超音速下正常工作。
```mermaid
flowchart TD
accTitle: 超级高铁设计哲学对比
accDescr: 对比马斯克的气垫轴承方案与电磁悬浮方案在管道建设成本上的差异
subgraph Cost["管道经济学：400英里双向 = 800英里管道"]
T["管道成本极其高昂\n是整个系统的核心优化指标"]
P["吊舱成本相对廉价"]
end
subgraph MuskSol["马斯克白皮书方案：气垫轴承 (Air Bearings)"]
Compressor["压缩机抽取车头空气"] --> Skis["流经气垫滑板消除阻力并悬浮"]
Skis --> CheapTube["管道端零复杂结构 = 管道极低成本"]
end
subgraph MagSol["学生优胜方案：电磁悬浮 (EMS)"]
MagTrack["需要在整个管道铺设电磁轨道/部件"] --> ExpensiveTube["管道造价急剧攀升"]
end
T --> CheapTube
T --> ExpensiveTube
```
马斯克深入解释了自己当初不推荐电磁悬浮的根本原因：这是基于严苛的经济学参数优化。如果修建一条 400 英里的双向超级高铁线路，总共需要铺设 800 英里的管道。由于吊舱本身造价低廉，真正决定项目生死的是那 800 英里管道的建造成本。只要任何技术方案对管道端提出了主动改造或铺设复杂设备的要求，管道造价就会不可避免地暴增；相反，利用气垫轴承则完全无需改造管道内壁，能最大程度压低基建成本。当被问及这套系统最终是否会成为现实时，马斯克展现了务实的态度：“我认为未来总会诞生类似的东西。”他表示，如果目前尝试商业化的公司最终未能做成，他未来可能会亲自下场打造一套原型演示系统，但他目前绝不会“抢跑”去和那些团队竞争，因为抛出免费开源概念却又亲自入场摘桃子绝非君子所为。
- **核心经济指标**：吊舱成本极低，管道建设成本极高，400 英里双向（即 800 英里）管道的每英里建造成本是首要优化参数。
- **气动方案验证**：气垫轴承方案在实验中已被证明在高达 1.1 马赫的速度下仍可稳定运行。
- **竞赛态度**：鼓励学生打破白皮书原有框架，开放探索包括电磁悬浮在内的不同架构。
---
### [55:02](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3302s) 当前科技版图与人才分布的失衡
卡拉·斯威舍将话题转向宏观科技格局，询问作为公认“远见者”的马斯克如何看待当前的技术发展态势，以及行业是否正在走向更加严肃的深水区。马斯克给出了相当积极的评价，认为技术世界并未陷入停滞或缺乏创新的低谷期，反而在众多前沿领域正涌现出惊人的突破，尤其是人工智能的演进步伐令人赞叹，基因技术的发展同样令人瞩目。
然而，在对全局创新保持乐观的同时，马斯克对硅谷的人才资源配置表达了微妙的担忧。他指出，当前的创业生态中存在人才分布不均的现象，有太多极具天赋的优秀创业者被过度吸引并集聚在纯互联网领域。马斯克直言：“从某种理想的分布来看，互联网领域聚集了过多的顶尖人才，若能将其中一部分天赋转移到其他实体与深科技行业中去，往往能创造出更大的社会价值。”
- **高速创新前沿**：马斯克重点列举了人工智能（AI）与基因工程（Genetics）作为最具突破性的两个技术领域。
- **资源错配担忧**：互联网应用和软件层吸收了过多的工程与商业天才，硬件、能源及重工业创新存在人才短缺。
---
### [56:29](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3389s) 超级人工智能危机：人类沦为“家猫”与神经织网（Neural Lace）的构想
马斯克随后引出了一个至关重要却极少有公司在严肃攻关的终极命题——“神经织网”（Neural Lace）。在重审 AI 风险时，他指出，只要假定人工智能保持任何持续的进步速率，人类的智力水平很快就会被远远甩在身后。即便迎来最温和、最仁慈的结局，人类面对超级智能也将彻底沦为下位物种。斯威舍随口打趣道就像变成家猫一样，马斯克立刻认同了这个比喻：“我可不怎么喜欢人类变成家猫的想法。”他强调，哪怕沦为被宠物般好生供养的家猫，也已经是最良善的设想，因此人类必须寻找与数字智能共存并保持主动权的根本途径。
在马斯克看来，最佳解决方案是在人类大脑之上构建共生的“第三数字层”。人脑结构本身就是由边缘系统（Limbic System）和大脑皮层（Cortex）协同构成的，正如大脑皮层能够与更底层的边缘系统紧密合作一样，一个架设在皮层之上的数字层也可以与人脑实现有机共生。马斯克指出，人类其实早已经变成了某种程度上的“生化人”（Cyborg）——每个人都在社交网络和电子邮件中拥有一个数字孪生体，手持智能手机的现代普通人所拥有的信息调用与瞬时通讯能力，甚至远远超越了 20 年前的美国总统。
```mermaid
flowchart LR
accTitle: 人脑三层架构与输入输出瓶颈
accDescr: 展示边缘系统、大脑皮层与神经织网构成的数字层，以及视网膜输入与拇指敲击输出之间的带宽失衡
subgraph BrainArch["大脑共生架构"]
direction TB
Limbic["边缘系统 (Limbic System)\n原始本能 / 情感反应"]
Cortex["大脑皮层 (Cortex)\n思考 / 理性分析 / 认知处理"]
Digital["数字层：神经织网 (Neural Lace)\n高速数字智能共生"]
Limbic <--> Cortex
Cortex <--> Digital
end
subgraph IOBottleneck["带宽约束对比 (I/O Bound)"]
direction TB
Input["信息输入 (Input)\n高带宽视觉界面 (双眼)\n每秒接收巨量数据流"]
Output["信息输出 (Output)\n极其受限 (双拇指屏幕打字)\n带宽低至数个字节每秒"]
end
IOBottleneck -.->|"核心痛点：解决 I/O 约束"| Digital
```
人类与数字智能融合的最大瓶颈在于输入/输出（I/O）极度受限，尤其是“输出受限”。我们的双眼构成了高带宽的视觉接口，能以极高速度摄取外界信息；然而在输出端，人类却只能依靠两根大拇指在手机屏幕上笨拙地敲敲打打，两者的带宽差异达到了数个数量级。马斯克断言：“有效实现与数字智能共生融合的核心，就在于消除这种 I/O 瓶颈。”而打破这一瓶颈的唯一手段，就是建立某种直接对接大脑皮层神经元的界面，即“神经织网”。
针对斯威舍关于这是否意味着开颅手术的疑虑，马斯克澄清并不一定需要打开头骨。神经元作为大脑内部的极度高能耗单元，需要极其充沛的血液滋养，因此人体的大脑动静脉本身就是一张通往全部神经元的现成网络。理论上，科学家完全可以通过颈静脉等血管途径将微型电极装置微创注入，沿着血管网络顺流定植到大脑各处神经元周围，无需实施传统开颅大手术即可建立起皮层级的超高带宽数据通道。
- **三层大脑模型**：边缘系统（情感/本能） + 大脑皮层（理性/智力） + 数字共生层（神经织网）。
- **I/O 速度断层**：人类属于典型的“输出受限型”物种，双眼接收信息属于超高带宽，而拇指打字或语言输出的带宽极其极其低下。
- **血管输送路径**：提出利用高血流量的动静脉血管网作为通道，通过颈静脉介入微创植入神经织网，规避开颅手术风险。
### [01:01:40](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3700s) 神经接口与 AI 带来的紧迫威胁
埃隆·马斯克（Elon Musk）接续关于人类未来地位的讨论，指出如果人类能建立一种高带宽的神经接口（high-bandwidth neural interface），有效地将肉体与自身的数字自我连接起来，人类就不会沦为人工智能面前的“宠物猫”（house cat）。卡拉·斯威舍（Kara Swisher）与沃尔特·莫斯伯格（Walt Mossberg）随即追问他本人是否打算亲自付诸实践，马斯克直言这必须有人去做：“有人必须去做这件事。我的意思是，如果没人做，那我想我可能就得去做。”（“Somebody's got to do it. I mean, I I I I mean, I So, somebody should do it. And I mean, if somebody doesn't do it, then I then I think I should probably do it, but uh”）。该设想的核心目标在于防止出现一个由少数人控制、力量远超人类且具有神级能力的外部 AI 系统。
莫斯伯格语带讽刺地调侃这场对话真是“令人心情愉悦”，斯威舍也提起访谈开头他们还在讨论小行星撞地球的危机。马斯克回应道，在与人类相关的时间尺度上，小行星撞击属于低概率的生存危机，而超级人工智能的发展则完全不同，它需要人类拿出极大的紧迫感去应对。
```mermaid
flowchart TD
accTitle: 神经接口与 AI 风险防范机制
accDescr: 展示马斯克通过建立高带宽神经接口来避免人类被集中化 AI 边缘化的逻辑链条。
subgraph Risk["潜在风险"]
A["中心化超级 AI<br/>（由少数群体掌控）"] --> B["人类沦为边缘存在<br/>（如宠物猫般无能为力）"]
end
subgraph Solution["应对路径"]
C["高带宽神经接口"] --> D["人脑与数字自我直连"]
D --> E["提升人类整体认知能力"]
end
E -->|"抵消垄断风险"| B
```
---
### [01:02:57](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3777s) 马斯克的日常减压方式、多任务处理与育儿生活
斯威舍顺势将话题转向轻松的个人生活，询问马斯克平时有什么真正的娱乐活动。马斯克坦言自己平时会和孩子们一起打电子游戏、看电影，过着相对普通的生活。现场一位观众提问，面对如此庞大的野心、繁重的工作以及高压对话，他是如何在日常生活中平衡工作与生活、承受巨大压力的。马斯克解释说，他的工作状态多数时候处于一种“工作分诊模式”（work triage mode）。只要 SpaceX 和特斯拉两家公司不同时爆发严重危机，局面就还算可控。他把成长型企业和准初创公司的波动比作正弦波（sinusoidal wave），只要两家公司的危机波峰不重叠在一起，压力就不至于失控；目前两家公司都在平稳推进，他的大脑中也装载了这两家公司的上下文，并能看清走向好结果的路径。
马斯克特别提到，无论工作多忙，他都会尽量留出时间陪伴孩子，因为孩子是他生活中最大的快乐来源：
- **幸福感来源**：“在我的生命中，比起任何事物，可以说孩子们带给我的快乐是遥遥领先的。”（“...of anything in my life, I would say kids by far make me the happiest.”）
- **相处模式**：孩子大约 99% 的时间让人感到幸福（剩下的 1% 另当别论）。由于孩子们往往沉浸在自己的小世界里，并不需要父亲连续几小时陪着聊天，因此马斯克经常与他们待在同一个房间内，一边处理工作邮件，一边随时准备回答孩子的提问。
- **家庭活动**：全家会一起旅行、打游戏。就在受访当周的周一，他们刚去了好莱坞环球影城新开的哈利波特乐园（The Wizarding World of Harry Potter），马斯克对乐园的建造质量赞不绝口，还特意夸赞那里的黄油啤酒（butterbeer）非常棒。
---
### [01:06:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=3960s) SpaceX 的火星商业生态与地火运输愿景
奥德赛（Odyssey）创始人埃文·伯恩斯（Evan Burns）在观众席提问：SpaceX 宣布过很多衍生业务（例如通过发射巨量卫星提供全球 Wi-Fi 覆盖），那么 SpaceX 未来的定位究竟是一个让其他企业开展业务的底层平台，还是一个自身孵化诸多业务线的大型集团？
马斯克阐明了 SpaceX 的根本战略：建立火星城市需要庞大的资金储备，因此 SpaceX 当前的策略是最大限度地获取地球上与火箭、航天器相关的收入，以此来资助火星运输系统的研发。一旦 SpaceX 拥有了将大量人员与物资送往火星的能力，它将成为火星商业生态的巨大推动者：
- **赋能火星创业**：火星上几乎一切都是空白，从建造第一座铁矿石精炼厂，到开办第一家披萨店，乃至创立地球上从未存在过的新型商业，都将大有可为。
- **类比联合太平洋铁路**：马斯克将 SpaceX 的角色类比为当年的联合太平洋铁路（Union Pacific Railroad）。当时许多人认为这条跨大陆铁路极其愚蠢，因为加利福尼亚根本没多少人居住；但只要建立了运输通道，繁荣的人口和商业便会随之而来。
- **航行时间预期**：针对现场开玩笑问从火星送一份披萨要多久，马斯克指出，目前完全有办法在 3 个月以内抵达火星；长远来看，航行时间有望压缩到 1 个月以内，尽管耗时越短技术难度会呈指数级上升。
---
### [01:08:40](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4120s) 交付延期挑战与 Model 3 的可制造性设计
科技媒体 The Verge 的尼莱·帕特尔（Nilay Patel）提出了一个尖锐问题：马斯克的雄心曾多次导致关键时间节点跳票，例如重型猎鹰火箭（Falcon Heavy）最初计划于 2012 年发射，Model X 也经历了漫长的延期。眼下大众消费级车型 Model 3 已经收取了大量普通消费者的定金，面对激进的生产进度表和巨额订单，特斯拉将如何确保按时交付？
马斯克表示，按期交付 Model 3 最核心的改进在于**面向制造的设计**（designing the car for manufacturing）：
- **历史吸取的教训**：在最初制造 Roadster 时，底盘和车身由莲花汽车（Lotus）负责，特斯拉仅负责动力总成集成；Model S 是特斯拉第一次完全独立造车，当时团队根本不懂什么是“为了易于制造而设计”，导致 Model S 的制造难度极大。随后的 Model X 虽然沿用了 Model S 的平台，却堆砌了大量过于炫酷的复杂新技术（wiz-bang technologies），使得制造难度雪上加霜。
- **Model 3 的制造哲学**：对于 Model 3，特斯拉采取了与 Model X 完全相反的策略——大幅简化整车结构，同时保留用户喜爱的核心特性。
- **设计与制造绑定**：团队在做出每一项设计决策时，都会严格权衡可制造性，确保零件能够在既定时间表内、以可承受的成本实现大规模量产。
- **供应链挑战**：马斯克补充道，全新汽车的制造极其困难，因为汽车是一个高度集成的产品，包含数千个独特的独立零部件，任何一个部件的供应链卡顿都会影响整车生产。
### [01:10:44](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4244s) 供应链的不可控风险与 Model 3 的设计定型
Elon Musk 详细探讨了制造业复杂的供应链管理挑战。他指出，如果算上二阶（Tier 2）和三阶（Tier 3）供应商，特斯拉的总供应商数量多达数千家，因此整条生产线的整体推进速度，“取决于最倒霉、最无能的那家供应商”。Musk 如数家珍般列举了供应商遭遇过的各种离奇灾难：工厂失火、地震、海啸、特大冰雹、龙卷风、货船沉没，甚至还有在墨西哥边境发生的枪战。在那次枪战事件中，运送后备箱地毯的卡车被打成了筛子，边境巡逻队（Border Patrol）因为车身布满弹孔而拒绝放行，导致生产线整整停工数天。Musk 笑着调侃：“我们只是想要后备箱地毯而已……里面又没有可卡因，就只是正经零件。”
为了应对这些不可控因素，特斯拉正在建立更高的冗余与自研能力，确保在供应商“无法或不愿”按时交付时能迅速在内部转为自行生产。关于备受瞩目的 Model 3，Musk 透露绝大部分设计已经完成，团队正全力争取在接下来的 6 周内实现“停笔定型”（complete pencils down）。针对外界各种酷炫新功能的构想，他明确表示这些都会留到第二版或第三版（Version 2、Version 3）中再行迭代。当被问及设定的交付时间表时，Musk 坦承自己可能有时略显不切实际，但强调自己每次给出的截止日期都是发自内心认同的真实目标，绝非刻意画饼。
- 供应商层级延伸至二、三级时，涉及的供应商总数达到数千家。
- 墨西哥边境枪击导致弹孔卡车被扣，曾致使 Model 3 后备箱地毯缺件并造成停产数天。
- Model 3 计划在约 6 周内彻底完成设计定型（pencils down）。
- Musk 确认 6 周后不会举行专门的自动驾驶产品发布会。
- “So things move as fast as the least lucky and least competent supplier.”（事情的进展速度取决于最倒霉、最无能的那家供应商。）—— Elon Musk
- “When I sort of cite a schedule, it is actually the schedule I think is true. It's not some fake schedule that I don't think is true.”（当我给出一个时间表时，那确实是我认为能实现的时间表，绝不是一个连我自己都不信的虚假日程。）—— Elon Musk
---
### [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) 模拟假说：我们处于“基础现实”的概率只有十亿分之一
现场观众 Josh 提出了一个关于文明与哲学的经典假说：一个足够先进的技术文明是否有能力创造模拟现实，而我们是否正生活在这样的虚拟模拟之中。Musk 坦言自己对这个问题思考了非常久，甚至一度到了走火入魔的地步——由于过去几乎每一次闲聊最终都会演变成关于 AI 和模拟现实的争执，他和弟弟 Kimbal Musk 甚至定下了一条硬性家规：严禁在泡热水澡（hot tub）时讨论模拟现实理论，因为这实在太破坏气氛了。
随后 Musk 给出了支持“我们生活在模拟之中”的最核心推论。他回顾道，40 年前人类只有像《乓》（Pong）这样由两个矩形和一个小光点组成的最简陋像素游戏；而 40 年后的今天，人类已经拥有了数千万人同时在线、画面逼真如实景的 3D 拟真游戏，并且虚拟现实（VR）和增强现实（AR）正在飞速发展。他推导指出，即使技术进步的速度放慢一千倍，只要时间尺度放大到一万年——这在宇宙进化尺度上根本微不足道——游戏就会变得与现实无法区分。考虑到未来将存在数以十亿计的计算终端运行这些以假乱真的模拟世界，从统计概率来看，我们处于最初“基础现实”（base reality）的几率仅有十亿分之一。此外 Musk 补充道，我们甚至应该庆幸并希望世界是一场模拟，因为如果文明停止进步，通常意味着发生了导致文明灭绝的灾难性事件。
```mermaid
flowchart TD
accTitle: 模拟假说逻辑推导流程
accDescr: 展示从早期视频游戏发展到未来拟真模拟并推导出基础现实概率的逻辑关系
subgraph TechEvolution["技术演进轨迹"]
A["40年前：Pong（两个矩形与一个像素点）"] --> B["现在：数千万人在线的光影逼真 3D 游戏"]
B --> C["未来演进：VR/AR 及完全不可区分的拟真现实"]
end
subgraph CivilizationalOutcomes["文明的终极分支"]
C --> D{"文明是否持续演进？"}
D -->|"持续发展"| E["数以亿计的模拟终端同时运行"]
D -->|"停止演进"| F["遭遇毁灭性灾难，文明灭绝"]
end
E --> G["统计概率：处于基础现实的几率仅有十亿分之一"]
```
- 40 年间视频游戏从《Pong》（双矩形与像素点）演化为写实级 3D 多人联机世界。
- Musk 和弟弟曾明令禁止在泡热水澡（hot tub）时谈论 AI 与模拟假说，称其“大煞风景”。
