Meta Creator Chief Warns Fully Synthetic Media Destroys Viewer Trust

@Scale

As automated content floods feeds with low-effort synthetic media, creators who attempt to replace themselves risk alienating audiences while authentic human connection commands an increasing market premium.

AI-assisted summary of Keynote: Creativity in the Age of AI | Yair Livne from Meta — original source. Check the original for full context.

Source published: 2026-10-30

Key ideas

The creator economy is approaching an estimated $500 billion annually, with top individual creator enterprises commanding multi-million to billion-dollar valuations. (0:00)
Over-relying on synthetic AI threatens creator authenticity; viewers react negatively to low-effort 'slop,' broken physics, and unstated automation, leading to severe trust collapse. (5:00)
Meta has deployed an agentic AI Creator Assistant coordinating 4 subagents, 16 skills, and 40 tools to help millions of weekly creators analyze analytics and co-create strategy without replacing the creator. (10:00)
Automated cross-lingual dubbing and lip-syncing across 18 languages now serves over 500 million daily active users on both Facebook and Instagram, relying on transparent UI labeling to protect viewer trust. (11:15)
As generative AI matures and automates sectors like micro-dramas, raw human authenticity—captured via first-person POV wearables and unscripted live streaming—will command an even higher premium. (13:50)

Detailed notes

## Background Published on YouTube by @Scale on October 30, 2026, this keynote features Meta's Yair Livne presenting on 'Creativity in the Age of AI.' No additional show notes or description were provided with the upload.

## TL;DR Yair Livne analyzes how artificial intelligence is transforming digital content creation through the lens of platform economics and Maslow's hierarchy of needs. He warns that low-effort generative 'slop' and synthetic shortcuts risk alienating audiences and shattering creator trust, as seen in controversies surrounding N3on, Hank Green, and MrBeast. To resolve this dilemma, Livne presents Meta's approach: deploying invisible, hyper-personalized agentic assistants behind the scenes and utilizing transparent, disclosed augmentations like cross-lingual dubbing. Ultimately, he argues that while synthetic entertainment will inevitably professionalize, the proliferation of AI will only increase the emotional and cultural premium placed on raw, authentic human connection.

## Atmosphere & style Livne delivers a measured, intellectually rigorous keynote that reflects his background as a tech economist, trading hype for disciplined market analysis and historical precedent. The pacing is crisp and authoritative, blending analytical breakdowns of technological adoption waves with cautionary case studies of audience backlash. The overall experience feels grounded and quietly optimistic, offering reassurance to creators by framing synthetic tools not as replacements for the human spirit, but as behind-the-scenes leverage that elevates authentic self-expression.

## Bulletin - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) The creator economy is approaching an estimated $500 billion annually, with top individual creator enterprises commanding multi-million to billion-dollar valuations. - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) Over-relying on synthetic AI threatens creator authenticity; viewers react negatively to low-effort 'slop,' broken physics, and unstated automation, leading to severe trust collapse. - [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) Meta has deployed an agentic AI Creator Assistant coordinating 4 subagents, 16 skills, and 40 tools to help millions of weekly creators analyze analytics and co-create strategy without replacing the creator. - [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) Automated cross-lingual dubbing and lip-syncing across 18 languages now serves over 500 million daily active users on both Facebook and Instagram, relying on transparent UI labeling to protect viewer trust. - [13:50](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=830s) As generative AI matures and automates sectors like micro-dramas, raw human authenticity—captured via first-person POV wearables and unscripted live streaming—will command an even higher premium.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) Content creation platforms sit at the apex of Maslow's hierarchy of needs, democratizing self-actualization and creative expression for both creators and audiences. - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) Media disruption consistently follows three phases: old-guard gatekeepers, chaotic mass-market democratization, and the rise of a new professional creator class. - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) User consumption patterns demonstrate that video engagement has surged continuously, establishing creators as the new cultural guard and forcing legacy platforms like Netflix to incorporate creator programming. - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) Multimodal generative AI triggers an acute dilemma: its ease of production floods feeds with cheap 'slop' and broken physics, eroding the fragile bond of audience trust. - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) High-profile backlash against experiments like N3on's autonomous stream, Hank Green's suspected LLM phrasing, and MrBeast's AI thumbnail tool show how sensitive audiences are to perceived artificiality. - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) Instead of replacing on-screen human presence, creators should leverage AI as an invisible, always-on personal agency to streamline research, planning, and backend distribution. - [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) Meta's AI Creator Assistant uses historical performance data to pinpoint viral storytelling inflection points, co-creating customized distribution strategies. - [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) Direct generative augmentation succeeds when it is transparently labeled, such as automated multi-language dubbing with lip-syncing or creative Restyle filters. - [12:55](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=775s) A powerful countervailing cultural trend is driving demand toward unvarnished, authentic experiences, seen in the resurgence of live streaming and first-person POV smart glasses. - [13:50](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=830s) Synthetic media will follow the professionalization curve of prior formats, eventually automating low-budget productions like micro-dramas from script to video. - [13:50](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=830s) The proliferation of synthetic content will not replace real artists; it will create an authenticity dividend, increasing the emotional and commercial value of genuine human connection.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) Self-Actualization and the History of Content Creation Platforms

