Caterpillar Pairs with Nvidia on Physical AI to Beat Labor Crunch

Smarter Jobsites: Coordinating People, Tools, and Machines with Caterpillar and TechCrunch — Ai4

Ai4

Autonomous equipment fails to deliver economic returns unless companies redesign jobsite workflows and upskill operators to bridge chronic labor shortages.

AI-assisted summary of Smarter Jobsites: Coordinating People, Tools, and Machines with Caterpillar and TechCrunch — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by Ai4 on August 22, 2026, this fireside chat features Caterpillar CTO Jaime Mineart and TechCrunch editorial contributor Kate Park. The discussion examines how Caterpillar is taking AI and edge computing out of digital software labs and deploying physical autonomy into heavy industrial environments like construction and mining.

## TL;DR Jaime Mineart outlines Caterpillar's trajectory from three decades of remote mining autonomy to modern physical AI powered by 1.6 million connected assets, 16 petabytes of structured data, and strategic collaboration with NVIDIA. The conversation emphasizes that the true challenge of autonomy lies in human and workflow integration rather than the underlying algorithms. Facing widespread machinery operator shortages, Caterpillar is embedding veteran expertise into machine learning models and committing $100 million over five years to workforce upskilling. Ultimately, Mineart frames Caterpillar as the essential 'invisible layer of the tech stack,' simultaneously building the physical infrastructure for global AI data centers and deploying those advanced AI tools back onto the jobsite.

## Atmosphere & style The session has a pragmatic, grounded, and forward-looking tone that connects cutting-edge computational technology with heavy industrial machinery. Kate Park’s targeted interview style allows Mineart to speak with measured executive authority and an evident respect for the hands-on craft of equipment operators. Watching the chat feels less like a speculative tech keynote and more like a sober, operational masterclass on integrating autonomous systems into complex, real-world human environments.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) Caterpillar began developing autonomous haulage more than 30 years ago to address remote mining hazards and fly-in/fly-out staffing bottlenecks. - [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) Early mining autonomy pioneered relocating operators from isolated machinery cabs into centralized metropolitan command centers, forming the foundation for modern physical AI. - [04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) Caterpillar powers its AI models with operational intelligence gathered from 1.6 million connected assets and more than 16 petabytes of structured telematics data. - [04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) A multi-layered partnership with NVIDIA provides the high-performance compute and simulation speeds required to scale manufacturing, in-cab AI assistants, and physical AI platforms. - [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) The hardest part of scaling physical AI is not the underlying technology, but integrating it into dynamic jobsite workflows and redesigning human processes to capture tangible business value. - [08:02](https://www.youtube.com/watch?v=vII4u3nfyks&t=482s) Recent advances in on-machine edge computing and physics simulations allow heavy machinery to process billions of environmental data points in real time on unstructured worksites. - [08:02](https://www.youtube.com/watch?v=vII4u3nfyks&t=482s) In-cab AI assistants are transitioning heavy equipment operators from manual machine steerers into site orchestrators who supervise multiple automated tasks simultaneously. - [10:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=604s) Severe construction labor shortages leave dozens of machines sitting idle, prompting Caterpillar to capture veteran operator intuition in AI models so novices reach baseline productivity faster. - [10:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=604s) Caterpillar committed $100 million over five years to its 'Building the Future Workforce' initiative to upskill workers in robotics, autonomy, and AI tools. - [12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) To scale AI successfully, enterprise leaders must resist technological novelty and anchor every deployment in a clearly defined customer problem with measurable outcomes. - [12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) Caterpillar acts as an 'invisible layer of the tech stack' by mining chip minerals, building physical data centers, and supplying the industrial power generation required to run digital AI. - [12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) A reinforcing physical-to-digital loop exists where heavy machinery builds and powers AI infrastructure, and the resulting AI capabilities are reinvested back into machines to coordinate safer, more productive jobsites.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) Introduction and Caterpillar's 30-Year History in Autonomy

TechCrunch contributing writer and *Signals* newsletter author Kate Park opens the late-afternoon fireside chat by welcoming the audience, noting that while artificial intelligence discussions usually center on software labs and digital models, there is an entire physical world of machines, tools, and people operating behind the scenes. She introduces Jaime Mineart, Chief Technology Officer and Senior Vice President of Cat Technology at Caterpillar, to discuss how heavy industry is deploying physical AI. Addressing the common perception of Caterpillar strictly as a traditional iron machinery manufacturer, Kate asks why the company invested in autonomous haulage more than three decades ago and how that early foundation is evolving in the modern AI era.

Mineart explains that Caterpillar’s technology strategy has always been anchored in solving acute, real-world customer problems. Mining autonomy originated thirty years ago out of urgent operational necessity: mine sites are typically remote, hazardous, and difficult to staff, historically requiring companies to fly workers in and out for multi-week rotations. By automating massive haul trucks in the early days of autonomous systems, Caterpillar removed operators from high-risk cabs and relocated them to centralized command centers located in major metropolitan areas. This shift not only eliminated severe labor constraints and significantly improved safety and comfort for workers, but also unlocked continuous productivity gains across mining operations. Mineart points out that this three-decade mining foundation is now serving as the springboard for physical AI, enabling Caterpillar to deploy edge computing and real-time planning into far more dynamic, unstructured environments like commercial construction sites and quarries.

- Mineart serves as both Chief Technology Officer and Senior Vice President for Cat Technology at Caterpillar. - Autonomous operations began over 30 years ago specifically to address remote mining conditions, fly-in/fly-out staffing bottlenecks, and extreme worksite hazards. - Early autonomy allowed operators to transition from isolated truck cabs into centralized, urban command centers.

---

[04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) The Cat AI Assistant and Strategic Partnership with NVIDIA

Transitioning to Caterpillar's newest digital developments, Park asks about the Cat AI assistant and the company’s strategic collaboration with major tech players like NVIDIA. Mineart explains that modern physical AI relies on combining massive operational data with deep machine domain expertise. Caterpillar currently monitors an expansive fleet of 1.6 million connected assets generating over 16 petabytes of high-quality structured data. By feeding this proprietary telemetry and equipment knowledge into advanced AI models, Caterpillar delivers actionable recommendations, predicts site-level risks, and guides operators directly inside the machine cab to help new workers ramp up productivity quickly.

Mineart details Caterpillar’s multi-layered partnership with NVIDIA, which spans manufacturing environments, the Cat AI assistant, and Caterpillar's broader physical AI platform. Partnering with NVIDIA provides the high-performance compute and simulation speeds necessary to run complex foundation models at scale while staying tethered to Caterpillar’s mechanical engineering and job site intelligence. She reiterates that true transformation occurs at the exact intersection of cutting-edge computational capability and specific customer workflow challenges.

- Caterpillar maintains an ecosystem of **1.6 million connected assets** globally. - The company leverages over **16 petabytes of high-quality structured data** to train its intelligent systems. - The NVIDIA collaboration covers three core pillars: factory manufacturing operations, the in-cab Cat AI assistant, and the next-generation physical AI platform.