- 文明的未来在逻辑上只有两条路：要么研发出无法区分的虚拟模拟，要么在停滞中走向灭亡。
- “It would seem to follow that the odds that we're in base reality is one in billions.”（这样推导下来，我们身处真实基础现实的概率似乎只有十亿分之一。）—— Elon Musk
- “Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options.”（我们要么创造出与现实毫无二致的模拟世界，要么文明走向毁灭。摆在我们面前的只有这两种可能。）—— Elon Musk
---
### [01:17:56](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4676s) 完全自动驾驶的技术实现与监管审批时间表
来自 Box 公司的 Jeetu Patel 向 Musk 提出了两个尖锐问题：一是经监管审批且能在复杂城市路况下行驶的“完全自动驾驶”何时能够成为现实；二是距离人类驾驶汽车被立法禁止或因保险高昂而承担不起还有多久。Jeetu Patel 特别描绘了一个应用场景：一个完全不会英语、也不会开车的 75 岁老人，能否仅凭独自坐进车内就被安全运送到目的地。
Musk 斩钉截铁地表示，从技术研发层面看，自动驾驶“基本上已经是一个被解决了的问题”（basically a solved problem）。他分析称，在像北京这样极端拥堵的市区，自动驾驶反而相对容易，因为车流缓慢，系统只需与前车保持固定距离，且绝无超速撞击行人的风险；而在封闭的高速公路上，由于没有行人和对向交叉干扰，自动驾驶同样简单，现款的 Model S 和 Model X 已经在安全性上超越了人类司机。真正具有挑战性的是时速在 30 至 40 英里（约 48 至 64 公里）之间的城市混合复杂路段。Musk 给出明确预测：特斯拉在技术上实现“比人类更安全”的完全自动驾驶只需不到 2 年时间；但监管机构的批准预计至少还需要再等 1 年以上，因为不同国家和地区的监管部门必然要求积累数十亿英里的真实路测数据，以在统计学上证实其安全性显著优于人类驾驶。
- 城市严重拥堵路段因车速极低、制动余量大，对算法而言技术难度反而偏低。
- 最具挑战的场景是车速在 30 至 40 mph（约 48-64 km/h）的城市非封闭道路。
- Musk 预估完全自主驾驶技术（Full Autonomy）的工程成熟期在 2 年以内。
- 监管机构为放行完全无人驾驶，预计需要数百亿英里（billions of miles）统计级别的安全对照数据。
- “I really would consider autonomous driving to be basically a solved problem.”（我真的认为自动驾驶在本质上已经是一个被解决了的问题。）—— Elon Musk
- “I think we're basically less than 2 years away from complete autonomy.”（我认为距离实现完全自动驾驶，我们基本上只差不到两年的时间。）—— Elon Musk
### [01:20:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4841s) 自动驾驶的技术落地、监管门槛与人类驾驶权
马斯克继续深入探讨自动驾驶汽车的安全标准与普及时间表。他强调，自动驾驶在安全性上至少要达到人类驾驶员的两倍水平，甚至可能需要达到五到十倍的安全提升。而且，这一结论必须建立在具有统计学意义的数据集之上，也就是要在各种复杂多变的路况与行驶环境中累积数十亿英里的测试里程。基于此，他给出了技术与法律监管的两个时间预期：从技术实现的角度来看，完全自动驾驶在不到两年内就能成真；但要跨过监管门槛并获得正式批准，大概还需要三年的时间。
现场有人顺势追问，未来是否会出现彻底立法禁止人类手动开车的那一天。马斯克明确表示反对这种剥夺驾驶权的做法，他阐述道：“我们生活在一个民主社会，所以这大概取决于全体民众的决定。我的意思是，我并不赞成禁止人们开车。我支持自由，反对限制人们的行为。”不过，他预测获取驾照的门槛可能会大幅收紧。在他看来，提高考试要求是一个合理的方向——未来如果一个人想要手动驾驶汽车，就必须证明自己具备远高于现行标准的驾驶技能。
- 自动驾驶安全基准：需达到人类安全水平的 2 倍至 5 到 10 倍，并基于数十亿英里的多样化路测数据进行统计验证。
- 落地时间表预估：技术上可行预计在 2 年以内；通过监管审批预计需要大约 3 年。
- 人类驾照未来趋势：不应剥夺驾驶自由，但手动驾驶牌照的考核与准入门槛将变得更加严格。
---
### [01:22:03](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4923s) 火星文明的政体构建：直接民主与对抗法律惯性
在全场访谈进入尾声之际，主持人开放了最后一个提问机会。一位观众再次将目光投向火星生活，询问在火星拓荒初期，应该如何考虑文化融合、治理制度以及法治规则的建立。马斯克先是开玩笑地打趣说：“我想我就直接宣布自己是火星之王，然后就去统治吧。”引得全场哄堂大笑。
玩笑过后，马斯克严肃地分享了他对火星政治架构的深层构想。他认为，火星最有可能采取的政体是直接民主制，而非代议制民主——让公民直接就各项议题进行投票。马斯克解释说：“我认为直接民主可能更好，因为与代议制民主相比，直接民主中发生腐败的可能性会大幅降低。”
为了防止人类社会常见的“法律堆积与官僚化”弊病，马斯克特别提出了一套对抗“法律惯性”（inertia of laws）的立法框架建议：废除法律应当比制定法律更容易。他解释道，法律一旦通过往往具有无限期的生命力，除非被刻意废止。因此，他建议设置不对称的投票门槛以及自动失效机制：
```mermaid
flowchart TD
accTitle: 火星立法与废除机制
accDescr: 马斯克构想的火星法律制定、日落条款及废除流程图
subgraph Legislate ["立法阶段"]
A["提出新法律草案"] --> B{"投票赞成率 ≥ 60%?"}
B -->|"是"| C["法律正式生效并附带日落条款"]
B -->|"否"| D["草案不予通过"]
end
subgraph Repeal ["存续与废止阶段"]
C --> E{"公民动议废除: 赞成率 > 40%?"}
E -->|"是"| F["法律立即废除"]
E -->|"否"| G["法律继续生效至到期"]
G --> H{"日落期满: 重新投票达到 60%?"}
H -->|"否"| F
H -->|"是"| C
end
```
- 不对称门槛：制定一项新法律可能需要 60% 的赞成票，但任何时候只要有超过 40% 的人投票反对，就可以将该法律废除。
- 强制日落条款：任何法律必须自带日落条款（Sunset Provision）；如果一项法律到了期限不足以让大家再次投票赞成它，那它就不配继续留存在法典之中。
马斯克总结道，这就是他为火星政府设想的蓝图：一个推行直接民主、制定法律比废除法律更难、且法规不会自动永久生效的社会。主持人卡拉·斯威舍与沃尔特·莫斯伯格以一句幽默的“你会成为一个好国王的”作为总结，并在全场的热烈掌声中向埃隆·马斯克致谢，结束了本次访谈。
## Notable quotes
> [11:39](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=699s) 生活不能仅仅是为了解决一个又一个问题，还必须有那些令人振奋和激动的事情，让你发自内心地为活着而感到高兴。 — Elon Musk
> [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) It would seem to follow that the odds that we're in base reality is one in billions. — Elon Musk
> [40:41](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=2441s) freedom consists of the distribution of power and despotism in its concentration — Elon Musk
> [01:33](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=93s) 在太空中根本不存在所谓的脱离高度，只有脱离速度。 — Elon Musk
> [01:22:03](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4923s) 我认为直接民主可能更好，因为与代议制民主相比，直接民主中发生腐败的可能性会大幅降低。 — Elon Musk
> [01:14:00](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4440s) Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options. — Elon Musk
> [26:05](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=1565s) 我认为目前为止没有任何一家汽车公司造出了一款真正优秀的电动汽车。 — Elon Musk
> [01:10:44](https://www.youtube.com/watch?v=wsixsRI-Sz4&t=4244s) So things move as fast as the least lucky and least competent supplier. — Elon Musk
## Mentioned
Elon Musk, Kara Swisher, Walt Mossberg, Jeff Bezos, Matt Damon, Kimbal Musk, Nilay Patel, Jeetu Patel, Evan Burns, SpaceX, Tesla, OpenAI, Google, Apple, NASA, Volkswagen, Fiat, Lotus, Box, Odyssey, Falcon 9, Falcon Heavy, Dragon 2, Saturn V, Model S, Model X, Model 3, Roadster, Gigafactory, Hyperloop, Neural Lace, Iridium NEXT, Pong, The Martian
## Video description
SpaceX CEO Elon Musk talks with Recode's Kara Swisher and The Verge's Walt Mossberg about his plans to send a one-way rocket to Mars in 2018. He estimates colonists could start arriving on the Red Planet by 2025. Musk also talks about the proliferation of electric vehicle initiatives that compete with his other company, Tesla, and why autonomous cars will become the norm. He says he doesn't see Google as a competitor, but that "Apple will be more direct." Plus: Why Musk wants more people to have access to the power of artificial intelligence.
[Source](https://www.youtube.com/watch?v=wsixsRI-Sz4)
Source quotes
[11:39] Elon Musk: 生活不能仅仅是为了解决一个又一个问题，还必须有那些令人振奋和激动的事情，让你发自内心地为活着而感到高兴。
[74:00] Elon Musk: It would seem to follow that the odds that we're in base reality is one in billions.
[40:41] Elon Musk: freedom consists of the distribution of power and despotism in its concentration
[1:33] Elon Musk: 在太空中根本不存在所谓的脱离高度，只有脱离速度。
[82:03] Elon Musk: 我认为直接民主可能更好，因为与代议制民主相比，直接民主中发生腐败的可能性会大幅降低。
[74:00] Elon Musk: Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options.
[26:05] Elon Musk: 我认为目前为止没有任何一家汽车公司造出了一款真正优秀的电动汽车。
[70:44] Elon Musk: So things move as fast as the least lucky and least competent supplier.
Source quotes
[74:00] Elon Musk: Either we're going to create simulations that are indistinguishable from reality or civilization will cease to exist. Those are the two options.
[11:39] Elon Musk: Life can't be just about solving problems. Like there have to be things that are inspiring and exciting that make you glad to be alive.
[77:56] Elon Musk: I really would consider autonomous driving to be basically a solved problem.
[40:41] Elon Musk: Freedom consists of the distribution of power and despotism in its concentration.
[23:28] Elon Musk: It's going to be odd to have a car without autonomy in the future.
[70:44] Elon Musk: So, things move as fast as the least lucky and least competent supplier.
[47:16] Elon Musk: I’m glad that the framers of the Constitution saw fit to ensure that the president was someone who was captain of a large ship with a small rudder.
Original transcript
[0:00] Elon, let me start by saying we're very
[0:02] glad you're here safe and sound. Uh
[0:04] thanks. I actually had to sorry for
[0:06] being late. I was uh
[0:07] uh flying I flew here with the landing
[0:09] gear down cuz there was like some kind
[0:10] of landing gear issue. Landing gear was
[0:12] stuck?
[0:13] There's some kind of warning light uh my
[0:16] pilot said that if they were to retract
[0:18] the landing gear it may not go back down
[0:19] again. So
[0:21] So this happens to you this even happens
[0:23] to you.
[0:24] Yeah. Yeah. Yeah. Okay. I have landing
[0:26] gear problems of various kinds. Um so um
[0:30] anyway we're happy to wait for you and
[0:32] then we're glad you got here safe.
[0:33] Thanks for having me. It's great to see
[0:34] you guys. Thank you for coming. I really
[0:35] appreciate it. You kept your promise
[0:37] which was nice. I think you were drunk
[0:39] when you promised me but that's okay.
[0:41] I'll take it. So you know in the couple
[0:43] of years since you've come you've done
[0:46] some astonishing things in in terms of
[0:48] substance of stuff with your with both
[0:50] of your companies. Um let's start
[0:52] talking about space and what you've been
[0:54] doing there. Um obviously you've had
[0:56] some success uh landing the landing the
[0:58] rocket. You've had you know you've done
[1:00] a bunch of other things that where
[1:01] people thought you weren't going to be
[1:02] successful. Talk a little bit about sort
[1:05] of the progress you think you've made
[1:06] with SpaceX. Sure well I mean there's a
[1:08] lot of things where I I think I I didn't
[1:10] think we'd be successful. Um
[1:11] so the
[1:13] the um
[1:14] probably and the most significant thing
[1:15] is being able to land an
[1:18] an orbit class rocket uh booster stage
[1:21] um
[1:22] and uh and bring it both back to Cape
[1:25] Canaveral uh land to land and be able to
[1:27] land on a drone ship out in the ocean.