Yair Livne opens his keynote by recalling a previous talk by his colleague Ime Archibong, who discussed how the age of AI introduces serious challenges to authentic human connection alongside new opportunities to connect people. While Archibong grounded his perspective in the middle layer of Maslow's hierarchy of needs—specifically belonging and connection among friends, family, and strangers—Livne focuses his attention on the very apex of the pyramid: self-actualization. When basic safety, nourishment, connection, and respect are secured, individuals gain the autonomy to express themselves creatively. Creative expression matters deeply not only on a personal level, but also societally through culture, entertainment, and art, which allow audiences to participate in the creator's self-actualization.

Introducing his background, Livne describes himself as a “recovering economist” who has dedicated most of his career to building public content ecosystems and supporting the creators within them, beginning at Quora and continuing at Meta. There, his team focuses on Facebook creators while collaborating closely with teams across Instagram, Threads, and news. He explains that these platforms operate fundamentally as economic markets ruled by supply and demand, and analyzing these dynamics helps build superior creator products. Looking at historical supply shifts, Livne traces how people posting monthly online across text, photos, and video transformed in three distinct waves: the early 2000s saw text democratized via personal websites, MySpace, Facebook, and Twitter; the early 2010s turned everyone into photographers via the iPhone, mobile cameras, Instagram, and Snapchat; and the late 2010s transformed the masses into video creators through improved phone cameras, cellular connectivity, and short-form video platforms.