---

[06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) Scaling Beyond Proof of Concepts: People and Process Integration

Park notes a widespread industry hurdle: while building an initial AI pilot or proof of concept is relatively easy, scaling it into daily operations proves elusive for most organizations. Mineart agrees, asserting that the hardest part of physical AI and autonomy is almost never the underlying technology itself. Instead, the true challenge lies in integrating that technology into human workflows and transforming site operations to capture concrete economic value.

"The hard part about autonomy and about physical AI is actually not the technology," Mineart observes. "It is incorporating that technology into the customer job site and into the workflows and getting the actual value out of that technology." Simply replacing a human-operated machine with an autonomous one without redesigning the surrounding ecosystem will not automatically produce "mass productivity" or superior business outcomes. Because Caterpillar has spent decades guiding heavy industry clients through operational changes, the company focuses on partnering with customers to adapt site processes and build the trust required to transition from a novelty autonomous demonstration to a fully scaled deployment.

``` [ 1.6M Connected Assets + 16 PB Structured Data ] │ ▼ [ NVIDIA Compute & Simulation Speed ] │ ▼ [ Base Machine Architecture + Edge AI ] │ ▼ ┌───────────────────────────────────────────────┐ │ Customer Site Integration │ ├───────────────────────┬───────────────────────┤ │ Workflow Redesign │ Operator-as-Director │ └───────────────────────┴───────────────────────┘ ```

---

[08:02](https://www.youtube.com/watch?v=vII4u3nfyks&t=482s) Smarter Jobsites: Dynamic Environments, Edge Compute, and In-Cab Assistance

Park shifts the conversation to the complexity of unpredictable construction environments, asking what technological shifts finally made AI feasible on unstructured job sites and what friction points remain. Mineart highlights the recent convergence of on-machine edge computing, advanced AI models, and high-fidelity physics simulations, alongside deeper integration directly into Caterpillar's base machine architecture. Modern earthmoving machines can now interpret billions of data points in real time to perceive their surroundings, make autonomous decisions, and actuate controls with extreme mechanical precision.

This setup transforms the Cat AI assistant into an ecosystem-wide copilot that bridges office management, machine sensors, and the physical cab. Rather than acting strictly as manual equipment operators, workers are evolving into orchestrators who oversee multiple automated tasks and dynamic site workflows. However, Mineart reiterates that the primary friction point remains unchanged: helping organizations execute the cultural, process, and human adjustments required to integrate these advanced capabilities seamlessly into chaotic, everyday job sites.

- Base machine architectures now integrate control systems that process billions of environmental data points at the edge. - AI systems leverage high-fidelity simulation environments to model dynamic construction variables before field deployment. - Connected digital ecosystems allow equipment operators to transition from manual steering to site orchestration.

---

[10:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=604s) Bridging the Labor Shortage and the $100 Million Workforce Initiative

Addressing the human dimension of automation, Park asks how Caterpillar ensures workers grow alongside the technology rather than getting displaced, particularly as AI elevates novices to perform closer to experienced veterans. Mineart emphasizes that operating heavy machinery is a specialized craft, praising 30-year veteran operators for their intuitive judgment and technical mastery. However, this craft faces an acute global labor shortage amid massive infrastructure demands. Mineart shares firsthand observations of visiting customer yards where as many as 30 machines sit completely idle, not due to a lack of contracted work, but simply because there is nobody available to staff the cabs.

Caterpillar bridges this gap by capturing the knowledge of veteran operators to train their AI models and algorithms, allowing novice operators to achieve high baseline productivity from their very first day on site. Meanwhile, veteran operators can transition from the physical cab into air-conditioned remote command centers, where their judgment and ingenuity can orchestrate multiple machines simultaneously. To support this workforce transformation, Mineart highlights Caterpillar's commitment to upskilling:

"Caterpillar has the initiative, 'Building the Future Workforce' initiative. We've committed $100 million over the next 5 years to that effort to upskilling and creating more capability for the future workforce so people to be comfortable with AI, with autonomy, with robotics, and so they will be able to be more skilled and not be left behind by the technology."

- Heavy construction faces widespread labor deficits where clients have up to 30 idle machines due to a lack of qualified operators. - Veteran operator domain expertise is incorporated into machine learning models to accelerate training curves for new workers. - Caterpillar committed **$100 million over 5 years** toward its **Building the Future Workforce** initiative to train personnel in robotics, autonomy, and AI operations.

---

[12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) Hard-Earned Leadership Lessons and Caterpillar's Role in the AI Tech Stack

When asked for the single most important lesson learned the hard way when scaling AI, Mineart offers clear advice for enterprise leaders: "Always start with the problem to be solved." She urges organizations not to get distracted by pure technological novelty, but to continually anchor every AI project to clear business metrics and tangible customer outcomes. Success should be judged strictly by whether the technology solves the operational challenge, after which organizations can safely iterate, deploy tools into human hands, and build institutional trust as deployments expand.