[1:31] Um
[1:32] the
[1:33] um it it there is a bit of an education
[1:36] process that's needed to understand
[1:37] orbital dynamics um cuz
[1:40] a lot of people can are confused of like
[1:42] why the heck are you landing a ship uh
[1:44] landing a rocket in a on a ship in the
[1:46] ocean? That seems pretty inconvenient.
[1:48] Um
[1:49] and the the reason is because that uh
[1:52] going up and staying up is actually
[1:54] about velocity horizontal to the earth's
[1:56] surface.
[1:58] So
[1:59] um there there's a huge difference
[2:01] between space and or space and or and
[2:03] orbit. It like space you can think of as
[2:05] like say being the international waters
[2:08] boundary for the Pacific Ocean. Like if
[2:10] you go you know
[2:11] uh 100 miles offshore you're technically
[2:13] out of
[2:14] Yes.
[2:15] coastal waters now you're in the
[2:16] Pacific. So it's like technically you're
[2:18] in the Pacific but but it's but orbit is
[2:20] like circumnavigating the globe.
[2:23] It it's a really giant difference and
[2:25] the the reason that things go up and
[2:28] stay up is because
[2:30] you're you're zooming around the earth
[2:31] so fast that your outward radial
[2:33] acceleration is equal to the inward
[2:35] acceleration of gravity.
[2:37] And so those balance out and you have a
[2:39] net zero gravity.
[2:41] So when you see the space station
[2:43] the thing that's little little sort of
[2:45] um counterintuitive is that the space
[2:47] station is actually zooming around the
[2:48] earth at 17,000 miles an hour.
[2:52] Even though it seems like it just
[2:53] really still you know but it's moving
[2:56] really really fast. Um
[2:58] I mean to put that into perspective
[3:01] um a bullet from a 45 um
[3:04] gun you know a handgun um is is is is uh
[3:08] just below the speed of sound. So
[3:10] the space station is going more than 25
[3:13] times faster than that.
[3:16] Um and that's what's needed actually to
[3:17] go up and stay up. Um and that's why
[3:19] that's why there's the term escape
[3:21] velocity not escape altitude. There's no
[3:23] such thing as an escape altitude there's
[3:25] only escape velocity. You need to be a
[3:27] certain speed to escape the gravity of
[3:29] the earth. Yeah you can think of gravity
[3:30] as kind of a funnel in space time.
[3:33] Um so uh
[3:35] or think of it like a coin funnel. Like
[3:37] it's it's really it's very much like
[3:38] that in in you know but it's obviously a
[3:40] sort of a four-dimensional coin funnel
[3:42] but uh
[3:44] if if you if you spin a spin a marble or
[3:46] a coin on a coin funnel the
[3:50] when it's when it's far out it sort of
[3:52] spins slowly and then as it gets closer
[3:54] it spins faster and faster
[3:56] and if you want if you want if you were
[3:57] to start at the bottom of the coin
[3:59] funnel and you wanted to to to to to to
[4:01] exit you'd spin it horizontally and it
[4:04] would it would spin out and and and
[4:07] that's really how you
[4:08] how you get to orbit. Um
[4:11] Yeah so the gravity well is like a
[4:13] funnel.
[4:13] why you want to land on the on a ship in
[4:15] the ocean?
[4:17] Uh because
[4:18] um in order to get to orbit you all that
[4:20] matters is your horizontal velocity.
[4:22] Your altitude is doesn't doesn't really
[4:23] matter.
[4:24] Um in fact the the um
[4:28] the force of gravity at uh say
[4:30] the
[4:31] the nominal
[4:33] um boundary of space 100 km is almost
[4:35] exactly the same as it is on the surface
[4:37] of the earth.
[4:38] Hm.
[4:39] Um it's it's like if it's a few percent
[4:41] lower than than the surface of the
[4:42] earth.
[4:43] Um
[4:45] uh so in in order to go up and stay up
[4:48] the only thing that matters is how fast
[4:49] are you going horizontal to the earth's
[4:50] surface.
[4:52] So you have that outward radial
[4:53] acceleration or think of it like maybe
[4:54] like tetherball or something like that.
[4:56] It's really that outward acceleration is
[4:58] the thing that matters. Um and so when
[5:00] the rocket is going to orbit um
[5:03] the only reason it's going up is to get
[5:05] out of the thick part of the atmosphere
[5:06] cuz that
[5:07] at high velocity the atmosphere is thick
[5:09] as molasses.
[5:10] Um
[5:12] and so it goes up very briefly but if
[5:13] you look at a long exposure of the the
[5:16] rocket's uh trajectory you'll see it it
[5:19] goes up but immediately curves over and
[5:21] starts going horizontal.
[5:23] Um
[5:24] and so
[5:25] the um
[5:27] at at the at the point at which the
[5:29] uh
[5:31] the uh the the point at which the stage
[5:33] is separate those two stages
[5:35] um the the primary booster stage which
[5:37] is the most expensive part of the rocket
[5:39] the point at which that that staging
[5:40] occurs uh can be um as high as uh Mach
[5:44] 10.
[5:46] Um but it's it's so it's going away from
[5:48] the launch site at 10 times the speed of
[5:50] sound.
[5:52] So in in in order to get back to the
[5:55] launch site you would have to have
[5:56] enough I
[5:57] fuel and oxygen to reverse out that
[6:00] velocity
[6:02] and and and boost back all the way to
[6:04] the launch site.
[6:05] Um and you just don't have the physics
[6:07] of it don't really allow you to have
[6:09] that much it's it's not about saving
[6:11] money on fuel or anything it's just
[6:13] physically impossible.
[6:15] Um
[6:16] so um
[6:18] because another sort of thing about uh
[6:21] if you're if you're in space is that
[6:22] there's nothing to react against. So
[6:24] like whereas an aircraft can can circle
[6:27] very easily because it's reacting
[6:29] against air.
[6:31] In vacuum there's nothing to react
[6:33] against. So the only way to go back the
[6:35] other direction is to apply just as much
[6:37] energy as it took you to go it we if you
[6:39] want to go backwards you have to apply
[6:41] just as much energy as it took you to go
[6:42] forwards.
[6:43] In fact it would twice as much really
[6:45] cuz you got to zero it out and then
[6:47] you've got to You've got to land
[6:48] elsewhere.
[6:49] Yeah so bottom line is this thing is
[6:50] zinging out to
[6:52] zinging out to
[6:52] the point it may well be over the ocean
[6:53] cuz the ocean covers most of the
[6:55] ocean covers most of the uh the earth.
[6:56] So that's the point
[6:57] it may Oh it's it's it's actually the
[7:00] point of separation it's not that far
[7:01] away. It's maybe 100 km away from the
[7:04] the launch site but it is going like
[7:07] hell in the opposite you know away from
[7:09] the launch site.
[7:10] So
[7:11] the
[7:12] the only way to really land it is to
[7:14] have it continue on that arc that
[7:17] ballistic arc and then land far out to
[7:20] sea on a ship that's that's
[7:22] pre-positioned to a particular
[7:25] latitude and longitude very very precise
[7:27] to within about a meter. Um
[7:30] and then
[7:31] the the rocket will um then go from
[7:33] vacuum through rarefied air at
[7:36] hypersonic velocity uh
[7:39] um it and and what so when it's when
[7:40] it's in vacuum it has to obviously you
[7:42] can't use air surfaces you have to use
[7:44] um nitrogen jets to control the um
[7:48] the attitude and position.
[7:50] And then um as it starts to encounter
[7:54] uh the air um we use um
[7:57] grid fins because grid fins
[7:59] uh look like sort of like a waffle um
[8:02] they work quite well across a wide
[8:04] regime from both very high velocity um
[8:07] hypersonic through supersonic transonic
[8:10] and subsonic. Um so it's hard to it's
[8:12] it's hard to have air surfaces that work
[8:14] well across that entire regime.
[8:17] And then uh
[8:19] so once the air air
[8:20] forces become high it uses the um the
[8:23] four grid fins to to sort of control its
[8:26] attitude
[8:26] itself. Yeah it's it's controlling its
[8:29] it's it's controlling pitch yaw and roll
[8:31] with with the grid fins.
[8:33] Um
[8:34] and uh and then once and those grid fins
[8:37] will then
[8:38] position it to where it's fairly close
[8:40] to the ship and then it will light in
[8:43] this case three of the nine engines to
[8:46] arrest the velocity and then drop to one
[8:48] engine for precision right before
[8:50] landing. Right. Okay. So that was a
[8:53] maybe a bit of a long explanation.
[8:54] but okay what we're going to get to is
[8:57] that's super [ __ ] hard.
[8:59] There's a video. So why why is that
[9:01] important? Why has that this moment been
[9:03] important for you?
[9:05] Um
[9:08] well so in order to reuse
[9:11] the the booster stage which is about 70%
[9:14] of the cost of the rocket so that Which
[9:15] cost is that? How much is that?
[9:17] Um
[9:19] well I mean it's sort of on the order of
[9:21] 30 to 35 million dollars. Right. So you
[9:23] want to save that. Yeah I mean it's like
[9:25] I try to I tell my team it's like
[9:26] imagine there was a pallet of cash that
[9:28] was
[9:29] falling through the atmosphere
[9:33] and it was going to burn up and smash
[9:35] into tiny pieces. Would you try to save
[9:37] it? Right. Right. Right. Probably yes.
[9:39] Yes. Okay. Yeah that sounds like a good
[9:41] idea. Right. Okay. Uh-huh. Um
[9:44] so
[9:45] so yeah so we we we
[9:46] we want to get it back and that way um
[9:50] we don't have to make another one.
[9:52] Right. Um and I think it's quite tragic
[9:54] if rockets like get smashed into tiny
[9:56] pieces and land at the bottom of the
[9:57] ocean.
[9:57] a question? We we've been in we've been
[9:59] going to space for
[10:01] uh what, 50 years or something like
[10:04] that?
[10:04] Yeah.
[10:05] Nobody, until you started doing this and
[10:08] Jeff Bezos' company has done it, uh the
[10:10] government never thought of saving the
[10:12] rockets.
[10:13] They never saved the pallet of cash.
[10:16] Why not? And the Russians didn't either.
[10:18] I mean, what what is the deal there?
[10:20] yeah, I mean there was some attempt uh
[10:21] made to do that with the the space
[10:23] shuttle but there was no um
[10:25] return. It's the first time that that a
[10:27] rocket booster has returned to launch
[10:28] site um
[10:30] from an orbital mission and and
[10:31] certainly the first time that there's
[10:32] been a a landing on a ship out at sea.
[10:34] regular rockets that went up that
[10:36] weren't designed like planes never tried
[10:39] to do this. Right. Um the plane thing is
[10:43] not not a good idea in my view. Um
[10:46] the
[10:47] so so the the the plane um
[10:50] and the the reason I think it's
[10:51] like intuitively it seems like a plane
[10:53] should work but but actually if you you
[10:55] consider that really every mode of
[10:56] transport has a design that is
[10:58] appropriate to its medium. Um and if
[11:01] you're in
[11:02] space, um wings are not very useful.
[11:06] Um cuz there's no air.
[11:08] Uh
[11:10] and
[11:11] and and then if you want to go somewhere
[11:13] other than Earth, there's also no
[11:14] runways.
[11:16] Uh-huh. So, this is these are important
[11:18] considerations.
[11:20] Um
[11:23] so that's why when they went to the
[11:24] moon, they used propulsive landing.
[11:26] Right. But what I'm saying was when they
[11:27] built the space shuttle, it sort of was
[11:29] like a looked like a
[11:31] kind of bulbous plane.
[11:32] appealed to Congress. Yeah. Yeah.
[11:34] Yeah.
[11:36] They went cool. It's cool. Yeah, looked
[11:38] like an airplane.
[11:39] Can you explain the the you know, Jeff
[11:41] Jeff Bezos Good one. Jeff Bezos was here
[11:44] last night and I asked him, what's the
[11:45] difference between what you're doing and
[11:47] what Elon Musk is doing and he said,
[11:49] well, I think we have
[11:51] I think he used the word like-minded
[11:54] in the general sense of it and then he
[11:56] went on to explain some differences. Do
[11:57] you
[11:59] And then he but he talked about and
[12:02] correct me if I'm misquoting him but I
[12:04] think he was saying,
[12:06] we're this is all about laying the
[12:07] foundations of being able to do greater
[12:11] things by getting the basic
[12:13] infrastructure of being able to reuse
[12:15] these rockets down.
[12:17] Right. Do you Is that correct? Do you
[12:20] have a similar starting point from him
[12:23] in your thinking process? I think
[12:25] there's certainly some similarities of
[12:26] opinion. Um I think both uh Jeff and I
[12:28] believe that it's important for the
[12:30] future to be a spacefaring civilization
[12:33] um and actually ultimately be out there
[12:34] among the stars and I think that's the
[12:36] that's the exciting, inspiring future
[12:38] that
[12:39] I think I think certainly people in this
[12:41] room are wanting and anyone
[12:43] after seeing that the asteroids are
[12:45] going to destroy this planet. I mean, Um
[12:47] I mean, I don't view it as um you know,
[12:50] we want we
[12:52] I mean, I I think I think what when I
[12:53] say, you know, multi-planet
[12:55] species, like that's really where we
[12:56] want to be. It's not like, you know,
[12:59] still being a single-planet species with
[13:00] moving planets. It's uh it's really
[13:02] being a multi-planet species um
[13:05] and having civilization and life as we
[13:07] know it extend beyond Earth to the rest
[13:09] of the solar system and ultimately to
[13:11] other star systems. Um I think that's
[13:13] the thing that that's the that's the
[13:15] future that's exciting and inspiring and
[13:16] I think that's what, you know, I think
[13:19] you know,
[13:20] you you need kind of you need things
[13:21] like that to make to to to be glad to
[13:23] wake up in the morning. You know, the
[13:24] like life life can't be just about
[13:26] solving problems. Like there have to be
[13:27] things that are inspiring and exciting
[13:29] that make you glad to be alive. So, what
[13:32] in the immediate time frame, what is
[13:34] what is your goal for SpaceX now that
[13:36] you've done this, which is a huge
[13:38] accomplishment? What is the
[13:39] plan for you in the immediate time and
[13:42] then the longer range?
[13:44] Sure. So, the
[13:45] um so we plan to re-fly
[13:49] uh one of the landed um
[13:52] rocket boosters hopefully in about
[13:56] two or three months, something like
[13:57] that.
[13:58] Um
[13:59] and um
[14:01] and then that that so that'll be an
[14:03] important milestone. Um so far the the
[14:05] the stages are looking like quite quite
[14:07] good uh even though they come through
[14:08] through quite there's a really difficult
[14:11] entry re-entry situation. Um but they're
[14:14] they're looking in fact they're in
[14:15] they're in good shape.
[14:17] Um we now have four of them. Um so, we
[14:20] want to start re-flying them
[14:21] um
[14:22] you know, towards the end of summer
[14:24] um
[14:25] and then uh hopefully by the end of this
[14:28] year we'll be launching Falcon Heavy uh
[14:30] which will will be the um the most
[14:32] powerful rocket
[14:34] uh in the world by more than a factor of
[14:35] two.
[14:36] So, Falcon Heavy is will be on the order
[14:39] of 5 million pounds of thrust on
[14:40] lift-off, which is about 2/3 the size of
[14:42] a Saturn V.
[14:44] Oh, really? Yeah, so it's That's the
[14:46] rocket that uh took the astronauts to
[14:48] the moon. Right, exactly. So, In fact,
[14:50] we're launching from the same from the
[14:51] same pad. From Very same pad.
[14:53] from the Apollo 11 pad. Wow. Yeah.
[14:55] That's amazing. So, you you're hoping to
[14:57] launch that Falcon Heavy by the end of
[14:59] this year, you said.
[15:00] Yeah, that's that's our aspiration. Is
[15:01] that Now, that's somewhat of a delay
[15:03] from when you first hoped to launch it,
[15:05] right? Um yeah, um but uh
[15:09] the I mean, it's not like we had a lot
[15:10] of pressing customers who wanted us to
[15:12] launch it.