Across every technological medium, disruption consistently advances through three predictable phases:

```mermaid flowchart LR accTitle: Three Phases of Media Disruption accDescr: A flow illustrating the shift from old guard gatekeepers to mass amateur creation and the emergence of modern professional creators. subgraph S1["Phase 1: Incumbents"] A["Old Guard Professionals (Newspapers, TV Networks, Movie Studios)"] end subgraph S2["Phase 2: Democratization"] B["Mass-Market Explosion (Messy, noisy, lacking quality & trust signals)"] end subgraph S3["Phase 3: New Class"] C["Modern Creators (Hone new arts & build scalable businesses)"] end A --> B B --> C ```

Turning to the demand side, Livne evaluates where users actually spend their daily minutes. While daily engagement for text and photo posts eventually broke through and flattened, video consumption surged continuously without any signs of slowing down. To illustrate the cultural pull of this medium, he highlights a viral Reel by electronic music producer Alan Walker. Because the vast majority of video viewing goes directly to creators rather than institutions, video creators have come to dominate consumer mindshare. They have become the “new guard,” forcing even legacy entertainment giants like Netflix—which traditionally prioritized polished in-house productions—to integrate creator-led programming, exemplified by internet personality Alix Earle achieving a top-10 US Netflix show.

- Creator economy market size: approaching nearly half a trillion dollars ($500 billion) annually. - Multiple creator revenue engines: brand partnerships, direct tipping, merchandising, affiliate marketing, and platform payouts. - Top-tier creator businesses: now commanding private market valuations ranging from hundreds of millions to several billion dollars.

[05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) The AI Dilemma: Slop, Broken Physics, and the Fragility of Audience Trust

Livne points out that the creative ecosystem is now entering the first wave of multimodal artificial intelligence, capable of manipulating and producing text, audio, images, and video simultaneously. Unlike earlier formats, multimodal models can transform every medium at once, and hundreds of millions of users are already publishing AI-generated posts every month. However, this creates an acute dilemma for creators. While AI expands the artistic canvas into boundless possibilities, the medium has quickly become stigmatized as low-effort “slop.” Because AI media is cheap and effortless to mass-produce, feeds are flooded with substandard material. The worst examples explicitly try to manipulate viewers by triggering negative emotions like pity, disgust, and anger, resulting in an immediate erosion of trust between creators and viewers.

To demonstrate the perils of full-scale AI adoption, Livne cites the livestreamer N3on, who announced his retirement from active streaming to launch the first-ever 24/7 autonomous AI streamer using Higgsfield AI. The project sparked an intense backlash from fellow creators and fans, who labeled the stunt lazy and an outright sellout. Livne plays a chaotic gameplay clip from the stream featuring muffled voices, plot holes, disappearing characters, and jarring rendering errors—such as an avatar's hand passing straight through a box rather than opening it. When viewers see broken physics and careless errors, they realize no human is invested in the craft. Livne reflects on the fallout: “This immediately puts the viewer in a place where trust breaks. You can't trust that somebody actually cares about the experience you're watching.”

```mermaid flowchart TD accTitle: The Trust Collapse Under AI Adoption accDescr: How synthetic shortcuts damage authenticity and undermine audience trust in creator ecosystems. A["Adopting Cheap / Fully Synthetic AI"] --> B["Glitches, Inconsistencies & Slop"] B --> C["Breach of Audience Trust"] C --> D["Failure of Human Connection"] D --> E["Collapse of Maslow's Creative Hierarchy"] ```

Livne emphasizes that audience backlash is not limited to extreme experiments; even subtle hints of artificial production can derail established creators: - **Hank Green:** The popular science communicator and educational YouTuber had to temporarily pause production across several channels after fans accused him of using LLMs because a single scripted phrase mirrored the tone of Anthropic's Claude. Green had to issue extensive public apologies before resuming uploads. - **MrBeast (Jimmy Donaldson):** The world's largest digital creator faced immediate pushback after releasing an AI thumbnail generator designed to help smaller accounts emulate top creator styles. He promptly pulled the product off the market after his audience criticized it for encouraging mass inauthenticity.

Livne synthesizes the trilemma creators now face: they can refuse AI and risk technological obsolescence, go all-in by replacing themselves and alienate their followers, or strategically dabble while safeguarding audience trust. Livne argues: “I'm here to argue that while fully replacing your current creation or yourself with AI really risks creators breaking trust with their audience, we are already seeing how creators are able to react to this AI moment by leaning into both AI technologies and others, experimenting, but really keeping that most expensive currency, trust.”

The primary path forward is treating AI not as an on-screen replacement, but as an invisible, personal superintelligence. Large creators have long relied on dedicated production, research, and distribution crews. AI tools can now act as an always-on, low-cost social media agency for emerging creators—assisting with brainstorming, administrative planning, and post workflows across YouTube, TikTok, Facebook, and Instagram Edits without misrepresenting who created the final work.

[10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) Meta’s AI Creator Assistant: Hyper-Personalized Superintelligence

Meta has developed a dedicated AI Creator Assistant designed to realize Mark Zuckerberg’s vision of personalized superintelligence, giving creators substantial leverage in managing their presence across Facebook and Instagram. Rather than offering generic conversational output, the assistant is engineered to understand each creator's unique audience and past performance. Operating via an agentic architecture, the system coordinates four specialized subagents, 16 distinct skills, and 40 integrated tools—a suite that continues to expand monthly. These tools assist creators across their primary operational bottlenecks, including surfacing deep content analytics, drafting captions, managing comment replies, and brainstorming creative strategies.