Park points out that every AI data center requires substantial physical power, positioning Caterpillar directly at the center of the broader AI hardware supply chain. Mineart references Caterpillar Chairman and CEO Joe Creed’s keynote at CES, where he described Caterpillar as building the "invisible layer of the tech stack." She outlines a full-circle dynamic: Caterpillar equipment mines the raw minerals required for advanced microchips, builds the physical data centers, and supplies the critical power generation via industrial engines and turbines. Caterpillar then takes those resulting AI tools and feeds them back into its heavy machinery to make future construction and mining sites safer, more efficient, and far more productive.

Mineart concludes that looking to the future, Caterpillar’s primary definition of success is delivering measurable customer outcomes, helping contractors complete their surging backlogs and supporting the global AI infrastructure buildout. Park thanks Mineart for her insights, and Mineart closes by thanking the audience before releasing them to the evening networking sessions.

- Core scaling philosophy: Anchor every AI and autonomy investment in an identifiable customer problem and measurable business outcome. - CES keynote theme: Caterpillar serves as the "invisible layer of the tech stack" by mining chip minerals, grading data center sites, and providing primary/backup power via industrial engines and turbines. - The AI loop: Caterpillar powers and builds the physical infrastructure for AI, then deploys those resulting AI models back into heavy equipment to coordinate people, tools, and machines.

## Notable quotes > [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) The hard part about autonomy and about physical AI is actually not the technology, — Jaime Mineart

> [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) It is incorporating that technology into the customer job site and into the workflows and getting the actual value out of that technology. — Jaime Mineart

> [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) mass productivity — Jaime Mineart

> [10:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=604s) Caterpillar has the initiative, 'Building the Future Workforce' initiative. — Jaime Mineart

> [10:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=604s) We've committed $100 million over the next 5 years to that effort to upskilling and creating more capability for the future workforce so people to be comfortable with AI, with autonomy, with robotics, and so they will be able to be more skilled and not be left behind by the technology. — Jaime Mineart

> [12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) Always start with the problem to be solved. — Jaime Mineart

> [12:12](https://www.youtube.com/watch?v=vII4u3nfyks&t=732s) invisible layer of the tech stack. — Joe Creed

> [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) Signals

## Mentioned Caterpillar, TechCrunch, Jaime Mineart, Kate Park, Joe Creed, NVIDIA, Cat AI assistant, Cat Technology, Building the Future Workforce, Signals, CES, Ai4

## Video description Jaime Mineart - CTO, Caterpillar Kate Park - Editorial Contributor, TechCrunch

AI is moving from digital tools into the physical world, where machines, data and people work together to make jobsites safer, more productive and more efficient. In this fireside chat, Jaime Mineart will share how Caterpillar is applying AI, autonomy and edge computing to real customer challenges — from labor constraints and operator support to complex workflows in mining and construction. The discussion will also explore what it takes to scale these technologies responsibly: starting with the customer outcome, integrating with existing operations, and building the right guardrails for safety, governance and trust.

[Source](https://www.youtube.com/watch?v=vII4u3nfyks)

中文简报

卡特彼勒携手英伟达布局物理AI以破解用工荒 除非企业重新设计施工现场工作流程并提升操作员技能以弥补长期劳动力短缺，否则自主设备无法带来经济回报。 Jaime Mineart 概述了 Caterpillar 从三十年前的远程采矿自动化走向现代物理 AI 的发展历程；如今这一进程正由 160 万台互联资产、16 PB 结构化数据以及与 NVIDIA 的战略合作共同驱动。对话强调，自主化运营的真正挑战不在于底层算法，而在于人员与工作流程的整合。面对普遍存在的机械操作员短缺问题，Caterpillar 正在将资深专家的经验融入机器学习模型，并承诺在五年内投入 1 亿美元用于劳动力技能提升。最终，Mineart 将 Caterpillar 定位为“技术栈中不可或缺的隐形层”，既在为全球 AI 数据中心建设物理基础设施，又将这些先进的 AI 工具重新部署回作业现场。 卡特彼勒（Caterpillar）30多年前为应对偏远矿区恶劣环境与人员短缺启动了自动运输研发，将操作员从危险驾驶舱迁入城市调度中心，为现代物理AI（Physical AI）奠定了基础。 (0:04) 基于全球160万台互联资产和超16PB结构化数据，卡特彼勒与英伟达（NVIDIA）围绕工厂制造、驾驶室内Cat AI助手及物理AI平台展开深入合作，提供高算力与仿真支持。 (4:18) 随着机载边缘计算与高保真物理仿真的突破，重型机械可实时处理数十亿环境数据点，使设备操作员从机械操控者转变为统筹多项自动化任务的现场调度者。 (8:02) 为应对部分工地多达30台设备因人手匮乏而闲置的困境，卡特彼勒将资深机手经验转化为AI模型帮助新手快速提升产能，并承诺5年内出资1亿美元用于“构建未来劳动力”技能培训倡议。 (10:04)

Reading summary

Caterpillar is expanding its three-decade foundation in autonomous mining to bring physical artificial intelligence into dynamic, unstructured construction environments. According to Chief Technology Officer Jaime Mineart, successfully scaling physical AI depends less on algorithmic capability than on adapting jobsite workflows and managing human operational transitions. From Mining Autonomy to Edge-Driven Construction (0:04) Caterpillar's autonomy initiatives originated thirty years ago to resolve operational bottlenecks in mining, where remote sites, hazardous operating conditions, and fly-in/fly-out rotations created severe staffing constraints. Automating large haul trucks allowed operators to relocate from isolated machinery cabs to urban command centers, establishing safer conditions and sustained productivity gains. Mineart explains that this mining foundation now serves as the launching pad for physical AI in less structured settings such as commercial quarries and construction jobsites. Modern earthmoving machines combine base-level machine architecture with on-machine edge computing and high-fidelity physics simulations. This setup processes billions of environmental data points in real time, shifting equipment operators from hands-on manual steering to system-level orchestrators who oversee multiple automated tasks. Proprietary Fleet Telemetry