[15:12] Uh-huh. Okay.
[15:14] So,
[15:15] the
[15:17] uh in fact, the first launch will will
[15:18] not have any operational satellites.
[15:20] It'll be a demonstration launch. Mhm. Um
[15:22] and the first operational flights where
[15:24] where customers actually want us to
[15:26] launch it are next year.
[15:28] You know, whereas there's there's a lot
[15:29] of customers who want us to launch uh
[15:32] Falcon 9. Um
[15:34] so, about about a quarter of our launch
[15:35] of our flights are for uh for NASA um
[15:37] but 3/4 are for uh commercial satellites
[15:40] like broadcasting communication
[15:41] satellites um or science missions for
[15:44] other countries. Um
[15:46] and um
[15:48] and this is there's quite a quite a
[15:49] backlog and we had we had an issue with
[15:51] a rocket last year so that
[15:53] um put about a 6-month hold on our
[15:54] schedule so we're sort of backlogged on
[15:56] our launches and we're trying to get
[15:57] them out as as as quickly as we as we
[16:00] can um and
[16:02] you know, so
[16:03] you know, service our customers.
[16:05] The the uh so we're we're the launches
[16:07] will take place you know, every two to
[16:09] four weeks. It's quite a quite a high
[16:11] launch cadence.
[16:11] That's a much faster cadence than NASA
[16:13] had, right? Um yeah, it's I mean, it's
[16:16] it will be more launches than any
[16:18] anything else in the world. Um so, more
[16:20] than Russia, more than Europe, more than
[16:23] well, more than China by now. Sure,
[16:25] certainly.
[16:26] to deliver customers.
[16:28] Satellites.
[16:29] Yeah, it's it's um
[16:30] there's a lot of broadcasting
[16:31] communication satellites that are going
[16:33] to geosynchronous orbit. Mhm.
[16:36] And um
[16:37] and then there's we're we're we'll also
[16:39] be launching the new uh Iridium
[16:42] constellation. So, the Iridium's got a
[16:44] next generation uh constellation of uh
[16:47] satellites, I think 60 or 70 satellites,
[16:49] quite you know, decent-size satellites
[16:51] uh um
[16:52] that that'll be
[16:54] like many orders of magnitude
[16:55] improvement over the current Iridium
[16:56] system. So, we'll be able to have global
[16:58] broadband um
[17:00] so that that'll be a whole bunch of
[17:01] launches and um
[17:04] yeah, and then and then next year we'll
[17:06] be flying
[17:08] um Dragon version two, which is the one
[17:11] that's capable of taking up to seven
[17:13] astronauts to the space station. Mhm. Um
[17:16] in fact, Dragon two really is
[17:19] it's a propulsive lander as well um and
[17:22] it'll be the uh it's it's it's intended
[17:25] to carry astronauts to the space station
[17:27] but it's also capable of being a general
[17:30] science delivery platform to anywhere in
[17:32] the solar system.
[17:34] So, um So, where are you going with it?
[17:37] We're going to we're going to send one
[17:38] to Mars in 2018. Okay.
[17:40] Now, let's talk
[17:41] Wait. Wait. 2018, that's for sure. Yeah,
[17:43] a couple years. Couple years. Now, will
[17:45] you be on that flight?
[17:46] No. You have talked about this. You said
[17:49] you don't want to you want to die on
[17:51] Mars, just not on landing, right? Is
[17:52] that correct?
[17:54] Well, I mean, I think if you're going to
[17:55] choose a place to die,
[17:57] then Mars is probably
[17:59] you know, not a bad choice. All right.
[18:02] Um
[18:03] But you're not ready
[18:04] it's not some sort of Martian death wish
[18:05] or something.
[18:07] But but yeah, I mean, if you're going to
[18:09] be born on Earth, die on Mars. So,
[18:11] sending this up to Mars, 2018, right?
[18:14] Sending this up to Mars, 2018.
[18:17] When will someone like you get there
[18:19] from your plans? Sure. So, so the 2018
[18:22] mission would be um our Drag- Dragon
[18:24] version two.
[18:25] Right. Um and
[18:27] that um
[18:29] I wouldn't recommend traveling to Mars
[18:31] in in that cuz I mean, it has the
[18:32] interior volume of a large SUV. Okay. Uh
[18:35] the the the trip the trip for Dragon
[18:38] would be on the order of six months.
[18:39] Mhm. It's a long time to spend in an
[18:41] SUV, I think. Mhm. Um
[18:44] That's yes,
[18:45] can be done. Can be done but not not
[18:48] probably not
[18:49] ideal. Um
[18:51] and it also doesn't have the capability
[18:52] of getting back to Earth. Right. That's
[18:57] That seems more important than the space
[18:59] is.
[18:59] Hopefully. Yeah, we can put that in the
[19:00] in the fine print, Yeah.
[19:02] Sorry. We were rooting for It's It's
[19:03] like the side effects in a drug ad.
[19:05] By the way, cannot get back to Earth.
[19:08] Yeah. Yeah, we saw the movie. We saw
[19:09] what happened. He got back. Yeah, yeah.
[19:12] Um
[19:13] So, it's good. I I I actually enjoyed
[19:14] the movie. Um
[19:17] So,
[19:17] he could have gotten back like that?
[19:19] That was that plausible?
[19:19] Well, I thought there was some, you
[19:21] know, connection. It was It was mostly
[19:23] It was like 80% scientifically correct.
[19:26] Um but they did connect a series of
[19:28] improbable events.
[19:29] Such as?
[19:31] Well, I mean, I don't think you can sort
[19:32] of just uh take off from Mars um on an
[19:35] unguided rocket really and and then
[19:38] prick your finger on a space suit and
[19:40] navigate to a to a spaceship.
[19:42] Right?
[19:45] Yeah.
[19:46] Not Not Not impossible, just extremely
[19:48] unlikely.
[19:50] So, the sand storm in the movie
[19:52] if you're Matt Damon maybe. Maybe. You
[19:54] have some mad skills for sure.
[19:56] So so so when will people like yourself
[20:00] get there and I assume you'll be first
[20:01] in line for that? Yeah so
[20:04] later this year in September at the
[20:07] IAC which is the big
[20:10] sort of world space conference industry
[20:11] space conference. I'm going to be
[20:12] presenting the
[20:14] the architecture for Mars colonization.
[20:17] Mhm.
[20:19] So I think what really matters is being
[20:20] able to transport large numbers of
[20:22] people and
[20:25] ultimately millions of tons of cargo to
[20:26] Mars.
[20:29] And and that's what's necessary in order
[20:30] to create a self-sustaining
[20:35] not really self-sustaining but a growing
[20:39] city on Mars. So I'm curious have you
[20:41] been to space yet? No. Why?
[20:44] I mean you could just go up right for a
[20:45] little bit or not?
[20:46] I could I suppose. Yeah. Why haven't
[20:48] you? Like walked around or something?
[20:51] Yeah well yeah probably will. Will you
[20:52] do a moon test before you
[20:55] go to Mars?
[20:57] Yeah I'm going to probably probably I
[20:58] don't know go to orbit in four or five
[21:00] years or something like that. But again
[21:01] space and orbit are very different
[21:02] things. But on the Mars thing would you
[21:05] send up
[21:07] two or three
[21:09] whether it's you or not. I kind of would
[21:11] prefer if you tried it frankly but
[21:15] because it would be exciting but
[21:18] would you send up some people some
[21:20] people before you do this whole
[21:21] architecture for colonizing Mars just a
[21:24] handful of people to kind of see what I
[21:26] mean the the the basic game plan is like
[21:28] we're going to
[21:29] send um
[21:31] a mission to Mars with every Mars
[21:33] opportunity from 2018 onwards. So
[21:36] and they occur approximately every 26
[21:38] months. So um
[21:40] you know we we're establish we're
[21:42] establishing cargo flights to Mars that
[21:44] people that people can count on for for
[21:46] cargo. Um
[21:48] and it's like I said the the the
[21:50] Earth-Mars orbital rendezvous is only
[21:52] every 26 months. So there's one in 2018
[21:54] there'll be another one in 2020.
[21:56] Um
[21:57] and I think if things go
[22:00] according to plan we should be able to
[22:03] uh
[22:06] we should be able to launch people
[22:07] probably in 2024 with arrival in 2025.
[22:13] So Is that is that a more certain
[22:15] schedule than United Airlines?
[22:20] Well um I don't know.
[22:23] Uh There's certainly some uncertainties
[22:25] associated with that.
[22:27] Um So um let's um I'm going to share
[22:29] Anyway that's the game plan like
[22:30] approximately 2024 to do the first to to
[22:33] to to launch the first um
[22:37] of the Mars colonial transport system
[22:39] with I want to get back to what you said
[22:40] earlier about a multi
[22:41] a very big rocket.
[22:43] Okay. A very big bigger than Saturn V?
[22:46] Twice as big or what?
[22:49] September. I'll tell you.
[22:51] Not going to say anything till
[22:52] September? Come on.
[22:53] Very big.
[22:54] Come on. Has to be very big.
[22:56] I how big is very big? So big.
[23:02] Um
[23:05] Do you think we should abandon the Earth
[23:06] at some
[23:07] No. No I No I think it's great. But you
[23:10] have said things
[23:12] Why would we abandon Earth? It's really
[23:14] nice here.
[23:14] You've said things about we may have to
[23:16] abandon the Earth so it's good to have a
[23:18] plan B. No I haven't. Or No that was
[23:21] amazing. That's I think that's maybe
[23:23] maybe I don't know but it wasn't me. All
[23:25] right. Okay.
[23:26] Wasn't me like Jackie you All right. Um
[23:28] So let's move to things on this Earth.
[23:31] Let's move to Hyperloop Tesla
[23:34] other things. Let's talk about Tesla
[23:35] first.
[23:37] Where do you feel like the company is at
[23:39] at this point and there's been lots of
[23:41] activity in self-driving cars and
[23:44] autonomous autonomous. How do you look
[23:46] at how everybody's jumped in Google
[23:48] Apple others? Um and all the car
[23:51] manufacturers.
[23:53] Um
[23:54] Yeah I mean there have been so many
[23:55] announcements of like autonomous EV
[23:57] startups. I'm waiting for my mom to
[23:58] announce one. Okay. Um it's like mom you
[24:01] too?
[24:03] Um
[24:06] I mean there's a lot.
[24:07] So
[24:08] um
[24:12] Yeah I mean in in
[24:14] in in the US alone there are there are
[24:16] four I think maybe five China-funded EV
[24:19] startups.
[24:21] At the billion dollar plus level like
[24:24] serious funding.
[24:25] Um
[24:26] and there's a bunch of startups and then
[24:28] of course the you know the car industry
[24:30] as a whole seems to be moving in that
[24:31] direction.
[24:34] Volkswagen just I think announced a huge
[24:37] battery factory that they're going to
[24:38] build.
[24:39] Um
[24:40] And I think these are all good you know
[24:41] it's good it's good for
[24:43] the industry to be moving towards
[24:45] sustainable transport as as quickly as
[24:47] possible.
[24:48] Um we open sourced our patents to try to
[24:50] be helpful in that regard.
[24:52] And um
[24:53] yeah so it's it's encouraging to see all
[24:55] this activity.
[24:57] Um
[24:58] From a Tesla standpoint you know we're
[25:01] just we want to take the set of actions
[25:03] that are
[25:05] uh likely to accelerate the advent of
[25:06] sustainable energy. Um so
[25:09] um
[25:11] scale up production as fast as we can so
[25:13] we accelerated plans for the Model 3 by
[25:15] two years.
[25:17] Um
[25:18] and to so we we want to try to get to
[25:20] half a million cars in total in kind of
[25:23] the 2018 timeframe which is an
[25:25] aggressive schedule but I think
[25:27] achievable.
[25:28] Um and and then maybe a million cars a
[25:31] year by 2020. Um
[25:34] and um
[25:37] you know I I can see a I I think a
[25:39] pretty clear path to get there.
[25:41] Uh autonomy is obviously extremely
[25:42] important. Uh
[25:44] people are going to want want autonomy.
[25:46] It's going to be odd to have a car
[25:48] without autonomy in the future.
[25:50] Um
[25:52] But
[25:53] um yeah so that I think that's So what
[25:55] what what do you How do you look at
[25:57] How do you look at the all these
[25:58] efforts? Not your mother but
[26:00] Yeah my mom's yeah she's She's not going
[26:02] to do it. She may do a rocket situation
[26:03] but
[26:04] um
[26:05] how do you look at each of them? Let's
[26:06] go through them. What Google's doing?
[26:09] How do you assess what they're doing
[26:11] when you're looking at it cuz they'll be
[26:12] competitors at some point. These are all
[26:14] Um eventual competitors. Well you know
[26:17] what I think what what Google's I mean
[26:19] Google's done
[26:20] a great job of showing the potential of
[26:23] autonomous transport. Um
[26:25] but they're they're not a they're not a
[26:27] car company so they would potentially
[26:29] you know license their technology to
[26:31] other car companies and I think they
[26:32] announced something with Fiat. Um
[26:36] and
[26:37] so I I wouldn't say you know Google's a
[26:39] competitor because they're not a car
[26:41] company. They We would compete with
[26:43] somebody perhaps that they license
[26:44] technology to but not to them directly.
[26:46] Right. Um Apple?
[26:48] Uh
[26:50] Yeah that that'll be more direct.
[26:52] That'll be more direct? Yeah. You can
[26:54] tell that by the hiring pattern and that
[26:57] kind of stuff.
[26:58] So what do you
[26:59] Okay so they're going to be more direct.
[27:01] How do you assess it?
[27:03] Um
[27:06] I mean I I say like you know I
[27:08] I think it's great that they're doing
[27:09] this and
[27:11] um I
[27:12] um
[27:14] you know I hope they I hope it works
[27:15] out.
[27:19] What's What's the timeframe for them do
[27:21] you think?
[27:22] Um I don't know I mean
[27:24] um I I think they should have
[27:27] embarked upon this project
[27:30] sooner actually.
[27:33] Uh but that uh
[27:35] um
[27:38] but I don't know
[27:40] I don't know when they I mean they have
[27:41] they don't they don't share with me the
[27:42] details of their
[27:44] production plans but um I I I
[27:47] I don't think it's going to be I don't
[27:49] think they'll be in volume production
[27:50] sooner than
[27:52] maybe 2020. That'll be like the soonest.
[27:55] And that's Is that too late? When you
[27:56] say they should have embarked sooner is
[27:58] is that because 2020 will be too late to
[28:00] stop you or beat you or compete with you
[28:02] or No it's just like it's a missed
[28:04] opportunity. It's just a that they it's
[28:07] a it's
[28:09] um It'll be over by 2020?
[28:11] I wouldn't say that. It's
[28:12] It's just like it's it's it's a couple
[28:13] of years.
[28:14] I think they'll they'll probably make a
[28:16] good car and probably be successful. The
[28:19] car industry is very big so it's not as
[28:21] though there's um
[28:23] you know one company to the exclusion of
[28:25] others.
[28:26] Um
[28:27] I mean there's like a dozen
[28:29] car companies in the world of of of
[28:31] significance. So
[28:32] uh and the the most that any company has
[28:36] is approximately 10% market share.
[28:38] So it's it's not like um
[28:40] you know somebody comes up with a car
[28:42] and they're suddenly like they kill
[28:43] everyone else. It's not not that way.
[28:46] Um
[28:47] and and the sheer scale of automotive
[28:49] manufacturing is is is just
[28:52] it's hard to appreciate until you see
[28:54] the plants. I mean that they're gigantic
[28:56] like the industrial Yeah I mean the the
[28:59] the the sheer size of the industrial
[29:01] infrastructure is is staggering. Not
[29:03] just the assembly plant but everything
[29:04] else that goes
[29:05] Yeah the supply chain exactly the supply
[29:07] chain is just a little tip of the
[29:08] iceberg really. The supply chain is
[29:10] literally tip of the iceberg.