The assistant directly taps into historical performance data to co-create strategy alongside the creator. Yair Livne highlights instances where the assistant analyzed a creator's video history and pinpointed the exact moment within a clip that catalyzed its virality, allowing the creator to intentionally replicate that storytelling mechanism in future productions. The tool was developed by a lean, talent-dense team leveraging modern agentic flows, which compressed engineering cycles by 3x to 4x compared to traditional product development timelines. Four months after rollout, the assistant is already utilized weekly by millions of creators with exceptionally high user retention.

“We wanted to provide a tool that can help creators with all their needs in a hyper-personalized way,” Livne explains, emphasizing that “all these abilities are deeply rooted in the creator's insights and history, knowing which of their content resonates with their audience, and enabling them to co-create a content strategy.”

```mermaid flowchart TD accTitle: Meta Creator Assistant Architecture accDescr: Diagram illustrating the agentic framework of Meta's creator assistant orchestrating subagents, skills, and tools. subgraph Core["Creator Assistant Harness"] direction TB A["Creator Performance & History"] --> B["Agentic Orchestrator"] B --> C["4 Subagents"] end

subgraph Capabilities["Skills & Toolsets"] direction TB C --> D["16 Skills"] C --> E["40+ Monthly Growing Tools"] end

subgraph Actions["Creator Workflows"] direction TB D & E --> F["Content Insights & Virality Analysis"] D & E --> G["Drafting Captions & Ideation"] D & E --> H["Audience Comment Management"] end ```

- **Agentic architecture:** Powered by a dedicated harness invoking 4 subagents, 16 skills, and 40 expanding tools. - **Development velocity:** Built using advanced agentic pipelines, cutting engineering turnaround times by 3x to 4x. - **Scale and adoption:** Deployed for only four months, reaching millions of active creators weekly with strong engagement metrics.

---

[11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) Transparent Augmentation: AI Translations and Restyle Filters

Addressing the challenge of incorporating generative AI directly into creator media without undermining audience trust, Livne advocates for explicit, transparent augmentation. When enhancements are openly disclosed rather than hidden, creators can meaningfully expand the reach and creative expression of their work without deceiving viewers.

A prime example is automated cross-lingual translation, which incorporates synchronized dubbing and precise lip-syncing into 18 languages across both Facebook and Instagram. This feature touches over 500 million daily active users on each platform. Viewers are clearly informed when a video has been translated with AI and are given the option to toggle between languages, preserving the creator's original voice while removing global language barriers.

Transparent augmentation also enables playful cultural phenomena, such as Meta’s "Restyles" video filters. Livne points to Edits, a creative video editing tool that reached 50 million monthly active users, which sparked an organic cultural wave in India. Creators across regions that never experience snowfall used the app's "Snow" preset to transform their everyday cityscapes into winter wonderlands, prompting millions of downloads purely to participate in the shared cultural meme.