and Compute Partnerships (4:18) Developing reliable physical AI requires marrying algorithmic models with industrial domain expertise. Caterpillar monitors an active global fleet of 1.6 million connected assets, generating over 16 petabytes of structured telemetry data. This operational archive is used to train predictive models, flag jobsite risks, and feed recommendations directly into machinery cabs via the Cat AI assistant. To manage the computational demands of foundation models and dynamic physics simulations, Caterpillar established a three-part partnership with NVIDIA. The collaboration spans smart manufacturing facilities, in-cab AI guidance, and Caterpillar's broader physical AI platform. Mineart notes that while technical infrastructure is critical, the primary barrier to value realization remains integrating tools into existing workflows rather than deploying standalone autonomous pilots. Workforce Upskilling Amid Widespread Operator Deficits (10:04) The construction sector faces an acute global labor shortage, with Mineart observing customer equipment yards where up to 30 machines sit idle solely due to an absence of qualified personnel. Rather than displacing labor, Caterpillar positions automation as an operational necessity to sustain productivity across major infrastructure backlogs. To bridge this labor gap, the company captures the craft and procedural knowledge of veteran operators with 30 years of experience, encoding their techniques into machine learning models. This enables novice operators to achieve baseline proficiency from their first day, while experienced personnel can transition into remote command centers managing multi-machine deployments. To facilitate this shift, Caterpillar committed $100 million over five years to its 'Building the Future Workforce' program, aimed at training personnel in robotics, AI, and autonomous systems. Powering the Physical Infrastructure of AI (12:12) Caterpillar operates on both sides of the modern artificial intelligence lifecycle. Referencing Chairman and CEO Joe Creed's address at CES, Mineart highlights Caterpillar's role as the 'invisible layer of the tech stack.' The company's heavy equipment extracts the minerals required for microchips, excavates foundations for data centers, and provides critical backup and primary power through industrial engines and turbines. This physical infrastructure directly powers the digital cloud models that Caterpillar subsequently feeds back into its field machinery. For enterprise leaders deploying AI, Mineart advises against chasing technical novelty for its own sake, emphasizing that every investment must anchor strictly to measurable operational problems and customer outcomes. Mineart concludes that the ultimate measure of success for physical AI is delivering concrete customer productivity gains to clear mounting project backlogs. However, unlocking this value remains fundamentally contingent on whether organizations can successfully execute the cultural and process redesigns necessary to integrate automated machinery into day-to-day operations.

精炼阅读摘要

卡特彼勒（Caterpillar）首席技术官兼技术业务高级副总裁Jaime Mineart指出，重工业领域的“物理AI”与自主化核心不在于纯技术堆砌，而在于将其与现场作业流程和人员转型深度融合，以解决极端工况与全球劳动力短缺问题。 矿山自主化积淀奠定物理AI基石 (0:04) 卡特彼勒的自主化技术布局始于30年前，其出发点并非追求技术概念，而是解决矿业客户面临的极端现实痛点。采矿作业区通常地处偏远、环境危险且招工困难，传统模式下企业必须依赖高成本的轮岗飞行班次运送工人。卡特彼勒通过对超大型非公路自卸车实施自动化改造，将操作员从高风险的现场驾驶室撤离，集中安置到大城市的远程指挥中心。这种调整不仅消除了严苛的用工瓶颈并保障了工人安全舒适，更带来了采矿作业连续生产率的显著提升。 Mineart强调，长达三十年的矿山自主化积累，现已成为卡特彼勒拓展“物理AI”的技术跳板。与结构相对规范的露天矿区相比，商业建筑工地和采石场更具动态性和非结构化特征。依托早期的控制工程沉淀，卡特彼勒正逐步将边缘计算、实时感知规划及精确实时控制扩展至更复杂的现实作业场景中。 数据资产与NVIDIA算力生态协同 (4:18) 物理AI的实际效能高度依赖海量工业运营数据与机械工程专业知识的深度融合。Mineart介绍，卡特彼勒当前在全球范围内监测着超过160万台互联资产，这些设备持续生成了超过16 PB的高质量结构化遥测数据。通过将专有设备数据与深厚工况知识注入先进算法模型，卡特彼勒不仅能为机队提供现场风险预测与运营建议，还能在驾驶室内直接辅助新手操作员，缩短其掌握复杂工程技能的周期。 为支撑大规模物理AI落地，卡特彼勒与NVIDIA展开了多层次战略合作，重点覆盖工厂制造环境、驾驶室内Cat AI助手以及下一代物理AI平台三大支柱。NVIDIA提供的高性能算力与高保真物理仿真能力，使卡特彼勒能够在兼顾机械工程机理与工地实况的前提下运行复杂基座模型。Mineart重申，真正的行业变革仅发生在顶尖计算能力与具体客户工作流痛点精准对接的交汇点上。 突破试点瓶颈：现场流程与指挥者重塑 (6:11) 行业内普遍面临“概念验证容易但规模化推广极难”的困境。Mineart明确指出，物理AI和自主化最艰难的部分往往并非技术本身，而是如何将技术无缝整合到客户工地的日常作业流程中并实现明确的经济回报。如果仅仅是将有人驾驶机械替换为自动化设备，而不对周边的作业生态和协同流程进行重构，并不会自然带来生产率的大幅飞跃或优异的商业结果。 借助在机械架构层集成的边缘计算能力，现代工程设备能够在几毫秒内解析数十亿个环境数据点并完成精确控制，这促使操作人员的角色发生了根本性改变。工人正在从单纯的手动操纵设备者，演变为监督多项自动化任务和动态现场工序的“现场指挥调度者”。卡特彼勒强调，推动客户企业进行文化适应与管理再造，是在动态环境下建立人机信任、实现规模化常态运营的关键条件。 缓解用工荒与一亿美元未来劳动力倡议 (10:04) 重型机械操作是一门高度依赖经验与技术直觉的专门技艺，但全球建筑与基础设施领域当前正遭遇严重的熟练工短缺。Mineart分享了其在客户现场的真实见闻：由于缺乏合格的操作人员，客户工地上有时会有多达30台重型设备处于完全闲置状态，即便手头拥有充足的工程合同也无法开工。劳动力危机已成为制约重工业产能扩张的直接瓶颈。 为化解这一结构性难题，卡特彼勒将具备三十年工作经验的资深机手知识提取并转化为算法模型，使初学者在进入驾驶室的第一天就能达到基准生产效率，同时让经验丰富的专家转入远程指挥中心统筹多台设备。此外，卡特彼勒设立了“建设未来劳动力”（Building the Future Workforce）倡议，承诺在未来5年内投入1亿美元，专门用于技术升级与技能培训，帮助一线人员掌握AI、自主化和机器人系统的协作能力，防止技术迭代造成从业人员脱节。 贯通AI底层基础设施与工程应用闭环 (12:12) Mineart向企业管理者建议，扩展AI规模最深刻的教训是“必须始终以待解决的问题为起点”，切忌沉迷于单纯的技术新奇感，每个项目都必须锚定明确的业务指标和客户成果。只有在技术切实解决了现场运营阻碍后，企业才能在迭代中将工具交付给一线工人并逐步建立组织层面的信任机制。 卡特彼勒在整个人工智能供应链中还承担着双重基础设施角色。正如卡特彼勒董事长兼首席执行官Joe Creed在CES主题演讲中所阐述，公司构成了“技术栈的隐形层”：卡特彼勒机械负责开采微芯片所需的关键矿产原材料、平整大型AI数据中心地基，并通过工业发动机与燃气轮机提供主要及备用电力保障。由此，卡特彼勒不仅在物理层面上建造并支撑了AI基础设施，更将生成的先进AI模型反哺至重型工程装备中，形成了驱动实体工程进化的闭环体系。 Mineart总结指出，卡特彼勒衡量AI成功的最终标准是能否为客户带来可衡量的实际商业成果，并缓解施工积压与基建需求。然而，面对高度无序、动态且碎片化的真实工地环境，如何在庞大的存量机械生态中全面完成组织协同与作业文化重构，依然是物理AI从技术愿景走向全面商业兑现所必须面对的长期课题。

Detailed notes — zh