[29:12] The the supply chain is um
[29:15] you know once you go to tier two tier
[29:17] three tier four suppliers
[29:19] that's
[29:20] uh probably an order of magnitude more
[29:22] uh So okay so you think Google will not
[29:25] be a competitor Apple probably will be a
[29:28] direct competitor.
[29:29] Yeah yeah sure. What about the car
[29:31] companies? The the I think they're all
[29:33] they're all going to be competitors.
[29:35] Yeah sure. Who do you see out there that
[29:37] has done a nice job so far? Mercedes or
[29:40] GM
[29:41] of of what? Of a competitive car.
[29:44] incumbents.
[29:46] Potentially competitive car, I guess.
[29:50] I mean
[29:52] I don't think anyone's any of the car
[29:53] companies as far as I've made
[29:55] a really great electric car.
[29:58] I mean, you tell me if I'm if you
[29:59] disagree, but I don't think yet that any
[30:03] of them have made a great electric car.
[30:06] Okay. And they, you know, presumably
[30:07] will continue to improve on what they've
[30:09] done so far and and then at some point
[30:11] they may make a car that's
[30:13] that's a
[30:15] you know, but that's a great car, but
[30:17] not they haven't done that yet. Can I
[30:18] ask you about batteries for a second?
[30:20] Yeah, sure. So you're building this
[30:22] gigafactory, right? You've got it's
[30:24] built.
[30:25] It's
[30:26] Well, it's not completely built. Okay,
[30:27] but it part of it's up and running. A
[30:29] chunk of it is built, yeah. Part of it
[30:30] is up and running.
[30:31] gigantic thing. It's like when the
[30:33] gigafactory is done, it'll be the
[30:35] largest footprint building of any kind
[30:37] in the world. Of any kind, not just
[30:38] factories, it's literally of any
[30:40] What is this? The largest rocket, the
[30:41] largest building. I mean, every
[30:43] Well, I mean, I think this it's not
[30:44] scale for scale's sake. It's just like
[30:46] if you say, "Well, we want to accomplish
[30:48] these goals, then um
[30:50] then you kind of have to be make a big
[30:52] thing."
[30:54] Okay. Um
[30:56] You've got this big thing. It's this big
[30:58] giant building. Yeah. It's going to make
[30:59] batteries. The batteries it's going to
[31:01] make are lithium
[31:03] to have an opening. Well, it's not
[31:04] technically an opening party since it's
[31:05] been operating for a little while, but
[31:07] we're going to have a party soon. You
[31:08] guys maybe want to come. We can come.
[31:10] Yeah, sure. We'll come to the battery
[31:12] party. But they can't come. What I was
[31:13] saying this is like
[31:15] Right? Just this This is crazy. No. This
[31:17] is like an alien dreadnought. It's
[31:18] really nutty.
[31:20] I love a battery party, but
[31:22] Right. Well, but but but But talk about
[31:24] where it's going. Are these lithium ion
[31:26] batteries?
[31:26] Yeah, sure. So they're the same
[31:28] batteries that's in our phone? No.
[31:31] Explain. Please explain.
[31:32] Yes, so so lithium
[31:34] Have you made a a battery breakthrough
[31:36] is something I'm interested in.
[31:38] Yeah, I mean, generally the I mean,
[31:40] there's there's so much nonsense out
[31:42] there about batteries. Like about you
[31:43] can believe about 1% of what you read.
[31:45] Um on a you know, maybe. Um
[31:48] Uh
[31:49] Lithium ion covers a very broad range of
[31:52] technologies. Um
[31:54] and you can have an enormous difference
[31:57] in the power density and the energy
[31:58] density and the cycle life um
[32:03] between one chemistry and another. They
[32:05] can be really enormously different. Um
[32:08] So
[32:10] Uh what you really actually want to ask
[32:12] is what is the cathode and what is the
[32:13] anode? Right. Um
[32:16] So in our case
[32:17] That's right.
[32:18] Um
[32:19] I just put it in the The the lithium is
[32:21] actually 2% of the cell mass.
[32:24] So so like it's like the salt in the
[32:25] salad. It's it's a very small
[32:28] um
[32:29] amount of the cell mass and a very small
[32:31] amount of the cost. Um
[32:33] But it sounds like it's big because it's
[32:35] called lithium ion, but it it really
[32:37] like our battery should be called
[32:39] nickel graphite
[32:40] cuz it's mostly nickel and graphite.
[32:43] Okay, and Um it's nickel cobalt aluminum
[32:46] But battery
[32:46] little things in graphite with a silicon
[32:48] oxide layer.
[32:49] Battery efficiency or power you know,
[32:52] the power that you can store in a
[32:53] certain uh mass seems to be
[32:57] move very slowly, at least compared to
[32:59] you know, we're used to Moore's law
[33:01] pushing uh integrated circuits faster.
[33:04] Batteries kind of are always in our
[33:06] consumer devices always lagging behind.
[33:09] In your you've built this giant thing,
[33:11] the biggest building in the world that's
[33:12] ever seen.
[33:13] not fully built, but yes, it's Your
[33:15] building a A pretty big chunk of it is
[33:16] built, so yeah.
[33:17] Uh to make batteries. Your whole
[33:20] business depends on batteries in these
[33:21] cars. Have you figured out a way to do
[33:25] some significant
[33:27] uh increase in the yield of energy from
[33:31] a given amount of
[33:33] of space in the battery?
[33:35] Well
[33:37] Yeah, I mean, the the
[33:39] the energy density is increasing sort of
[33:41] maybe on the order of like 5-ish percent
[33:44] per year.
[33:46] Um
[33:46] And it doesn't sound like much, but you
[33:47] add that up over a number of years with
[33:50] compound interest, it ends up being
[33:51] quite quite a significant number.
[33:54] Um and a lot of people sort of think
[33:56] that, "Oh, well, we just sort of cobbled
[33:58] together some
[33:59] um laptop batteries and somehow made a
[34:02] great car." If it If it was that easy,
[34:03] then I think we would have quite a few
[34:04] competitors who did the same thing. But
[34:06] but it's it's it's really quite quite a
[34:08] lot harder than that. Um the it is a
[34:11] cylindrical form form factor, but the
[34:13] internals of the battery are quite
[34:14] different from what you'd find in uh in
[34:16] a laptop. Um
[34:19] And uh and and and it will be
[34:21] increasingly different with the what's
[34:23] built at the gigafactory, which is
[34:24] highly optimized for automotive. Um
[34:27] And um
[34:28] and with has improved energy density,
[34:30] but but mostly it's not the energy
[34:32] density that's the issue cuz
[34:34] you know, you can buy if you buy a Model
[34:36] S today
[34:37] um the range is um
[34:39] around 300 miles. Um
[34:42] And and yeah, that's quite a lot. Um
[34:45] So it's pretty rare that people really
[34:48] need to go more than 300 miles at a time
[34:50] without stopping. Right.
[34:52] um
[34:54] So I don't think we really have a range
[34:56] issue. And we could make a 400-mile
[34:57] range car today.
[34:59] Like that wouldn't be too big of a deal.
[35:01] Um
[35:03] The What What really matters is
[35:05] decreasing the cost
[35:07] uh per unit of energy of the battery
[35:09] packs so that you can make the car
[35:10] affordable. That's actually the the the
[35:12] important thing. Um
[35:14] So there's
[35:15] and there's really two main main
[35:17] dimensions along which uh
[35:19] cost optimization and making something
[35:21] available to mass market can be
[35:22] achieved. One is design iteration, going
[35:24] through multiple versions of something.
[35:26] And then the other is economies of
[35:27] scale.
[35:28] Um and you kind of need both of those
[35:30] those things in order to make a
[35:32] compelling mass market uh product. And
[35:34] you look at like cell phones and how
[35:35] many design iterations have we gone
[35:37] through with cell phones?
[35:39] Um and and then and and and what and
[35:42] look at the scale at which they're
[35:44] they're made is enormous. Uh and that's
[35:46] what enables everyone to have a
[35:48] supercomputer in their pocket.
[35:50] Um So speaking of that, the sales. When
[35:52] you're talking about the sales, you have
[35:54] booked how many orders for
[35:56] It's on the order of 400,000.
[35:58] 400,000. So obviously it's consumer
[36:00] interest in a promise. A lot of it
[36:02] around you, around the idea of you and
[36:04] Tesla and the excitement. Yeah, it's
[36:07] it's quite surprising, actually. I mean,
[36:08] I
[36:09] the
[36:10] um
[36:11] cuz we didn't do any advertising or
[36:14] there was no guerrilla marketing or
[36:16] anything. It was just basically like,
[36:17] "Yeah, we're going to have this
[36:18] webcast." There was only there were only
[36:20] about a thousand people in the audience.
[36:22] Um and um
[36:24] it really caught us by surprise. But I
[36:25] think you know, when you have a product
[36:27] that really resonates with with
[36:30] customers, the word of mouth uh
[36:32] grows like wildfire. And that seems to
[36:34] be what it
[36:35] Yeah, but it's a little bit. I mean,
[36:36] honestly, in some groups of especially
[36:39] men in Silicon Valley, if you show up
[36:40] and read like a label of a peanut jar,
[36:42] they'd be thrilled with the situation.
[36:44] So I mean, you a lot of this does base
[36:47] around you. Like the idea of you and the
[36:49] excitement around this exciting
[36:51] entrepreneur. Is that
[36:54] Is that enough to get it to to to this
[36:56] massive company you've been hoping to?
[36:57] The idea of
[36:59] this is the Elon promise or it's the
[37:01] Well, I I think actually it's not so
[37:03] much
[37:04] I mean, I sort of um
[37:07] um I I mean, I'm not sure. I I think I
[37:10] deserve less credit than that, actually.
[37:11] The the
[37:14] I
[37:15] What What Tesla's done with a phenomenal
[37:16] team is like 15,000 people at the
[37:18] company. Um
[37:20] worked super hard to create compelling
[37:22] products, to create great cars. Um and
[37:24] we started off with the roadster and
[37:26] then the Model S. The Model S was rated,
[37:27] you know, by Consumer Reports as the
[37:28] best car ever. Um
[37:31] Got the Model X, which you know, had
[37:32] some has had some teething issues, but
[37:35] um I think it's now at the point where
[37:36] it's it's really starting to
[37:39] I think it's really
[37:41] I think quite sublime at this point. Um
[37:45] And uh and and so people look at that
[37:47] and say, "Okay, well, if Tesla's made
[37:50] these cars, then probably the next car
[37:52] they make is going to be
[37:54] you know, the less expensive one.
[37:55] also a great car. And Um yeah, but you
[37:58] know, so it'll be a great car, but it'll
[38:00] be affordable. And it's like, "Great.
[38:01] Okay, that sounds like something I
[38:02] want." Um So this car, this next car,
[38:04] the price is the It's starting at
[38:06] 35,000. Okay, affordable. Okay, when is
[38:09] the When do you get to the When do you
[38:10] get to the When do you get to the really
[38:12] affordable, the way down much lower than
[38:14] that? Yeah, I mean, it's more important
[38:16] that the 35,000, particularly when
[38:18] you're factoring in the lower cost of
[38:20] electricity versus gasoline and that the
[38:22] maintenance cost is much less. You don't
[38:23] have to have oil changes. You never need
[38:25] to replace your brakes because the the
[38:27] car uses regenerative braking, so the
[38:29] brakes last as long as the car do at the
[38:31] car does.
[38:32] Um it's basically you just need to
[38:33] replace the tires. Like that's about
[38:35] about all.
[38:36] Um So the operational cost of the car is
[38:38] much lower fundamentally than than a
[38:39] gasoline car.
[38:41] And and so
[38:42] um
[38:43] And and I think I think the the the
[38:45] average price for cars
[38:48] for gasoline cars is around 30, 32,
[38:51] something like that, yeah. Uh I mean,
[38:52] there are starting prices that are
[38:53] lower, but but when people pick pick
[38:55] options, I believe it's in the
[38:57] around 32 or so in the US. So we're
[38:59] pretty close to the that that that But
[39:01] that's your base price, right? I mean,
[39:03] the Yeah, but it's going to be a great
[39:04] car.
[39:05] ASP for the car.
[39:06] No, but it's going to be a great car
[39:07] even at 35.
[39:09] So it's like even if you order nothing
[39:10] no options at all, it'll be great. But
[39:12] you're at but you're likely to have a
[39:13] mix where the average
[39:16] car that you actually sell sells for a
[39:18] little more than that. Yes, probably
[39:21] it's probably going to be Yeah, it's
[39:22] probably going to be some higher number.
[39:24] Um
[39:25] But it's really important to emphasize
[39:26] like the the 35 if if somebody orders
[39:28] the 35,000 car, they'll be very happy.
[39:30] Like it's not like you need to order a
[39:31] bunch of options in order without which
[39:32] the car is is, you know, not not good.
[39:35] That car will have autonomous for 35? Um
[39:40] I have a
[39:41] uh
[39:44] I'm going to do another of Tesla van
[39:46] maybe at the end of the year. Um,
[39:48] talk more about that.
[39:50] And say
[39:50] Well, so you could start here.
[39:55] Um, it would be real big news if I start
[39:57] here.
[39:58] Um We don't mind that.
[40:01] Let me just say that we're going to do
[40:02] the obvious thing.
[40:05] Okay.
[40:06] Okay.
[40:07] Got it. It's really obvious.
[40:10] So cup
[40:11] So
[40:12] So cup holders good. Okay. Um
[40:17] So brilliant brilliant But those things
[40:18] are nuanced. All right, absolutely. Um,
[40:21] let's talk about two more things. I want
[40:23] to talk about AI cuz we've been talking
[40:24] about it a lot here. Um,
[40:26] which
[40:27] I want to get it clear what your
[40:29] thoughts are because it's mostly Elon's
[40:31] scared of robots. I mean, that kind of
[40:32] thing or what How do you I'm just scared
[40:35] of robots.
[40:36] Um
[40:37] Or artificial intelligence. Can you like
[40:39] clarify exactly what the issue you have
[40:41] now?
[40:41] And and you deserve the background.
[40:43] We've been talking to
[40:45] uh Jeff Bezos, Sundar Pichai. Uh, we
[40:51] talked to Mark Fields from Ford about
[40:53] it. Um it's
[40:54] Microsoft for Uh
[40:55] yeah, the Facebook folks. Um, there
[40:57] certainly seems to be uh in the I in the
[41:01] tech companies a big
[41:03] tremendous new drive or interest to
[41:06] believing they will be all competing
[41:09] for intelligent assistants and
[41:11] good. It'll make your life better Make
[41:13] your life better. Siri is going to
[41:15] suddenly get smart. Microsoft's one is
[41:18] going to get smart. And Google is going
[41:20] to cream them all.
[41:22] a happy version of this is going to
[41:23] Sometimes technology hurts you, but not
[41:25] as much as it helps you. That's really
[41:27] Yeah. So that's there's been a lot of
[41:29] conversation here about that.
[41:31] Um and yet and you've staked out a
[41:32] slightly different position. So can you
[41:34] talk about that?
[41:35] Well, I mean, I think my sort of full
[41:37] position would
[41:39] require quite a long explanation. Um
[41:42] I mean, I
[41:44] I am concerned about um
[41:46] certain
[41:48] directions that AI could take that would
[41:50] be uh
[41:53] not good for the future. That the I
[41:55] mean, it I I I think it'd be fair to say
[41:57] that like not all AI futures are benign.
[42:00] Not not all. Okay. Um
[42:02] And and so if you have something if if
[42:04] if there's if we create some digital
[42:07] superintelligence that exceeds us in
[42:08] every way by a lot, um
[42:12] it's very important that that be benign.
[42:14] Um
[42:16] And um
[42:17] and so actually with with the with a few
[42:19] others um
[42:21] I created uh Open AI.