“Creators can keep their content real but explicitly augment it with AI using the latest and greatest models, but what's key about the augmentation is that it needs to be transparent to the user so it maintains trust, but still enables the content to go further and more meaningfully than before,” Livne states.

- **AI translations:** Fully dubbed and lip-synced across 18 languages, serving over half a billion daily active users on Facebook and Instagram respectively. - **Trust safeguards:** UI labels explicitly designate AI-dubbed content and provide manual language switching to ensure viewer transparency. - **Restyles adoption:** Creative filter presets like "Snow" drove virality in India and propelled the Edits editing tool past 50 million MAU.

---

[12:55](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=775s) Raw Human Connection: Point-of-View Smart Glasses and Live Video

In stark contrast to synthetic generation, Livne identifies an equally powerful countervailing trend: audiences gravitating toward hyper-authentic, unvarnished human-to-human experiences. Wearable hardware such as Meta Ray-Ban smart glasses delivers a literal first-person point of view (POV), capturing raw human perspective in ways conventional cameras cannot replicate. Livne demonstrates this intimacy through athletic training footage, putting the audience directly inside an NBA practice through the eyes of Minnesota Timberwolves star Karl-Anthony Towns.

This desire for unmediated reality is simultaneously fueling a dramatic resurgence in live streaming. While live video experienced an expected post-pandemic lull as physical gatherings resumed, consumer engagement is rebounding rapidly. The inherent unpredictability, lack of rehearsal, and unfiltered presence of live broadcasts offer a level of intimacy that highly polished synthetic media cannot match, prompting creators to double down on live formats to satisfy audience demand.

“Live streams create a sincere, unrehearsed, and authentic vibe, which is why I believe consumers are flocking back to them now,” Livne observes.

- **First-person perspective:** Meta smart glasses allow fans to directly inhabit a creator's perspective, exemplified by NBA drills seen through the eyes of Karl-Anthony Towns. - **Live streaming recovery:** Rebounding past its post-pandemic low due to rising user preference for unscripted, real-time spontaneity.

---

[13:50](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=830s) The Maturation of AI Media and the Enduring Value of Human Storytelling

Looking toward the broader horizon of creative media, Livne acknowledges that generative AI tooling will inevitably mature. The low-grade "slop" that characterizes much of today’s generative media will disappear as AI content undergoes the exact same professionalization curve previously traversed by digital text, photography, and video. This transition is already well underway in the micro-drama sector—a market that emerged only in recent years and is already seeing end-to-end synthetic micro-dramas that rival or surpass the quality of mid-tier live-action alternatives.