## Background 本内容由 YouTube 频道 Ai4 于 2026 年 8 月 22 日发布，记录了 TechCrunch 特约撰稿人 Kate Park 与卡特彼勒（Caterpillar）首席技术官兼 Cat Technology 高级副总裁 Jaime Mineart 之间的一场炉边对话。视频简介指出，内容聚焦于 AI 如何从数字工具迈向物理世界，探讨卡特彼勒如何将 AI、自主作业与边缘计算应用于矿山与建筑施工现场，解决劳动力短缺并负责任地实现规模化落地。 ## TL;DR 对话深入探讨了卡特彼勒三十年来在矿山与建筑场景中应用物理 AI、边缘计算和自主机械的实践经验。卡特彼勒依托全球 160 万台联网设备与 16 PB 高质量数据，携手 NVIDIA 打造驾驶舱 Cat AI Assistant，将现场机手升级为全局调度者，并设立 1 亿美元基金用于未来劳动力技能重塑。Jaime Mineart 强调，推动 AI 规模化的铁律是始终从客户的实际业务痛点出发，同时揭示了卡特彼勒作为 AI 数据中心基建与电力底座的闭环支撑价值。 ## Atmosphere & style 整场对话专业、务实且充满工业底蕴，打破了传统科技会议纯粹讨论算法模型的虚浮感，将焦点牢牢锚定在重工现场的现实挑战上。主持人 Kate Park 的提问精准敏锐，直击 POC 落地难和全球招工荒等行业痛点；CTO Jaime Mineart 的回应则沉稳笃定、兼具一线洞察与宏观格局，尤其在阐述卡特彼勒作为 AI 算力‘隐形底层’时展现出强烈的工业自豪感，整体氛围发人深省且极具启发性。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) AI 正在从纯数字软件工具深入重工业的物理世界（Physical AI），结合边缘计算与自主设备解决严峻的现场生产力挑战。 - [01:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=85s) 卡特彼勒早在 30 年前便开启矿区自主化研发，核心目的是将工人从高危偏远的恶劣驾驶舱解放至远程指挥中心，从根本上解决用工荒并保障安全。 - [01:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=85s) 矿山封闭场景积累的自主作业与安全调度经验，正随着物理 AI 和边缘算力的成熟，进一步推广至环境更动态、更复杂的采石场与建筑施工现场。 - [04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) 卡特彼勒在全球拥有超 160 万台互联设备资产，沉淀了超过 16 PB 高质量工业数据，结合深厚机理知识驱动可落地执行的 Cat AI Assistant。 - [04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) 卡特彼勒与 NVIDIA 展开全面合作，涵盖智能制造、驾驶舱 AI 助手以及利用 NVIDIA 软硬件生态构建下一代物理 AI 平台。 - [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) 跨越概念验证（POC）实现规模化落地的难点在于工作流重构，单纯撤换无人设备无法直接提升生产力，必须协助客户完成组织与业务流程的深层变革。 - [08:01](https://www.youtube.com/watch?v=vII4u3nfyks&t=481s) 建筑工地的动态多变性通过车载边缘算力、AI 模型与高保真物理仿真的底层硬件级集成得以征服，使机械具备毫秒级实时感知与高精度控制能力。 - [08:01](https://www.youtube.com/watch?v=vII4u3nfyks&t=481s) 在 AI 助手与自主设备的辅助下，一线操作员的角色正从传统的繁重驾驶员转变为整个工地的“全盘编排调度者”（Orchestrators）。 - [10:05](https://www.youtube.com/watch?v=vII4u3nfyks&t=605s) 将几十年经验的老机手驾驶技巧与判断力训练进 AI 模型，不仅让新手第一天即可高效作业，也让资深工人得以转向远程决策岗位。 - [10:05](https://www.youtube.com/watch?v=vII4u3nfyks&t=605s) 卡特彼勒设立“建设未来劳动力计划”，承诺在 5 年内投入 1 亿美元，专门用于提升劳动者在 AI、自主化与机器人技术上的应用技能。 - [12:13](https://www.youtube.com/watch?v=vII4u3nfyks&t=733s) 推动 AI 规模化落地的首要铁律是“永远从待解决的问题出发”，必须紧扣客户预期的业务成果而非被技术本身所迷惑。 - [13:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=805s) 卡特彼勒构成了 AI 算力时代的“隐形底层技术栈”：从开采芯片矿物、建造数据中心到提供电网与动力系统，再反哺应用先进 AI 赋能机械，形成全周期闭环。 - [14:49](https://www.youtube.com/watch?v=vII4u3nfyks&t=889s) 卡特彼勒在自主化与 AI 布局上的终极目标，是为面临巨大工程交付压力的客户带来实际业务成果与生产力形态的实质性变革。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) 开场介绍：物理世界中的 AI 与重工业赋能 在大会尾声的炉边对话中，TechCrunch 特约撰稿人、《Signals》通讯主理人 Kate Park 担任主持人，向与会者介绍了本场对话的背景：当大众谈论人工智能时，往往首先联想到软件算法或前沿 AI 实验室，但在数字世界背后，存在着一个由人员、工具和重型机械构成的广阔物理世界。Caterpillar（卡特彼勒）首席技术官兼 Cat Technology 高级副总裁 Jaime Mineart 随后登台，共同探讨物理 AI（Physical AI）、边缘计算与自主作业如何深入重工与建筑施工场景，解决极其严峻的现实生产力挑战。 --- ### [01:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=85s) 矿山无人化三十年：从偏远作业到动态物理 AI 针对卡特彼勒为何在 30 多年前就远超行业涉足自主卡车研发，Jaime Mineart 强调，卡特彼勒的一切技术投入始终锚定在解决客户的真实世界痛点上。矿区环境普遍极其偏远、恶劣且伴随高危风险，传统模式下企业必须将操作工通过飞机轮班运送至矿场驻留数周，面临巨大的人员调度与用工短缺压力。 卡特彼勒早在 30 年前便开始探索矿山自主作业，核心目标是将操作人员从重型自卸卡车的驾驶室中解放出来，转移到环境舒适且更安全的远程指挥中心进行统筹调度。这一演进不仅显著提升了人工作业的安全感与作业舒适度，还大幅消除了偏远地区的用工荒，同时拉升了整个矿区的综合产出效率。随着当前物理 AI 技术与边缘计算能力的成熟，卡特彼勒正将在矿山领域积累的成熟经验推广至更为动态、复杂的采石场和建筑施工工地。 ```mermaid flowchart TD accTitle: 卡特彼勒自主技术演进路径 subgraph Phase1["三十年前：矿区封闭自主作业"] A["高危/偏远矿区现场"] --> B["自卸卡车操作员撤出驾驶室"] B --> C["城市远程指挥中心统筹调度"] end subgraph Phase2["当前：动态物理 AI 与边缘计算"] D["复杂建筑施工/采石场"] --> E["高保真仿真与边缘算力"] E --> F["动态环境实时感知与快速路径规划"] end Phase1 --> Phase2 ``` --- ### [04:18](https://www.youtube.com/watch?v=vII4u3nfyks&t=258s) 16 PB 高质量数据与 Cat AI Assistant 的协同突破 谈及近期推出的 Cat AI Assistant 以及与 NVIDIA 的深度合作，Jaime Mineart 解释道，物理 AI 的落地依赖技术能力与行业专有知识的交汇。卡特彼勒在全球拥有超过 160 万台联网设备资产，沉淀了超过 16 PB 的高质量结构化工业数据。将这些庞大的数据资产与卡特彼勒对重型设备和施工现场的深厚机理知识相结合，能够生成可直接指导操作的行动洞察与风险排查建议。 通过将 Cat AI Assistant 引入设备驾驶舱，新入职的操作员能够迅速上手并高效操作机械。在与 NVIDIA 的合作中，双方不仅在制造环节和 AI 助手上协同，还在物理 AI 平台层面全面展开合作，借助 NVIDIA 的软硬件生态为卡特彼勒的行业专业经验赋予极高的算力扩展速度与计算规模。 - **联网资产规模**：全球超 160 万台互联设备。 - **工业数据储备**：超 16 PB 高质量结构化作业与机械数据。 - **与 NVIDIA 合作范畴**：涵盖智能制造、驾驶舱 Cat AI Assistant 以及下一代物理 AI 平台构建。 --- ### [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) 跨越概念验证（POC）：从单机自动化到工作流重构 针对“人人都可做 AI 原型，但极少数人能实现规模化落地”的行业普遍困境，Jaime Mineart 指出，物理 AI 与自主技术最艰难的部分其实并非技术算法本身，而是如何将技术无缝融入客户现有的工地作业场景与工作流中，以真正释放商业价值。 "单单撤下一个人工操作的机械，换上一台自动驾驶机械，并期望它能直接带来生产力的爆发，这种想法是极其不切实际的。" 卡特彼勒凭借三十年来的经验积累，核心优势在于协助客户经历组织流程与人员协同的深层变革，准确定位生态链中能产生颠覆性价值的节点，从而跨越单纯的技术演示，与客户建立起基于实际业务成果的深度信任。 --- ### [08:01](https://www.youtube.com/watch?v=vII4u3nfyks&t=481s) 征服不可预测的施工现场：边缘计算与底层架构集成 建筑施工现场相比矿山更为不可预测且动态多变，传统上被视为 AI 最难落地的行业之一。Jaime Mineart 表示，当前技术之所以迎来爆发拐点，关键在于边缘端计算能力、现代化 AI 模型与高保真物理仿真的成熟结合。卡特彼勒直接在基础机械硬件架构层面深度集成了感知与控制系统，使重型机械具备了极高的控制精度。 当下的智能工程机械能够实时采集、处理并解析数十亿个数据点，使其在极为动态复杂的工地上具备自主“观察与行动”的能力。结合覆盖全数字生态并延伸至驾驶室的 Cat AI Assistant，现场作业人员的角色正在转变为工地的“全盘编排调度者”（Orchestrators），彻底改变了传统施工的操作形态。 ```mermaid flowchart LR accTitle: 施工现场动态物理 AI 闭环架构 subgraph Sense["感知与边缘处理"] A["工地现场数十亿数据点"] --> B["车载边缘计算与 AI 模型"] end subgraph PlanControl["规划与机械控制"] B --> C["高保真物理仿真与控制系统"] C --> D["基础机械高精度物理执行"] end subgraph Human["人机协同"] E["驾驶室 Cat AI Assistant"] --> F["操作人员升级为全场调度者"] F -.-> C end ``` --- ### [10:05](https://www.youtube.com/watch?v=vII4u3nfyks&t=605s) 应对劳动力断层：经验传承与一亿美元未来人才计划 面对全球重工行业严峻的用工短缺现象，Kate Park 提及在日美等国均存在大量工程由于缺少机手而被迫搁置的情况。Jaime Mineart 证实了这一点：“我走访客户时，常看到场地里停着 30 台机器闲置，并非因为没有工程需求，而是根本招不到操作员。”与此相对的是，行业内拥有三十年驾龄的资深老机手，其高超的操作技巧堪称一门手艺。 为了避免操作人员在技术浪潮中被边缘化，卡特彼勒将资深机手的专业经验与判断力用于训练 AI 模型，使得新手操作员在辅助下第一天即可具备高效作业能力。与此同时，老机手得以从繁重的座舱驾驶转入远程控制站，发挥其统筹与复杂决策价值。 为了系统性推动劳动力技能重塑，卡特彼勒设立了“建设未来劳动力计划”（Building the Future Workforce Initiative），承诺在未来 5 年内投入 1 亿美元，用于提升劳动者在 AI、自主化和机器人技术领域的技能水平，确保一线工人不会被技术革命落下。 --- ### [12:13](https://www.youtube.com/watch?v=vII4u3nfyks&t=733s) 规模化 AI 的沉痛教训：永远以业务痛点为基准 当被问及在推动 AI 规模化过程中吸取的深刻教训时，Jaime Mineart 直言不讳地指出，核心铁律只有一条：“永远从待解决的问题本身出发（Always start with the problem to be solved）。” 在各种令人眼花缭乱的技术可能面前，企业极易迷失方向。技术团队必须时刻反躬自省：具体的商业问题是什么？客户期望达成的业务结果是什么？以此作为衡量项目成功与否的唯一标准。只有在明确目标后，才能将工具交付给一线人员进行现场学习与迭代，并通过不断规模化来建立对新技术的深厚信任。 --- ### [13:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=805s) 支撑 AI 时代的隐形基石：卡特彼勒的“全周期闭环” 谈及 AI 数据中心对电力基础设施的爆发式需求，Jaime Mineart 援引了公司在 CES 上的主题演讲内容指出，卡特彼勒构成了整个科技技术栈的“隐形底层”（the invisible layer of the tech stack）： 1. **矿产开采**：使用卡特彼勒机械开采制造芯片和电子元件所需的关键矿物。 2. **基础设施建设**：直接参与全球大型数据中心与电网的实体建设施工。 3. **电力能源保障**：通过旗下的大型发动机、燃气轮机和动力系统，为 AI 数据中心提供关键的备用与主干电力支持。 这一闭环极富戏剧性：卡特彼勒为支撑全球 AI 算力的基础设施提供基础保障与能源动力，随后再反哺利用这些先进 AI 算法与芯片能力，为全球工地打造更安全、更高产的智能机械。 --- ### [14:49](https://www.youtube.com/watch?v=vII4u3nfyks&t=889s) 展望未来：推动客户现场产生实质性变革 在对话的尾声，Kate Park 询问卡特彼勒在 AI 与自主化道路上最希望做对的一件事。Jaime Mineart 总结道，卡特彼勒最核心的目标是“确保客户业务成果的实质性变革”。 无论是前沿技术研发还是组织流程变革，一切投入的终极目的都是赋能客户。当前客户面临着远远超出其实际交付能力的工程需求，卡特彼勒希望通过 AI、边缘计算与自主系统，将施工现场打造成更安全、更高产、更高效的作业环境，助力全球 AI 基础设施建设与大型工程的顺利落地。伴随着观众席的热烈掌声，对话圆满结束。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=vII4u3nfyks&t=4s) 在数字世界背后，存在着一个由人员、工具和重型机械构成的广阔物理世界。 — Kate Park > [06:11](https://www.youtube.com/watch?v=vII4u3nfyks&t=371s) 单单撤下一个人工操作的机械，换上一台自动驾驶机械，并期望它能直接带来生产力的爆发，这种想法是极其不切实际的。 — Jaime Mineart > [08:01](https://www.youtube.com/watch?v=vII4u3nfyks&t=481s) 现场作业人员的角色正在转变为工地的“全盘编排调度者”（Orchestrators）。 — Jaime Mineart > [10:05](https://www.youtube.com/watch?v=vII4u3nfyks&t=605s) 我走访客户时，常看到场地里停着 30 台机器闲置，并非因为没有工程需求，而是根本招不到操作员。 — Jaime Mineart > [12:13](https://www.youtube.com/watch?v=vII4u3nfyks&t=733s) 永远从待解决的问题本身出发（Always start with the problem to be solved）。 — Jaime Mineart > [13:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=805s) 卡特彼勒构成了整个科技技术栈的“隐形底层”（the invisible layer of the tech stack）。 — Jaime Mineart > [13:25](https://www.youtube.com/watch?v=vII4u3nfyks&t=805s) 为支撑全球 AI 算力的基础设施提供基础保障与能源动力，随后再反哺利用这些先进 AI 算法与芯片能力，为全球工地打造更安全、更高产的智能机械。 — Jaime Mineart > [14:49](https://www.youtube.com/watch?v=vII4u3nfyks&t=889s) 确保客户业务成果的实质性变革。 — Jaime Mineart ## Mentioned Jaime Mineart, Kate Park, Caterpillar (卡特彼勒), TechCrunch, Signals, Cat Technology, Cat AI Assistant, NVIDIA, Building the Future Workforce Initiative (建设未来劳动力计划), CES, Ai4 ## Video description Jaime Mineart - CTO, Caterpillar Kate Park - Editorial Contributor, TechCrunch AI is moving from digital tools into the physical world, where machines, data and people work together to make jobsites safer, more productive and more efficient. In this fireside chat, Jaime Mineart will share how Caterpillar is applying AI, autonomy and edge computing to real customer challenges — from labor constraints and operator support to complex workflows in mining and construction. The discussion will also explore what it takes to scale these technologies responsibly: starting with the customer outcome, integrating with existing operations, and building the right guardrails for safety, governance and trust. [Source](https://www.youtube.com/watch?v=vII4u3nfyks)

Source quotes

[0:04] Kate Park: 在数字世界背后，存在着一个由人员、工具和重型机械构成的广阔物理世界。 [6:11] Jaime Mineart: 单单撤下一个人工操作的机械，换上一台自动驾驶机械，并期望它能直接带来生产力的爆发，这种想法是极其不切实际的。 [8:01] Jaime Mineart: 现场作业人员的角色正在转变为工地的“全盘编排调度者”（Orchestrators）。 [10:05] Jaime Mineart: 我走访客户时，常看到场地里停着 30 台机器闲置，并非因为没有工程需求，而是根本招不到操作员。 [12:13] Jaime Mineart: 永远从待解决的问题本身出发（Always start with the problem to be solved）。 [13:25] Jaime Mineart: 卡特彼勒构成了整个科技技术栈的“隐形底层”（the invisible layer of the tech stack）。 [13:25] Jaime Mineart: 为支撑全球 AI 算力的基础设施提供基础保障与能源动力，随后再反哺利用这些先进 AI 算法与芯片能力，为全球工地打造更安全、更高产的智能机械。 [14:49] Jaime Mineart: 确保客户业务成果的实质性变革。

Source quotes

[6:11] Jaime Mineart: The hard part about autonomy and about physical AI is actually not the technology, [6:11] Jaime Mineart: It is incorporating that technology into the customer job site and into the workflows and getting the actual value out of that technology. [6:11] Jaime Mineart: mass productivity [10:04] Jaime Mineart: Caterpillar has the initiative, 'Building the Future Workforce' initiative. [10:04] Jaime Mineart: We've committed $100 million over the next 5 years to that effort to upskilling and creating more capability for the future workforce so people to be comfortable with AI, with autonomy, with robotics, and so they will be able to be more skilled and not be left behind by the technology. [12:12] Jaime Mineart: Always start with the problem to be solved. [12:12] Joe Creed: invisible layer of the tech stack. [0:04] Signals

Original transcript

[0:04] Hi everyone. Good afternoon. All right. Good  afternoon everyone. I know we're like end of the [0:10] day so this is good. This will be be a fun session  discussion. Yeah. Thanks for joining today. Uh for [0:18] those who don't know me, my name is Kate Park.  Um, I'm a contributing writer for TechCrunch [0:23] and I also run my own newsletter called Signals  focused on AI and technology across a APAC Asia [0:31] Pacific and beyond. Okay. So, um, I'll be your  moderator for the next 20 minutes. Uh I'm very [0:39] excited about this one because when we talk about  AI we usually mean AI labs or like software but um [0:47] there's a whole physical world behind it like of  people tools and um machines right yeah so you're [0:55] going to talk about Yeah he's she's going to talk  about that so this will be conversation with Jamie [1:00] Minard CTO at Caterpillar so Jamie thanks for  being here yeah thanks for having me and it's [1:07] so great to to get to share this topic with you.  Yeah. Could you please introduce uh yourselves to [1:12] the audience? Absolutely. So, I'm Jamie Miner. I'm  the chief technology officer for Caterpillar and [1:17] I am the senior vice president for cat technology  for our organization. Thank you. Perfect. Uh let's [1:24] dive in. Yeah, let's do it. Yes. So, uh my first  question is uh most people picture Caterpillar [1:33] uh and think a machinery company not AI, right?  Uh and but you started your company autonomous [1:38] m machine uh autonomous trucks over 30 years  ago. Uh what made you invest so early ahead of [1:46] people everyone and um how is that changing in the  era age of AI? Yeah, it's it's such an exciting [1:55] opportunity and we have been in autonomy for as  you said over 30 years and everything that we do [2:02] at Caterpillar is anchored on solving real world  problems, real customer challenges and the reason [2:09] that we started 30 years ago in mining autonomy  because there was an urgent customer need. You all [2:15] uh may not live and work on mine sites or even  think about mine sites on a regular basis, but [2:20] it doesn't take a lot of imagination to understand  that mine sites tend to be very remote. They tend [2:26] to be hazardous environments. They may be hard to  staff. And so 30 years ago, as technology became [2:34] more capable, we really leaned into autonomy in  the very early days to help our customers solve [2:40] those challenges, to be able to take operators  out of the cab of these hall trucks and be able to [2:47] put them in remote command centers where they can  orchestrate the mindsite and do that autonomously. [2:54] And the the as we went on that evolution, we went  on that journey with our customers, one of the [3:00] things that was really exciting and about that  is now these remote mine sites where you had to [3:06] fly labor in and they had to stay for weeks at a  time and and fly labor back out and so forth. Now [3:12] you can put those labor forces in major centers  and they can have much more comfortable jobs, [3:18] certainly safer jobs, and you can make those  mine sites more productive. So the the business [3:25] benefits, the human benefit is certainly high.  