[42:23] Uh which is uh an AI
[42:26] uh
[42:28] It's a nonprofit actually. It's so so
[42:29] there's there's no I think the
[42:31] governance structure here is important.
[42:33] Um, so you want to make sure that there
[42:34] was not some fiduciary duty to uh
[42:37] generate
[42:38] um
[42:39] you know, profit or proof of the AI
[42:40] technology that's developed.
[42:42] Um So uh so we created this uh
[42:46] a 501c3 um but it but I think
[42:49] it's it's like quite different from I
[42:51] mean, like a lot of sort of
[42:54] 501c3s are, you know, they have
[42:57] they don't have a high sense of urgency.
[42:59] Um like they're they're not like
[43:01] um
[43:04] you know, they're not really
[43:05] sort of
[43:06] ex
[43:08] developing technology at at at a fast
[43:10] pace, but Open AI is. Uh so Open AI has
[43:13] a very high sense of urgency and the
[43:15] talent I think that the people that have
[43:17] joined are are really
[43:19] really amazing. Um
[43:21] um and um
[43:23] and the intent with Open AI is to
[43:26] democratize
[43:28] AI power. Um
[43:30] And there's a quote that I love from uh
[43:33] Lord Acton. He was the guy that came up
[43:34] with power corrupts and absolute power
[43:36] corrupts absolutely. Um which is that uh
[43:39] freedom consists of the distribution of
[43:41] power and despotism in its
[43:42] concentration.
[43:44] And so I think it's important if we have
[43:46] this incredible power of AI that it not
[43:48] be concentrated in the hands of a few
[43:49] and potentially lead to a world that we
[43:52] don't want. And what world is that? What
[43:54] is What do you see foresee that when you
[43:56] say
[43:57] It's difficult I mean, it's called the
[43:58] singularity cuz it's it's difficult to
[44:00] predict. Um
[44:02] what exactly what future that might be
[44:05] except um I don't know a lot of people
[44:07] who love the idea of living under a
[44:08] despot. Mhm.
[44:10] Um
[44:11] you know, I don't think people generally
[44:13] choose to live in a democracy over a
[44:15] dictatorship.
[44:15] Mhm. And the despot would be the
[44:17] computer?
[44:18] Well, the people controlling the
[44:19] computer. Mhm.
[44:21] And do you worry specifically about any
[44:25] of these companies I mentioned who've
[44:27] all seemed to
[44:29] now kind of be pivoting toward this as
[44:31] the battleground in the next 10 years? I
[44:34] won't name a name, but there is only
[44:35] one.
[44:37] There's only one you're worried about.
[44:40] And they're not preoccupied with making
[44:41] a car that will compete with you, I
[44:44] assume.
[44:46] There's only one.
[44:51] And what tell tell me This is an
[44:52] interesting
[44:53] competing It's not about competing. It's
[44:55] Like like this is sort of like who like
[44:56] what would be the point of competing for
[44:58] you know
[45:00] mutual destruction. It's like there's no
[45:01] It's not about competing. It's really
[45:02] just about
[45:04] um trying to increase the probability
[45:06] that the future will be good.
[45:09] That's all. Mhm. So the
[45:12] the goal of of Open AI is really just to
[45:15] take the set of actions that are most
[45:16] likely to improve the
[45:18] positive futures. Like if you can think
[45:20] of like the future as a set of
[45:22] of probability streams that that that
[45:25] branch out and then converge collapse
[45:27] down to a particular event and then
[45:29] branch out again. And uh
[45:32] there's a certain set of probability
[45:33] associated with the future being
[45:34] positive and different type flavors of
[45:35] that. And uh at Open AI we want to try
[45:39] to get do do whatever we can to guide to
[45:41] to increase the probability of of the
[45:43] good futures happening. I think that's
[45:45] that's really what we're trying to do
[45:46] there.
[45:46] worry that by making this open some bad
[45:49] actors may use some of what has been
[45:51] developed to do
[45:54] bad stuff uh with the power of the AI?
[45:56] Yeah, I mean, that that is certainly the
[45:57] the
[45:58] the I mean, a good rebuttal to that. Um
[46:01] however, I think if AI power is widely
[46:03] distributed um then and there is not uh
[46:07] say one entity that has
[46:11] some super AI that is a million times
[46:14] smarter than anything else.
[46:16] Uh you know, if it if instead the AI
[46:17] power is broadly distributed and to the
[46:20] degree that we can link uh AI power to
[46:24] um each individual's will. Um so like
[46:27] you you know, you would have your AI
[46:28] agent and you would like everyone would
[46:30] have their sort of AI agent. And then if
[46:33] somebody did try to do something really
[46:34] terrible, well, then uh the collective
[46:36] will of others could overcome that bad
[46:38] actor.
[46:40] Um
[46:41] which you can't do if if there's one AI
[46:44] that's
[46:46] you know, a million times better than
[46:47] everything else.
[46:48] and it's proprietary.
[46:49] And it's yeah, it's either has its own
[46:52] will or more likely at least in the
[46:53] beginning is controlled by
[46:56] you know, some small set of people. So
[46:59] um I think that's that's really the the
[47:01] risk. I mean
[47:02] um
[47:03] you know, there's always there's always
[47:04] been these arguments like what's the
[47:05] what's the best form of government? Um
[47:07] you know, big fan of
[47:09] I think it was Churchill like, you know,
[47:10] democracy is the the worst form of
[47:12] government except for all the others.
[47:13] Right. Um So speaking of that Yeah.
[47:16] this election. You are
[47:18] from
[47:18] Oh, no. No, no, no.
[47:20] Yes, yes, yes. How does that strike you
[47:23] what's happening now?
[47:25] You're you you come to this country or
[47:27] naturalized citizen?
[47:27] Uh you know, I think I'm glad that the
[47:29] framers of the Constitution saw fit to
[47:32] ensure that the president uh was someone
[47:34] who um
[47:36] was captain of a large ship with a small
[47:37] rudder.
[47:38] Okay.
[47:40] And
[47:42] There's a limit to how much harm any
[47:43] given president
[47:44] Are you sure about that? Oh, yeah. Yeah.
[47:47] Yeah. So you're not worried about Are
[47:49] you backing in either of the candidates
[47:51] at this point?
[47:52] stay out of this situation.
[47:54] Because
[47:56] I don't think this is the finest moment
[47:57] in our democracy. Well, given that it's
[48:00] not the finest moment in our democracy,
[48:02] do you think the best thing is to stay
[48:04] out
[48:05] or
[48:05] We'll see.
[48:06] get in I'm not sure what I what I can do
[48:08] to I mean
[48:09] worst. I don't know how much how much
[48:10] influence I could have as as one person
[48:12] on the outcome. Um so
[48:15] um
[48:18] I mean, if I think I could make a
[48:20] difference, I would probably do
[48:20] something. Um but um
[48:24] like I said, I think I'm just glad that
[48:25] that the pre you know,
[48:27] being the US president is like being
[48:29] captain of a large ship with a small
[48:30] rudder. Mhm. And so like there's just a
[48:33] limit to
[48:34] how much good or bad a president can can
[48:36] can actually do. I mean, obviously if if
[48:39] a president could make the economy great
[48:41] that and there was like a button he
[48:43] could press, they'd be pressing that
[48:44] button at the speed of light.
[48:45] Mhm. Um
[48:47] So
[48:49] you know, that they they but they
[48:51] they can't. So I can't they can't just
[48:52] magically make the economy good. Um no
[48:55] president wants the economy bad ever. Um
[48:57] but they you know,
[48:59] like there's just a limit to to how much
[49:00] they can do. Um
[49:03] and um
[49:05] yeah.
[49:07] I guess there is the nuclear thing which
[49:08] is Yeah, the nuclear
[49:11] I guess there is the nuclear thing.
[49:13] Yeah, but I I I don't know. I think you
[49:14] would
[49:15] I think I think
[49:16] I don't think we would like just
[49:18] arbitrarily launch nuclear missiles.
[49:19] Yeah. One would hope.
[49:20] President can do that.
[49:21] Uh I don't think so. Arbitrarily. I
[49:23] mean, I think that He's No, he's the
[49:25] commander-in-chief.
[49:26] but I don't I still I still don't I
[49:27] still don't think that means you can
[49:28] just launch nuclear missiles whenever
[49:29] you want. All right. Yeah.
[49:30] Um I I think Congress would be like
[49:32] quite upset about that and um They might
[49:35] not be consulted. Yeah, but I think
[49:38] I think like the military would be like,
[49:39] "Yeah, we really think Congress should
[49:40] be consulted on that before you launch a
[49:42] Yeah, that that might happen.
[49:44] preemptive nuclear strike. Are you
[49:45] willing that you're basing your faith in
[49:47] that though? I'm quite confident that
[49:49] the military would not just
[49:51] you know, randomly agree to launching
[49:54] nuclear missiles at somebody.
[49:56] Well, that's calming. Um so
[49:59] um
[50:00] we're going to put up just very quickly
[50:01] with and on Hyperloop. You've been
[50:03] involved with it. Your level of
[50:04] involvement is what at this point? Just
[50:06] Yeah, and I know it's a it's a bit it's
[50:08] a bit confusing cuz um
[50:10] I am
[50:12] You talked about it when you were here
[50:13] last time.
[50:14] Yeah, um
[50:15] I I actually came up with the idea um
[50:18] I came up with an initial idea which
[50:20] which turned out to be wrong and
[50:22] wouldn't work um several years ago
[50:25] and
[50:27] and then
[50:28] um but I sort of shot my mouth off and
[50:30] and said I like have an idea that would
[50:32] work and turned out that it didn't work
[50:33] but I
[50:34] with a lot of iteration was able to come
[50:35] up with something that
[50:37] where the physics hangs together.
[50:39] Um and then published the paper and just
[50:41] said like look anyone who wants to do
[50:42] this is great. Go you know, be my guest.
[50:45] Cuz I'm I sort of have a plate full
[50:47] running Tesla and SpaceX. Yeah, yeah.
[50:50] And
[50:51] so I think it'd be great. I mean I I
[50:53] think it'd be great to have any
[50:54] interesting new transport solutions. Um
[50:57] anything that gets people to their
[50:59] destination um in a way that's safer,
[51:02] costs less, is more convenient. Um
[51:06] that'd be great. I mean and so I think
[51:09] probably
[51:10] the most valuable thing that the
[51:11] Hyperloop paper that I published um
[51:16] uh has done is is to spur thinking in
[51:19] terms of new transportation systems. So
[51:21] it's not just oh let's you know, have a
[51:25] a fast train.
[51:27] Okay, that's not even as fast as what
[51:29] Japan did in the '80s. Like okay, why
[51:31] would I don't see what the point of that
[51:33] is, you know.
[51:34] Like we should really be trying to think
[51:36] of some something that's
[51:38] um I think particularly in California
[51:40] like we should be like saying, "Hey,
[51:41] what is the best Let's invent something
[51:44] new that's
[51:45] way better than anything else. Do you
[51:47] want to shoot your mouth off about that?
[51:50] Um
[51:51] well
[51:52] um you know, I so so I I mean I'm not an
[51:54] investor in any of the companies uh that
[51:56] that are working on it and I've tried to
[51:58] be neutral cuz I'm like trying to trying
[51:59] not to favor one company over another.
[52:02] Uh but just to encourage anyone that is
[52:04] interested to say like you know,
[52:07] the
[52:07] try try to give them moral support, you
[52:09] know. Um
[52:11] And I I hope they succeed. Um the only
[52:13] thing that um I am doing on Hyperloop
[52:15] front is like we're we're holding a
[52:16] student competition and the student
[52:18] competition is really just um aimed
[52:20] towards encouraging
[52:23] uh students to think about exciting new
[52:26] transport methods. Um and it's totally
[52:28] cool if they want to like do some
[52:30] architecture that's different from
[52:32] uh what I proposed in Hyperloop. In
[52:34] fact, the the winning team at the
[52:36] student competition that we held earlier
[52:37] this year used a different um suspension
[52:40] mechanism
[52:41] uh than what I proposed which is I you
[52:44] know, I I proposed using essentially
[52:46] taking taking the
[52:47] uh air that eventually is that builds up
[52:49] on the nose from the compressor and and
[52:51] flowing that through um air skis. Um
[52:55] So that you simultaneously remove the
[52:57] drag from the nose and provide a a a
[53:00] means of suspending the
[53:02] the pod. Um and and that's also
[53:04] something that that works well um even
[53:07] at uh super supersonic velocities. You
[53:09] can go it's been demonstrated up to Mach
[53:11] 1.1 in terms of using air bearings as
[53:14] But they use something different. Uh uh
[53:16] yeah, basically electromagnetic
[53:17] suspension.
[53:19] Um and and like the the the reason I
[53:21] didn't suggest um
[53:23] uh
[53:24] um
[53:25] sort of any any kind of um
[53:27] um magnetic suspension um is that
[53:30] it's very important that the cost of the
[53:33] of the tube be be minimized.
[53:36] So you really want because the the the
[53:38] pod is cheap, the tube is expensive.
[53:40] Um so if you if you want to go say 400
[53:42] miles and you and and two in two
[53:44] directions, so you've got 800 miles of
[53:45] tube.
[53:46] The the the critical uh economic
[53:49] optimization parameter is the cost of
[53:51] the tube. So you want that tube to be as
[53:53] low cost as possible. Um and so if you
[53:57] if you do anything that that
[53:59] um
[54:00] requires
[54:01] um
[54:03] action on the tube side, it's going to
[54:05] make that tube much more expensive.
[54:07] Um so if you use air bearings, it
[54:09] doesn't change like that's real cheap.
[54:11] Um
[54:12] and um
[54:14] yeah, so Do you think this is going to
[54:16] happen?
[54:19] Yeah, I think something like that I
[54:20] think something will happen in the
[54:21] future. Um
[54:23] You know, it's I think I think if if if
[54:25] the companies that are that are trying
[54:26] that are
[54:27] trying to make it happen now, if if for
[54:29] whatever reason that that doesn't work
[54:31] out, um
[54:32] then you know, I think I I'll you know,
[54:35] I'll I might I might do something myself
[54:38] in the future. I I don't want to do
[54:39] something
[54:40] I don't want to front I don't want to
[54:41] sort of front run them, you know, and
[54:42] say here's this free idea and then
[54:44] meanwhile I go and do it myself, you
[54:46] know, that wouldn't be nice. So
[54:48] um
[54:49] so so but if I if they if they if they
[54:52] if if much
[54:53] companies don't try and it doesn't work
[54:55] out then then I think
[54:56] I think um
[54:58] I think I'll try to just at least do a
[54:59] demonstration system, you know. Okay,
[55:02] last question. Do you think tech has
[55:03] gotten more serious? Do you How do you
[55:05] look at the tech landscape as someone
[55:07] who's you know, well known
[55:09] you probably qualify as a visionary. Um
[55:12] the concept What What do you imagine we
[55:14] are right now in the tech space and then
[55:16] we'll get to questions from the
[55:16] audience.
[55:20] I think there's a lot of innovation
[55:21] happening
[55:22] in in many different areas.
[55:25] Um
[55:26] The advancements in AI are I think are
[55:28] quite quite astonishing.
[55:30] Um the advancements in genetics are
[55:32] amazing.
[55:34] Um
[55:38] The
[55:39] So I I think that there is a lot of
[55:40] innovation going on.
[55:43] Um I think that there's probably a few
[55:45] too many talented entrepreneurs in kind
[55:49] of the internet space and and I think
[55:51] their talent actually would be better
[55:53] served in some other industries.
[55:55] Um
[55:57] But I do think I mean I don't think
[55:59] we're like facing some sort of low
[56:01] innovation period or anything like that.
[56:03] I think there's a lot of innovation
[56:04] going on. They need to move to other
[56:06] Uh I just think there's like if you had
[56:08] some ideal distribution there probably
[56:09] would be fewer
[56:11] like there's just a lot of talent
[56:12] focused on the internet and probably
[56:14] some of that talent um
[56:16] uh
[56:17] would
[56:18] um
[56:19] be be be better to have some of that
[56:21] talent in other industries.