This automation is accelerating with the introduction of novel toolchains, such as ByteDance's automated video framework, which generates complete, multi-scene micro-dramas directly from a raw script upload. Yet while high-quality synthetic media will inevitably capture a significant portion of overall entertainment consumption, Livne argues that it will simultaneously elevate the premium placed on genuine human identity. Rather than displacing real artists, the ubiquity of synthetic content will reinforce the appetite for genuine human vulnerability and connection.

“Humans will continue rooting for, identifying with, and looking up to other humans,” Livne argues, concluding his keynote by affirming that while the medium and tools of expression will evolve, “creativity fosters human connection, and that will always have a big place.”

- **Professionalization of AI content:** AI media quality will shed its current amateur reputation, mirroring the historical maturation of text, web graphics, and digital video. - **Micro-drama automation:** Tools like ByteDance’s script-to-video workflow enable instant end-to-end micro-drama generation, disrupting traditional low-budget live-action productions. - **The authenticity dividend:** As synthetic entertainment expands its market footprint, human creators who cultivate genuine emotional resonance and personal identity will command even higher audience value.

## Notable quotes > [13:50](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=830s) Humans will continue rooting for, identifying with, and looking up to other humans. — Yair Livne

> [12:55](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=775s) Live streams create a sincere, unrehearsed, and authentic vibe, which is why I believe consumers are flocking back to them now. — Yair Livne

## Mentioned Yair Livne, Meta, Ime Archibong, Quora, Facebook, Instagram, Threads, MySpace, Twitter, Snapchat, Alan Walker, Netflix, Alix Earle, N3on, Higgsfield AI, Hank Green, Claude, Anthropic, MrBeast, YouTube, TikTok, Mark Zuckerberg, Karl-Anthony Towns, Minnesota Timberwolves, NBA, ByteDance

[Source](https://www.youtube.com/watch?v=CYuO9-rau5Q)

中文简报

Meta创作者负责人警告：全合成媒体正在破坏观众信任随着自动化内容以低成本合成媒体泛滥于信息流中，试图取代自身的创作者面临疏远受众的风险，而真实的人类连接则享有日益上升的市场溢价。 Yair Livne从平台经济学和马斯洛需求层次理论的视角分析了人工智能如何重塑数字内容创作。他警告称，正如N3on、Hank Green和MrBeast引发的争议所示，低成本生成的“废料”和合成捷径可能会疏远观众并粉碎对创作者的信任。为化解这一困境，Livne介绍了Meta的应对方案：在幕后部署隐形、超个性化的智能体助手，并采用跨语言配音等透明且公开的增强工具。他最终指出，尽管合成娱乐不可避免地会走向专业化，但AI的激增只会进一步推高原生、真实的人类连接所具备的情感与文化溢价。创作者经济规模预计每年将接近5000亿美元，顶尖的个人创作者企业估值已达数百万乃至数十亿美元。 (0:00) 过度依赖合成AI会威胁创作者的真实性；观众对低成本“废料”、违背物理规律的内容以及未注明的自动化反应消极，导致信任严重崩溃。 (5:00) Meta已部署了一款由4个子智能体、16种技能和40种工具协同工作的智能体AI创作者助手，每周协助数百万创作者分析数据并共同制定策略，同时不取代创作者本身。 (10:00) 涵盖18种语言的自动化跨语言配音和唇形同步现已服务于Facebook和Instagram上超过5亿的日活跃用户，依托透明的用户界面标签来维护观众信任。 (11:15) 随着生成式AI走向成熟并实现微短剧等领域的自动化，通过第一人称视角穿戴设备和无脚本直播所捕获的原生人类真实性将获得更高的溢价。 (13:50)

Reading summary

Meta's Yair Livne argues that while generative AI expands content creation, fully replacing human creators with synthetic media risks breaking audience trust, making transparent augmentation and authentic human connection essential to sustaining the creator economy. Media Disruption and the Rise of Creators (0:00) Analyzing creator platforms through supply and demand economics, Livne outlines three historical waves of online supply democratization: text in the early 2000s, mobile photography in the early 2010s, and short-form video in the late 2010s. In each wave, disruption progressed from legacy institutional gatekeepers to mass amateur creation, culminating in a new professional class capable of building