Uh and you if you eliminate some of that labor [3:32] scarcity and labor challenges that those customers  um faced at that time. Now, we're at this super [3:40] exciting time where the uh physical AI capability  and the ability to take all of that learning that [3:48] we did in mining and bring that into much more  dynamic environments, much more dynamic sites, [3:56] job sites, quaries, and uh construction sites and  so forth, and be able to uh bring in compute at [4:04] the edge, be able to have the machines adapt apt  and plan much more quickly and be able to do a [4:12] lot more with the autonomy technology at scale.  Great. Thank you. And uh moving on to the next [4:18] question like you've had autonomy for decades that  you mentioned and but this the latest products the [4:25] kayaki assistant and you're working with big tech  companies like Nvidia uh feels like a different [4:31] chapter right uh what changed can you tell me  about uh can you can you tell me more details [4:36] about that corrosion and collaboration and  what changed yeah I think again it's about the [4:41] intersection of what's possible with technology  and the customer challenge to be solved and So, [4:47] uh, the the CAT AI assistant really is built on  all of that foundation and knowledge that we have. [4:56] We have 1.6 million connected assets. We have over  16 pabytes of high quality structured data that we [5:05] can combine with the knowledge and expertise that  we have of our machines and our customer job sites [5:11] to be able to serve up actionable insights and  recommendations to help our customers be more [5:16] productive, to be able to troubleshoot maybe um  areas of their job sites that are higher risk, [5:23] be able to help them run their their business  more effectively. we can bring that a that CAT [5:29] AI assistant into the cab of machines and be  able to help the the labor that is coming on [5:36] board be able to be more effective with those  machines much faster. That relationship that [5:41] we have with Nvidia, we've partnered with them  in our manufacturing environment. We partnered [5:46] with them and our CAT AI assistant and we are  now also partnering with them on our physical [5:52] AI platform. the ability that they have to uh  bring speed and scale to our knowledge and our [6:00] understanding of our machines and our job sites  just can be transformational for our customers. [6:05] So it's all about the the that intersection of the  technology and the customer problem to be solved. [6:11] Awesome. Yeah. And everyone can build AI pilot but  I think very few can scale it right. Uh what made [6:20] you what made the difference for Caterpillar? Can  you give me some use cases or Yes. Yes. I mean, [6:25] we talked about the the mining autonomy and we  we did that over three decades of evolving that [6:32] technology and scaling that with our customers.  And you know, I think you and I have even talked [6:37] about this. The hard part about autonomy and about  physical AI is actually not the technology. it is [6:45] incorporating that technology into the customer  job site uh and into the workflows and getting [6:51] the actual value out of that technology. So we  we have decades of experience doing this. We [6:58] partner with our customers and we help them go  on that journey to incorporate the technology [7:03] into their work sites into their workflows to  be able to understand okay what are the parts [7:09] of your workflow and the parts of your ecosystem  this technology can really create transformational [7:15] change and create value and create outcomes. So,  uh, just building on that people in process change [7:22] that we know know how to do for decades to take it  from an interesting idea or an autonomous machine, [7:30] that's that's pretty cool. I mean, it's pretty  it's pretty cool to see an autonomous machine, [7:34] but if you take a staff machine out and you  put an autonomous machine in and you think [7:38] you're just going to get bass mass productivity or  you're going to get great outcomes, you're really [7:44] you're really not thinking about what it's really  going to take. And that's the other thing that we, [7:49] you know, that's the thing that we bring to it  when you talk about these proof of concepts to [7:53] being able to actually build trust and scale  with our customers. We have that experience to [7:57] be able to help them. Okay, great. Yeah. And uh  this session is about smarter job site and coord [8:03] coordinating people tools and machines. And I  think construction sites are unpredictable and [8:11] the industry is very hard to you know sometimes  like a hard to adopt AI technology. So what [8:18] finally made AI possible and what's still hard  you know today? Yeah, the I think what is what [8:26] it makes this become more and more possible is the  the compute capability on the edge, the AI models, [8:33] the highfidelity simulation that's that's  available. We are incorporating more and more of [8:38] the technology into our base machine architecture.  So you can really control machines with the [8:45] precision that is needed to be able to operate in  that dynamic environment. So now machines can take [8:52] in process interpret billions of data points to be  able to see and act uh in in the job site. And so [9:01] that combination of you know the intelligence we  have, the data that we have, the understanding of [9:06] the and the knowledge of the machines themselves  makes this this a unique time where you can put [9:13] these large machines in construction sites that  are very dynamic and they can they can see and [9:19] they can adapt and they can operate in a way  that is it's just unique to today's technology. [9:26] Um, we we talked a little bit about the the  NVIDIA relationship and we talked about the CAD [9:32] AI assistant that runs across our entire digital  ecosystem. We'll bring it into the cab. All of [9:38] those things connecting to then allow operators  and