[56:22] Um that's about all but but there's
[56:25] tremendous amount of innovation that
[56:27] that's happening.
[56:28] Um
[56:29] It's something that I think is is
[56:31] going to be quite important. Um
[56:36] And and it's there's not I don't know of
[56:38] a company that's working on it seriously
[56:40] is
[56:42] um
[56:45] is a neural lace. Um
[56:48] so
[56:49] you know, going going back to the AI
[56:51] situation. Um
[56:52] this is quite important. Uh
[56:56] quite important debate. Like the
[56:59] if you assume any rate of advancement in
[57:01] AI, um
[57:03] we will be left behind by a lot.
[57:07] Um and so then we could be in like
[57:11] you know, a benign situation but the
[57:12] even the benign situation if you have
[57:14] some you know, if you have ultra
[57:15] intelligent AI, um we would be
[57:18] you know,
[57:20] so so far below them in intelligence
[57:22] that it would be would be like you know,
[57:24] a pet basically.
[57:25] Cat, that's what I was thinking.
[57:26] Like a pet, a cat, a cat, like a house
[57:28] cat. Yeah, we'd be like a house cat.
[57:30] Um
[57:32] and um yeah, so that's that's the end of
[57:34] the world, you know. Just
[57:36] sort of pets.
[57:37] Seen the movie, it could be. It could
[57:38] be, could be.
[57:40] Um
[57:41] the
[57:42] Yeah, so that but that honestly that
[57:43] would that would be a benign scenario.
[57:45] Um and
[57:46] So house cat is okay?
[57:49] I mean I don't love the idea of being a
[57:51] house cat.
[57:51] Okay.
[57:52] But but that
[57:53] So what's the solution? Yeah, so I think
[57:56] the
[57:58] um
[58:00] I I think I think it I think it's to
[58:02] essentially I think one of the solutions
[58:05] the solution that seems maybe the best
[58:08] one
[58:09] is
[58:10] to have an AI layer. Um if you think of
[58:13] like you've got your limbic system,
[58:15] um your cortex and then um a digital
[58:18] layer, a sort of a third layer above the
[58:20] cortex
[58:22] um
[58:23] that
[58:24] um could work work well and
[58:26] symbiotically with with you. I mean just
[58:29] as your cortex works symbiotically with
[58:31] your limbic system, your sort of a third
[58:33] digital layer could work symbiotically
[58:35] with the rest.
[58:36] This is something that's
[58:38] in surgically inserted or bred into the
[58:41] species or what?
[58:43] The the fundamental limitation is
[58:45] input-output. So
[58:47] uh we we already have uh we we're
[58:50] already a cyborg. Um it's just that I
[58:53] mean you have a digital
[58:55] version of yourself or or
[58:57] partial version of yourself online in
[58:59] the form of your emails and your social
[59:00] media and all the things that you do. Um
[59:04] And and you have basically superpowers
[59:07] in in that with your computer and your
[59:08] phone and and the applications that are
[59:10] there. Um you have more power than the
[59:13] president of the United States had 20
[59:14] years ago. You can answer any question.
[59:17] Uh you can video conference with anyone
[59:20] um Right. anywhere. You can send a
[59:22] message to millions of people instantly.
[59:24] Um
[59:26] you know, you just do incredible things
[59:28] and
[59:30] um
[59:32] but the constraint is is input out
[59:35] output. So, we're we're IO bound.
[59:37] Um
[59:38] particularly output bound. I mean, like
[59:40] the
[59:41] your output level is so low. It's like
[59:43] the particularly on a phone like your
[59:45] two thumbs are sort of tapping away.
[59:47] Um this is ridiculously slow.
[59:49] Um
[59:51] our input is much better cuz we have a
[59:53] high bandwidth visual interface with the
[59:55] brain. Like our eyes
[59:57] take in a lot of a lot of data. Um
[60:00] so, there's many orders of magnitude
[60:01] difference between
[60:03] um input and output. Um
[60:07] So,
[60:08] mostly
[60:10] um
[60:10] effectively merging in a symbiotic way
[60:12] with
[60:14] uh digital intelligence
[60:16] revolves around
[60:18] eliminating
[60:20] the IO constraint.
[60:22] Um
[60:23] so, it's with some sort of direct
[60:25] cortical interface.
[60:27] Um
[60:27] And you called it a neural Neural lace.
[60:29] Neural lace. Yeah.
[60:32] Um
[60:34] It's totally not Google Glass, right?
[60:36] No. I'm talking about something which
[60:38] wear it or No. I mean, it would be uh
[60:45] I mean,
[60:46] I mean, there are a few ways to approach
[60:47] this, but some sort of interface
[60:49] directly with your cortical neurons
[60:51] particularly. But doesn't that imply
[60:53] surgical insertion?
[60:54] Not necessarily. You could go through
[60:55] the veins and arteries cuz that that
[60:57] provides a a complete uh roadway to um
[61:00] all of your neurons. Your neurons are
[61:02] very heavy users of energy, so they need
[61:05] high blood flow. So, you automatically
[61:07] with your veins and arteries have um a
[61:09] road network to your neurons. Still
[61:11] search some kind of surgery, right? Um
[61:14] yes, but you could insert something, you
[61:16] know, basically
[61:18] you know,
[61:19] into the jugular and and have
[61:22] It gets my car, but
[61:24] really easy and It It doesn't involve It
[61:27] doesn't It doesn't involve, you know,
[61:28] like chopping your your your skull open
[61:31] or anything like that. Yeah.
[61:32] And plus you're not a house cat anymore,
[61:34] right?
[61:34] Not a house cat. So,
[61:35] Right. um I mean, essentially if if we
[61:37] can figure out how to
[61:40] establish a high bandwidth neural
[61:42] interface with ourselves. with with your
[61:45] digital self effectively.
[61:48] Um then
[61:51] uh
[61:52] then you're no longer a house cat. You
[61:54] know.
[61:54] All right. On that note. Now, on that Uh
[61:56] well, well, I just one
[61:59] closing thing.
[61:59] it's probably
[62:01] the best outcome. That's Are you Are you
[62:02] Are you interested
[62:03] you interested in exploring this
[62:05] possibility
[62:08] that you have just laid
[62:09] got to do it.
[62:12] I'm not saying that I will, but I'm
[62:14] somebody's got to do it.
[62:15] I mean, I I I I mean, I So, somebody
[62:17] should do it. And I mean, if somebody
[62:18] doesn't do it, then I then I think I
[62:20] should probably do it, but uh
[62:22] And and the goal of this is to prevent
[62:24] there being an external
[62:26] uh AI, particularly one controlled by a
[62:30] small group of people that could Yeah.
[62:32] be so much more powerful and intelligent
[62:35] than we are that the house cat
[62:36] would be godlike in its capabilities.
[62:38] Yeah. Well, this has been really
[62:40] cheerful. Thank you.
[62:41] Yeah.
[62:42] But But if we can establish
[62:44] about asteroids at the beginning of
[62:45] this.
[62:47] I mean, ast- asteroids are a low
[62:48] probability existential threat um on the
[62:51] time scale that's relevant to us. Okay.
[62:54] Okay. Um
[62:55] This is different. This requires
[62:56] urgency.
[62:57] So, what do you do for fun?
[62:58] Yes, this is much more
[62:59] do you do for for fun fun? What do you
[63:01] do? Anything?
[63:03] I play video games with my kids. All
[63:04] right. That sounds good. Let's get some
[63:06] questions for Elon. Elon the house cat.
[63:09] I watch movies, you know, that kind of
[63:10] thing. Normal things.
[63:11] Why don't we start over here? Yeah, hi.
[63:13] I I think this last question by
[63:16] uh Carl just just did that. Uh I want to
[63:19] know how do you uh live through the
[63:21] stress that kind of conversations we
[63:23] just heard that you went through and um
[63:26] kind of ambitions that you carry and
[63:27] then how do you adjust to the everyday
[63:30] work-life balance etc. things that in
[63:32] your life. So, a little bit of your
[63:33] personal side, actually.
[63:35] You're very busy. How does that work?
[63:37] Yeah, I mean, um
[63:39] I am sort of in kind of work triage mode
[63:42] um a lot of the time. So, I don't know.
[63:44] It seems to be
[63:46] as long as there's not like a crisis
[63:48] simultaneously at SpaceX and Tesla, it's
[63:50] okay. Um but you know, companies are I
[63:54] mean,
[63:55] the situation in any given company,
[63:56] particularly when, you know, if if if
[63:58] it's sort of growing fast and sort of
[63:59] quasi startup, it's it's somewhat
[64:01] sinusoidal. So, that I mean, it's okay
[64:03] if if you don't if the if the waves
[64:05] don't crest together. You know, um
[64:08] when that does happen, it then that's a
[64:10] huge strain. Um
[64:12] But right now, things are like, you
[64:13] know, motoring along okay. Um
[64:16] and I have like the context loaded
[64:19] for both companies. I can look sort of
[64:21] see a path to a good outcome. So, I feel
[64:24] pretty good right now. Uh but there've
[64:26] been super stressful times in the past.
[64:29] Um
[64:31] and uh and then,
[64:32] you know, and then I I always try to
[64:34] reserve time for my kids cuz I love
[64:36] hanging out with them. Like
[64:38] I mean, kids are really great. I mean,
[64:39] they're like the
[64:41] 99%
[64:43] um of the time they they they they make
[64:45] you happier. They're kids are awesome,
[64:47] you know.
[64:47] Yeah. Um Then there's that 1%.
[64:50] 1% you know, like yeah, 1%. But But like
[64:53] it's it's of anything in my life, I
[64:54] would say
[64:56] kids by far make me the happiest. I
[64:58] mean, I don't know if
[64:59] you know,
[64:59] I agree. Yeah. It's great.
[65:01] I agree. Yeah.
[65:03] At this Um so so I hang out with them
[65:04] like so like now the like things like a
[65:07] lot of times kids are kind of in their
[65:08] own world, so you don't need to like
[65:10] they don't want to like talk to their
[65:10] dad for hours on end generally.
[65:12] I've noticed that.
[65:13] Yeah. Um
[65:14] so like um so I can be in the same room
[65:16] with them. They can talk to me from time
[65:18] to time, but I you know, I can get you
[65:19] know, some emails done. Just get some
[65:21] work done. And then whenever they want
[65:22] to talk to me, they can. Um
[65:25] and then uh we uh we try to do things
[65:28] like um
[65:30] you know, travel places and uh
[65:33] like I said, we we play video games
[65:35] together or uh
[65:37] um actually on Monday, we went to the
[65:40] new Harry Potter land um at Universal.
[65:42] It was quite fun.
[65:44] Yeah. So, I think that Somebody from
[65:46] Universal is clapping. Yeah.
[65:48] I mean, whoever was in charge of Harry
[65:50] Potter land did a great job. It's really
[65:52] good.
[65:52] is good. I highly recommend it. Yeah.
[65:54] The butter beer is amazing.
[65:56] Yeah, I was just saying the butter The
[65:57] butter beer is amazing. Yeah. Okay.
[66:00] Over here. Hi. Uh my name is Evan Burns
[66:02] and the founder of Odyssey and I hope in
[66:04] to the future to be in something the
[66:05] space industry. And my curiosity is
[66:08] you've talked about SpaceX getting in
[66:10] many different businesses, for example,
[66:12] um global Wi-Fi through launching many
[66:13] satellites. Um do you hope SpaceX
[66:16] becomes a platform for others to launch
[66:17] businesses or you see SpaceX being a
[66:19] business that launches many business
[66:20] lines?
[66:23] Um
[66:24] well, I mean, the general or strategy of
[66:26] SpaceX is to like we we clearly need a
[66:29] lot of money in order to develop the
[66:31] transport system to establish a city on
[66:33] Mars. Um So, you know, we're we're like
[66:36] kind of gathering revenue
[66:39] uh
[66:40] like
[66:41] ear- earth-based revenue that's
[66:43] um
[66:46] we're we're we're we're we're we're
[66:47] trying to maximize earth
[66:48] other
[66:48] earth revenue. Well, right now it's only
[66:50] earth. So, we're going to maximize earth
[66:52] revenue as it relates to space, you
[66:54] know, as it relates to rockets and
[66:55] spacecraft. So,
[66:57] um
[66:58] but I think like what assuming SpaceX is
[67:01] able to transport large numbers of
[67:02] people and and goods to Mars, it will be
[67:04] an enormous enabler for entrepreneurial
[67:07] activity on Mars. Um because there's
[67:09] going to be so much to do. Um
[67:11] you know, everything from creating like
[67:13] the first iron ore refinery to the first
[67:15] pizza joint to
[67:17] um you know, something that doesn't even
[67:18] exist on earth. Um you know, it's kind
[67:21] of like when the Union Pacific uh
[67:22] crossed crossed, you know, and like ev-
[67:25] everyone thought Union Pacific, what a
[67:26] stupid idea.
[67:27] You know, like there's nobody living in
[67:29] in California.
[67:31] Well, okay. Now, there's quite a lot of
[67:32] people living in California.
[67:34] Um so so just uh
[67:37] having a You need the transport link. Um
[67:40] and so what SpaceX is trying to do is is
[67:41] establish a transport link
[67:43] um and then try to create a fertile
[67:45] environment for entrepreneurs on Mars
[67:48] uh to flourish.
[67:50] Um and and I think that will be an
[67:51] amazing
[67:53] um expansion of entrepreneurial
[67:55] would it take to deliver a pizza from
[67:56] Mars?
[67:57] Ha. Well, it's going to be a little cold
[67:59] if it
[68:01] But I mean,
[68:03] we
[68:04] we we can certainly see a way to get to
[68:05] Mars in under three months.
[68:07] Um and I think ultimately you'll be able
[68:09] to get to Mars in under a month. It does
[68:10] get exponentially difficult as you
[68:12] reduce the time.
[68:13] Um but um
[68:14] but you know, three three months is a is
[68:16] a way to think of it. Um and I think
[68:18] that's probably
[68:20] you know, that's that's really where
[68:21] where SpaceX will I think create a a
[68:23] great
[68:24] environment for entrepreneurial
[68:26] potential.
[68:27] Thanks.
[68:29] Nilay.
[68:30] I hope Domino's does not get to Mars.
[68:33] Please don't let it. Have a special
[68:35] special Mars pizza.
[68:36] going to be I'm assuming you're going to
[68:37] be king of Mars.
[68:38] Nilay.
[68:39] Yeah.
[68:40] Uh uh so, you're obviously very
[68:41] ambitious. Um
[68:43] that's led to some really ambitious
[68:46] deadlines have been missed. So, Falcon
[68:47] Heavy was originally 2012. Um the Model
[68:50] X was a little bit delayed. The Model 3
[68:53] The Model X was delayed. The Model 3
[68:55] seems to be stretching. But the Model 3
[68:56] in particular is a consumer product.
[68:59] You're taking money from people
[69:02] against a really aggressive production
[69:04] schedule and a huge amount of orders.
[69:07] What are you going to do to hit your
[69:09] deadlines on that because it's real
[69:11] consumers this time in a big class of
[69:12] people?
[69:14] Sure. Um
[69:16] the
[69:17] uh I I think the the biggest thing is
[69:19] just designing the car for um
[69:22] for for manufacturing. So, in the case
[69:25] of Model S, like the Model S was the
[69:27] first time we'd really built a car, um a
[69:29] whole car. Like with the Roadster, Lotus
[69:31] did the body and chassis, we did the
[69:33] powertrain, and we did the sort of final
[69:34] installation of the powertrain to the to
[69:36] the chassis, but the Model S was the
[69:37] first time we made a car. So, we're just
[69:39] trying to make a a great car, um
[69:42] and but we had no idea like what it
[69:44] meant to design something to be
[69:45] manufacturable. So, the Model S is super
[69:47] hard to make, and then the Model X is
[69:49] built off of the Model S platform,
[69:51] except it's got a bunch of other
[69:52] wiz-bang technologies that make it even
[69:54] harder to build. So,
[69:56] um
[69:57] and uh you know, so
[70:00] um
[70:01] like that I mean, definitely we want to
[70:03] do the opposite of what we did with the
[70:05] with the X, um
[70:07] which is make something that is is going
[70:08] to be
[70:09] a lot simpler, um but still a car that
[70:12] people will love, um and where every
[70:15] design decision is factoring in the
[70:17] manufacturability.