scalable businesses. Unlike text and photo engagement, which eventually plateaued, video consumption has surged continuously, driving the creator economy toward an estimated $500 billion market size. Modern creators now dominate consumer mindshare, forcing traditional entertainment networks like Netflix to incorporate creator-led programming. The AI Trust Dilemma and Audience Backlash (5:00) Multimodal AI enables the simultaneous generation of text, audio, images, and video, but its ease of production has flooded digital feeds with low-effort "slop" that often exploits negative emotions like disgust and anger. When creators attempt to substitute themselves with synthetic alternatives, audience trust frequently collapses. Livne highlights streamer N3on, whose 24/7 autonomous AI stream faced severe backlash due to rendering errors, broken physics, and muffled audio that signaled a total absence of human craft. Even subtle synthetic touches have disrupted established creators: educational YouTuber Hank Green paused production after facing accusations over Claude-like phrasing, and MrBeast withdrew an AI thumbnail generator following criticism that it encouraged mass inauthenticity. Livne cautions that creators must avoid both technological obsolescence and full replacement, framing trust as their most valuable asset. Invisible Assistance and Transparent Augmentation (10:00) Rather than deploying AI as an on-screen replacement, Livne advocates positioning it as an invisible superintelligence that handles operational bottlenecks like ideation, caption drafting, and performance analytics. To this end, Meta built an AI Creator Assistant using an agentic architecture featuring 4 subagents, 16 skills, and 40 tools. Reaching millions of weekly active creators within four months, the assistant examines historical performance to help creators co-create content strategies. When generative AI is applied directly to outward-facing content, Livne argues that explicit, transparent labeling preserves viewer trust. Examples include AI translations with synchronized lip-syncing across 18 languages—touching over 500 million daily users on both Facebook and Instagram—and creative editing presets like Meta's "Snow" filter in Edits, which surpassed 50 million monthly active users after sparking a viral trend in India. Raw Connection and the Maturation of Media (12:55) Parallel to generative tools, a countervailing consumer demand has emerged for unvarnished, authentic human experiences. This appetite is reflected in the rebound of live streaming—favored for its unscripted, unrehearsed nature—and the adoption of POV wearable hardware like Meta Ray-Ban smart glasses, which Livne illustrates through first-person NBA training footage from Karl-Anthony Towns. Looking forward, Livne anticipates that AI tooling will inevitably mature, progressing through the same quality curve that professionalized digital video. While automated pipelines—such as ByteDance's script-to-video micro-drama generators—will produce high-quality synthetic entertainment, Livne contends this abundance will create an authenticity dividend, ultimately elevating the emotional and economic value of genuine human vulnerability. Livne concludes that although automated pipelines will capture a substantial share of entertainment production, audiences will consistently seek out and identify with human experiences, ensuring that authentic creative expression remains central to media ecosystems.