allow the labor to now be orchestrators of [9:45] the job site. That that technology capability has  really changed the game. Now, what's still hard, [9:53] we talked about it, right? the people and process  changes to incorporate that technology into [9:58] the job site and help our customers on that  journey. And uh I'm wondering like if your if [10:05] your technology helps new operators perform much  more like uh experienced ones um as AI takes some [10:12] more of the work, how do you make sure your people  growing grow with it um not get left behind? Yeah, [10:19] that's a great question. Um, we have operators  that are 30-year veterans. They're incredible at [10:26] their craft. I mean, this is a craft. The their  ability to operate machines in a an environment [10:32] and be able to get productivity out of it. At the  same time, there is a huge labor shortage. There [10:37] is more work to be done with AI infrastructure  buildout than there are operators to be able to do [10:42] it. You've talked about it and written about it in  Japan. We see it here in the US. I visit customers [10:47] and they have 30 machines that are sitting because  they can't staff the machines, not because there's [10:52] not demand for the work to be done. And so using  that uh capability for those 30-year veterans to [11:00] train the the AI models, to train the intelligence  to to to help then those novice operators come [11:08] onto the job site and be productive from day one  using the technology. The 30-year veterans can [11:15] also move into a an orchestrator role. They can  move out of the cab and into a remote operator [11:22] station or a remote command center and use that  ingenuity, that judgment, that human capability [11:29] to be able to get more productivity out of these  job sites. Now, with all of that said, we I won't [11:36] be dismissive of that journey for upskilling the  the labor of the future. And so Caterpillar has [11:43] the initiative, building the future workforce  initiative. We've committed $und00 million over [11:49] the next 5 years to that effort to upskilling and  creating more capability for the future workforce [11:59] or people to be comfortable with AI, with  autonomy, with robotics, and so they will be able [12:06] to be more skilled and not be left behind by the  technology. Yeah. Yeah. Great. Great. Thank you. [12:12] Thank you for your insight. And uh for leaders in  this room uh in this audience trying to scale EI, [12:20] what's one lesson you learned the hard way?  Always start with the problem to be solved. [12:28] And I actually would be shocked if any of you find  that surprising. We're all facing these challenges [12:34] in our own world and we there's we're constantly  be hitting being hit with what's possible with [12:40] technology. We have to keep anchoring it back to  what is the business problem? What's the business [12:46] outcome? Um what's the and for us you know in this  this physical AI discussion that we talk about [12:52] what is the customer outcome and how are you going  to judge success based on that and then you can do [12:58] the learning on the job site. you can you do the  learning in in your um in all of our organizations [13:04] get the tools into the hands of people they can  learn and then you build trust through scale as [13:09] you as we go on this journey with our customers  and their job sites we're able to help them go on [13:15] that journey but you just have to stay anchored  on the problem to solve and the outcome that you [13:19] want to get from the technology investment all  right great thank you and every AI data center [13:26] depends on power and caterpillar helps provide  it, right? Um, how has the AI boom changed your [13:34] business? We did a keynote at CES at the beginning  of this year and our chairman and CEO, Joe Creed, [13:43] talked about the fact that Caterpillar builds the  invisible layer of the tech stack. We mine the [13:50] minerals that go into the chips and the technology  that we all use. We build out the infrastructure, [13:58] the data centers. We power those data centers  through our engines and turbines and power [14:03] systems. It it is a great opportunity for us  to be a part of building that infrastructure [14:12] and the the buildout that we're all going through  in order to power this AI uh capability that is [14:19] that we all get to take advantage of. And it's a  wild full circle moment. And then we take all of [14:25] those AI tools and capabilities and we build more  capabilities into our machines to build safer, [14:32] more productive job sites and help our customers  be more successful. So I do uh it is a really it's [14:39] a great opportunity for us to be at the center  of this and then use that capability to help [14:44] our customers be even more successful. Amazing.  Um and um looking ahead, what's the one thing [14:53] you hope Caterpillar to get right with AI and  autonomous? We have to get right the customer [15:01] outcome. We have to make meaningful change for  our customers and meaningful transformations [15:09] for our customers. Uh there there's the  technology element that we've talked about. [15:16] There's the people and process change element  that we've talked about. In the end, all of us are [15:22] investing in technology to get meaningful business  outcomes. We will be successful in this endeavor [15:29] when we help transform customer job sites into  safer, more productive, more efficient job sites [15:37] and that our customers are able to do more work.  They have more work than they they can get done [15:42] today. and it'll help with the AI infrastructure  buildout and hopefully continue to to evolve along [15:48] the way. Okay. Uh okay, that's all my questions  here. Thank you so much, Jamie. Hopefully it was [15:55] helpful. I know you all are between like now and  the end of the day and maybe networking and drinks [16:01] and and so forth. But um I really do appreciate  you all coming and joining us and hopefully [16:06] found the discussion to be valuable. Thank you.  Thank you. It was so good. Thank you. so much.