[70:19] Uh in fact,
[70:20] and making sure that when we design
[70:22] something,
[70:23] um that you can manufacture it at volume
[70:25] at an affordable price in the schedule
[70:27] that we're that we're on the schedule
[70:29] that we're that we're targeting. Um
[70:32] One of the things that makes a car very
[70:34] difficult, particularly if it's a new
[70:35] car, uh is
[70:37] is that it's an integrated product with
[70:39] several thousand unique components. Um
[70:41] so, we are somewhat at the mercy of
[70:44] whatever the slowest component is,
[70:45] whatever basically I mean, if you say go
[70:48] to tier two and three suppliers, they
[70:50] end up being several thousand suppliers.
[70:52] So, so things move as fast as the least
[70:55] lucky and least competent supplier.
[70:58] Um
[70:59] you know, but but just and you can think
[71:01] of like
[71:02] like any natural disaster you care to
[71:04] name, all of those things have happened
[71:05] to our suppliers. Their factory has
[71:07] burnt down, there's been an earthquake,
[71:08] there's been a you know, tsunami,
[71:10] there's been
[71:11] uh
[71:12] massive hail,
[71:13] uh there's been a tornado,
[71:16] uh the ship sank,
[71:17] uh
[71:18] there was a shootout at the Mexican
[71:20] border,
[71:21] um no kidding.
[71:22] Um
[71:24] that that delayed trunk carpet at one
[71:26] point.
[71:27] Well, like and we couldn't get and like
[71:28] the and like the border patrol wouldn't
[71:30] give us the truck cuz it had a bullet
[71:32] holes in it. Um
[71:35] We just wanted our trunk carpet. Um
[71:38] like it's pretty safe. There's like no
[71:40] cocaine or anything. It's just a good.
[71:42] Um
[71:44] but you know, that shut down the
[71:45] production line as an example for
[71:46] several days. Um so, so there's
[71:49] that's the biggest issue is like the
[71:50] supply chain stuff is really tricky. Um
[71:53] We're trying to
[71:55] anticipate as much of that as possible,
[71:57] um increase our optionality so
[72:00] that there's more internal capability at
[72:02] Tesla. Not that we want to do things
[72:03] internally, but if um if a supplier is
[72:06] unable and unwilling to
[72:08] uh deliver the part, we can quickly make
[72:11] that internally.
[72:12] Um
[72:14] So, I think the whole company is is
[72:16] geared geared
[72:17] geared for that, um
[72:19] and um
[72:22] I mean, right now it looks like, you
[72:23] know, we should be able to do that. Uh
[72:26] we expect to I mean, we we
[72:28] almost all of the Model 3 design is
[72:30] done, um
[72:32] and we're aiming for pencils down
[72:34] basically uh
[72:37] about 6 weeks, complete pencils down. Um
[72:41] And um
[72:42] and we're tabling all, you know, like if
[72:44] there are ideas for future cool things,
[72:47] we'll we'll have it in version two,
[72:48] version three, you know, in future years
[72:50] type of thing. So,
[72:51] um overall, I feel feel pretty good
[72:52] about it. Um
[72:55] And um our supplier um particularly our
[72:57] major supplier partners have been um
[73:01] very supportive and are are on board. Um
[73:04] But um
[73:07] you know, uh I mean,
[73:08] one thing I
[73:09] I should say is like the
[73:11] like when when I when I sort of cite a
[73:13] schedule, it is actually the schedule I
[73:15] think is true. It's it's not some fake
[73:18] schedule that I don't think is true. Um
[73:21] So, I mean, uh
[73:24] you know,
[73:25] it's never until you know, I it's
[73:28] um
[73:29] maybe delusional. That is entirely
[73:32] you know, possible. Maybe it's happened
[73:34] from time to time.
[73:36] Um but it's it's it's never um
[73:39] you know, some knowingly fake deadline
[73:41] ever.
[73:43] So, is there an event in 6 weeks where
[73:45] you're going to announce autonomous
[73:46] driving is included in the pencils down
[73:47] plan for the Model 3?
[73:54] We're We're not expecting any event in 6
[73:56] weeks. All right.
[73:59] Uh Josh.
[74:00] Hi. Um so, I have a this is a kind of a
[74:03] weird question. I feel like you would be
[74:05] the guy with the right answer for it.
[74:07] There's a
[74:08] um
[74:09] sort of a philosophic concept that a
[74:11] sufficiently advanced civilization would
[74:13] be able to create uh
[74:15] So, simulation.
[74:17] Yeah. Maybe you've answered this before.
[74:19] A simulation.
[74:20] simulation discussions, it's crazy.
[74:22] Okay. Um So, cuz cuz In fact, it it got
[74:24] to the point where basically every
[74:25] conversation was was the AI
[74:27] AI/simulation conversation,
[74:30] um and my brother and I finally agreed
[74:32] that um we would ban such conversations
[74:34] if we're ever in a hot tub. That was
[74:36] like
[74:38] Yeah, cuz that really kills the magic.
[74:40] So,
[74:41] So, so the idea is, right, any
[74:43] sufficiently advanced civilization would
[74:44] create could create a simulation that's
[74:47] like our existence. And so, the theory
[74:50] follows that may maybe we're in the
[74:52] simulation.
[74:54] Have you thought about this? A lot. Are
[74:56] we
[74:59] Are we Even in hot tub. So much so it
[75:01] had to be banned from a hot tub.
[75:06] Okay. It's not the sexiest conversation.
[75:08] Are we in Are we in? Um
[75:10] The the the I mean,
[75:12] I think here's
[75:13] in my view, like the the the strongest
[75:15] argument for
[75:16] the for us being in a simulation,
[75:17] probably being in a simulation, I think
[75:19] it's the following.
[75:20] Um
[75:21] That that 40 called 40 40 years ago, we
[75:25] had Pong, like two rectangles and a dot.
[75:28] That was what games were.
[75:30] Um now, 40 years later, we have
[75:32] photorealistic 3D simulations with
[75:36] millions of people playing
[75:36] simultaneously, and it's getting better
[75:38] every year.
[75:39] And soon we'll have you know, virtual
[75:41] reality, augmented reality.
[75:44] Um
[75:45] If you assume any rate of improvement at
[75:48] all, um then the games will become
[75:51] indistinguishable from reality. Just in
[75:54] indistinguishable. Um even if that rate
[75:56] of advancement drops by a thousand from
[75:59] what it is right now,
[76:01] um then you just say, "Okay, well, we'll
[76:03] let's imagine it's a
[76:04] 10,000 years in the future." Uh which is
[76:06] nothing in the evolutionary scale.
[76:08] Um so,
[76:10] um
[76:12] So, so given that we're clearly on a
[76:14] trajectory to have games that are
[76:16] indistinguishable from reality, and
[76:18] those games could be played on any set
[76:20] up box or on a PC or whatever, and they
[76:23] would probably be
[76:24] you know, billions of such
[76:26] uh
[76:27] you know,
[76:28] computers or set up boxes.
[76:32] It would
[76:34] seem to follow that the odds that we're
[76:36] in base reality is one in billions.
[76:41] Tell me what's wrong with that argument.
[76:42] Is the answer yes?
[76:44] The argument is probably
[76:46] I just like is there Is there a flaw in
[76:47] that argument?
[76:48] someone but someone
[76:49] sure what the error in All right, no,
[76:50] no. The argument makes sense. So, the
[76:52] assumption then is that
[76:54] some we beat us to it.
[76:56] And this is a game. No, no. There's a
[76:59] one in billions chance that this is base
[77:01] reality. Oh, okay.
[77:03] What do you think?
[77:05] Well, I think it's one in billions.
[77:06] Okay.
[77:08] I mean, this that seems to be like
[77:10] clearly
[77:12] what the you know, what the what it what
[77:14] it suggests. And and actually, I mean,
[77:16] ideally we should hope that that's true
[77:18] because
[77:19] otherwise, if if civilization stops
[77:22] advancing, then that may be due to some
[77:24] cataclysmic event that erases
[77:26] civilization.
[77:27] So, maybe we should be hopeful that this
[77:29] is a simulation cuz
[77:32] otherwise Cuz they could reboot it.
[77:35] Well, otherwise
[77:36] ei- either we're going to create
[77:37] simulations that are indistinguishable
[77:39] indistinguishable from reality or
[77:40] civilization will cease to exist. Those
[77:42] are the two options.
[77:44] I like those odds.
[77:47] Okay. All right.
[77:49] We're going to it's it's unlikely to go
[77:50] into some like, you know, multi-million
[77:52] year stasis. So, it's either going to
[77:53] increase or decrease.
[77:56] Hi, I'm Jeetu Patel from Box. Uh
[77:58] two-part question for you. One is um
[78:01] if you think about fully autonomous
[78:02] vehicles, um
[78:04] which have passed through regulatory
[78:06] approvals, have passed through in-city
[78:09] driving in traffic conditions, how far
[78:11] do you think from a time frame
[78:13] perspective we are for that hap- that
[78:14] becoming reality? And number two would
[78:17] be the second part of that question is
[78:20] how far before how long before you think
[78:22] it's either illegal or extremely pro-
[78:25] prohibitively expensive for humans to
[78:26] drive on the road?
[78:31] Well, I I mean, I think I mean, I really
[78:33] would consider autonomous driving
[78:36] to be basically a solved problem.
[78:39] Um
[78:41] Even in cities like Beijing and Yeah.
[78:44] Yeah. Actually, there's
[78:46] the
[78:47] There's really only one um area where
[78:50] it's like a little dodgy, and that's
[78:51] basically if you're at roughly the 30 30
[78:54] to 40 mph
[78:55] um
[78:56] in
[78:57] in urban environments, which is that's
[78:59] difficult to achieve in Beijing. Um
[79:02] It's like heavy traffic. Uh in in in in
[79:04] dense traffic situations, uh autonomy is
[79:06] really easy um cuz you can just maintain
[79:09] a set distance from various cars. It's
[79:11] actually quite quite easy. Um you're un-
[79:13] very unlikely to drive to run anyone
[79:14] over cuz you're just not moving fast
[79:16] enough, and you can brake in time.
[79:18] Um on highways, particularly highways
[79:20] that are um
[79:22] that that are barriers so that you you
[79:24] don't have pedestrians. That's also
[79:26] relatively easy and like a Model S and
[79:28] Model X at this point
[79:30] can drive autonomously with greater
[79:32] safety than a person. Right now. Last
[79:34] question. My my point is when does it
[79:36] get to be where you don't need to be
[79:38] sitting behind a vehicle and it actually
[79:40] the the way that society starts
[79:42] expecting this is I can have my
[79:44] 75-year-old mother who doesn't speak any
[79:46] English or doesn't drive
[79:48] be able to be transported from point A
[79:50] to point B by just sitting in a car by
[79:51] herself and being taken. I know it's
[79:54] technically possible, but how far do you
[79:55] think the regulatory approvals are for
[79:57] that happening?
[79:59] I I think we're basically
[80:02] um
[80:04] less than 2 years away from complete
[80:06] autonomy.
[80:07] Wow. Wow, complete. Safer than a human.
[80:10] Um however, regulators will take um
[80:14] I think at least another year, at least
[80:16] another year and it I guess it going to
[80:18] depend on which what part of the world
[80:20] you're in.
[80:21] Um
[80:22] because they'll want to see billions of
[80:24] miles of data to show that it is
[80:26] statistically true that there is a
[80:29] substantial improvement in safety if
[80:31] something's autonomous versus not
[80:32] autonomous. I don't think that
[80:34] regulators will accept something that's
[80:35] close to
[80:37] that's that's that's sort of
[80:38] approximately as good as a person. I
[80:39] think they'll have to be
[80:41] at least twice as good as a person,
[80:43] maybe
[80:44] five or 10 times
[80:46] um
[80:47] you know, better in terms of
[80:49] uh safety.
[80:50] Um
[80:51] and and and that will have to be have to
[80:53] be a statistically relevant data set.
[80:55] So, like billions of miles over widely
[80:58] differing
[80:59] uh roads and and situations.
[81:02] Um
[81:04] So,
[81:06] yeah, you know, that So, I think it's
[81:08] like probably 3 years before it's from a
[81:10] regulatory standpoint, but less than two
[81:12] before it is
[81:14] uh technically possible.
[81:16] And do you think there's a day when it's
[81:17] illegal to drive for humans or uh you
[81:20] know, Well, I mean, we live in a
[81:21] democracy, so it's presumably that would
[81:23] be a function of the population
[81:25] deciding.
[81:26] Um
[81:29] I mean, I'm not I I mean, I'm not in
[81:30] favor of banning people from driving
[81:32] cars. Um like I'm in favor of freedom.
[81:36] Um
[81:36] and and not restricting what people do.
[81:40] Um
[81:43] Yeah, but maybe the requirements for a
[81:45] license will get more stringent. I think
[81:48] that seems like maybe a good move, you
[81:49] know, so you have to demonstrate a
[81:51] higher level of skill to drive in order
[81:54] to be allowed to manually drive.
[81:56] Okay, very last question.
[81:57] Okay, this is the last question.
[81:58] one. Sorry, we you have to go.
[82:00] has to go. You're going to make it a
[82:01] great question. All right.
[82:03] Thinking about life on Mars again, how
[82:06] do you how do you think about cultural
[82:08] unification, systems of government,
[82:11] uh rules of law, establishing those uh
[82:14] very early on?
[82:16] Well, I think I'll just declare king of
[82:17] Mars and I'm going to go.
[82:19] Um
[82:20] I like that. Yeah. Take it. Yeah, thank
[82:22] you. Thank you. Thank you.
[82:24] Um so, the the uh
[82:27] I I think most likely the form of
[82:30] government on Mars would be a democ-
[82:32] direct democracy, um not representative.
[82:35] So, it would be people voting directly
[82:37] on on issues.
[82:39] Um
[82:40] and I think that's probably better cuz
[82:42] like the potential for corruption is
[82:45] substantially diminished in a direct
[82:47] versus a representative democracy.
[82:49] Um
[82:51] so, I think that's probably what will
[82:52] occur.
[82:53] Um
[82:54] the I I think there's some
[82:57] I think so I would recommend like some
[83:00] adjustment for the inertia of laws is
[83:03] what would be wise in that it should
[83:06] probably be easier to remove a law than
[83:08] create one.
[83:10] Um
[83:11] I think you know,
[83:13] the this is I would just be like let's
[83:15] just I mean, I think I think that's
[83:16] probably probably good cuz just laws
[83:18] laws are have infinite life unless
[83:20] they're
[83:21] taken away.
[83:22] Um so, I think
[83:23] I my recommendation would be like like
[83:25] something like let's say 60% of people
[83:27] need to
[83:29] uh vote in a law, but at any point
[83:31] greater than 40% of people can remove
[83:32] it.
[83:33] Um and any law should come with a sunset
[83:36] with a built-in sunset provision. If
[83:38] it's not good enough to be voted back
[83:39] in, maybe it shouldn't be there.
[83:42] So, that's that's the framework for the
[83:44] government on Mars. I mean, those that
[83:46] would be my those would be my
[83:47] recommendations. The democ- direct
[83:49] democracy where
[83:51] where it's slightly harder to to
[83:53] put laws in place than
[83:55] to take them away and where laws don't
[83:59] just automatically live forever. You'll
[84:00] be a good king. Thank you, Elon Musk.
[84:03] Thank you.
[84:04] Thank you.
[84:06] Thank you so much.