精炼阅读摘要

Meta创作者团队负责人Yair Livne在演讲中指出，在规模接近5000亿美元的创作者经济中，多模态AI带来了扩展创作边界与破坏受众信任的双重挑战；创作者若要破局，应当将AI定位为后台的隐形超智能与透明的增强工具，而非彻底替代人类表达。媒介颠覆规律与创作者经济的崛起 (0:00) 从马斯洛需求层次理论的顶层切入，Livne认为创作表达是人类实现“自我实现”的关键路径，而文化与艺术则让受众得以共享这种自我实现。回顾互联网内容供给的历史变迁，媒介历经了三次技术民主化浪潮：2000年代初个人网站与社交网络催生的文字普及、2010年代初智能手机与移动应用带动的全民摄影，以及2010年代末高速网络与手机摄像头驱动的视频爆发。每一种新媒介的兴起都遵循着从传统机构垄断、到大众创作爆发导致质量与信任混乱、再到全新专业创作者阶层建立成熟商业模式的三阶段演进规律。在需求侧，视频消费展现出不可逆的增长势头，且绝大多数时长流向了原生创作者而非传统机构。如今创作者经济的年市场规模已逼近5000亿美元，顶尖创作者机构估值达到数亿乃至数十亿美元，其影响力甚至迫使Netflix等传统流媒体巨头引入Alix Earle等网红主导的节目。创作者不仅彻底确立了文化话语权，也重塑了现代娱乐业的底层商业逻辑。合成劣质内容与受众信任崩塌的困境 (5:00) 随着多模态AI的介入，内容生产进入了全媒介同时重构的新阶段，但伴随而来的却是严重的信任危机。AI生成的低成本“劣质内容”（Slop）在信息流中泛滥，甚至通过挑动负面情绪进行算法操纵，迅速消耗着受众的耐心。Livne援引直播主播N3on的案例说明全面由AI替代人类的危险：其利用Higgsfield AI打造的24小时自主AI直播因角色手部穿透箱子、画面断裂和声音模糊等严重“物理法则破损”，引发了受众对创作者缺乏用心的强烈反感，导致信任链条彻底断裂。即使是细微的AI痕迹，也可能对头部创作者造成反噬。知名科普创作者Hank Green曾因文案单句与Anthropic的Claude模型语气高度重合而遭遇粉丝声讨，最终不得不公开道歉并暂停多频道更新；MrBeast推出的AI封面生成工具也因涉嫌鼓励虚假伪造，在受众抗议下迅速撤下。Livne总结道，创作者如今陷入了“拒绝AI导致落后”与“全盘AI化失去受众”的两难境地，而信任已成为内容生态中最昂贵的资产。后台智能辅助与透明增强的技术范式 (10:00) 面对信任危机，Livne提出AI的破局之道在于充当创作者的“隐形个人超智能”。类似于头部创作者依赖庞大的编导与分析团队，AI应当成为新兴创作者全天候、低成本的后台智囊，负责繁杂的数据分析、文案构思与发布排期。Meta推出的AI创作者助手即实践了这一逻辑：该系统依托智能体架构，协同4个专业子智能体、16项核心技能与40多项每月递增的工具，深入创作者的历史数据并精准定位视频爆点时刻，辅助其共创策略。借助现代化智能体开发管线，该工具的研发周期缩短了3至4倍，上线仅四个月便吸引了数百万创作者每周高频留存使用。在涉及前端展示时，Meta强调必须实行“透明增强”。例如，覆盖Facebook和Instagram双端各超5亿日活用户的跨语言自动翻译功能，在提供18种语言的高精度口型同步与配音的同时，通过明确的UI标识让观众知情并支持手动切换原声。此外，Edits应用推出的AI滤镜（如在从不下雪的热带地区风靡的“Snow”预设）已积累超过5000万月活用户，证明了透明、趣味化的AI视觉修饰能在保持真实感的前提下激发广泛的文化共鸣。具身视角、直播复兴与人类连接的溢价 (12:55) 在合成媒体迅速扩张的同时，受众对原生、未加工的真实人类体验展现出强烈的补偿性需求。硬件层面上，Meta Ray-Ban等智能眼镜通过第一人称视角（POV）记录人类主观体验，例如在NBA球员Karl-Anthony Towns的训练视角中，带来了传统摄像机无法复制的沉浸感与真实感。与此同时，网络直播在疫情后迎来强劲反弹，受众正重新流向那些充满不确定性、未经排练且未经滤镜修饰的实时互动，以此对抗精致却冰冷的合成媒介。展望未来，Livne认为AI内容同样会经历专业化演进，当前的粗糙与破损终将消失。类似字节跳动基于剧本自动化生成多场景微短剧的系统，已证明全流程合成内容完全具备媲美中端影视制作的能力。然而，即便合成娱乐在大众消费中占据一席之地，受众对于真实人类身份、情感共鸣与脆弱性的认同并不会褪色；相反，高度工业化的数字合成时代将赋予真实人类创作者更高的“真实性溢价”。 Livne坚信，尽管生成式AI的技术形态与创作工具将持续迭代并走向专业化，但大众追随、共情和崇拜真实个体的本能不会改变；在机器能够批量制造娱乐的时代，坚守透明度与真诚的人性连接将是创作者维持持久价值的核心基石。

Detailed notes — zh

## Background 本视频由 YouTube 频道 @Scale 于 2026 年 10 月 30 日发布，收录了 Meta 创作者产品负责人 Yair Livne 的主题演讲《AI 时代的创造力》（Keynote: Creativity in the Age of AI）。视频发布页面未提供补充说明或节目简介。 ## TL;DR Livne 从马斯洛需求理论顶层的“自我实现”切入，梳理了媒介形态从精英垄断到大众化噪音、再到新专业阶层崛起的三阶段演进规律，指出视频创作者已构建起数千亿美元的商业帝国。多模态 AI 虽然降低了生产门槛，但伴生的大量劣质垃圾内容（slop）正在严重侵蚀创作者赖以生存的受众信任纽带。他认为 AI 的最佳定位应是充当幕后“个性化超级智能”智囊团，以及进行透明可控的显式增强（如跨语言翻译与口型同步）。尽管端到端 AI 影视内容未来必将走向高度专业化，但机器生成的泛滥反而会放大市场对粗粝“真实感”的溢价，人与人之间的共情认同仍将是创造力的永恒核心。 ## Atmosphere & style 整场演讲呈现出经济学家式的沉稳与平台操盘手的敏锐，Livne 以清晰的宏观供求逻辑将技术浪潮抽丝剥茧。在剖析劣质 AI 翻车案例时言辞警醒且发人深省，而在演示智能体架构与透明化工具时又展现出务实的乐观主义。叙事节奏层层递进，从对抗技术焦虑的理性推演，最终升华为对人类本真价值与深层联结的温情确认，视听体验兼具前瞻视野与人文底色。 ## Bulletin - [02:30](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=150s) 创作者经济年市场规模已逼近 5000 亿美元，视频创作者逐步主导消费者心智并突破传统影视壁垒，甚至登上 Netflix 全美收视前十。 - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) 低质 AI 内容（slop）与虚假感正引发严重信任反噬，多位头部创作者因 AI 翻车或引咎停播，迫使生态将真实性视为最昂贵的核心资产。 - [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) Meta 推出基于 4 个子代理、16 项技能与 40 种工具的创作者专属 AI 助手，上线 4 个月即实现数百万创作者周活并具备强留存。 - [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) Meta 支持 18 种语言的透明化 AI 翻译与口型同步功能在 Facebook 和 Instagram 上每日各触达超 5 亿用户，兼顾触达效率与信任维护。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) 媒介形态演进遵循三阶段规律：旧行业精英垄断、技术下放导致的大众化与噪音期，以及最终崛起的新专业创作者阶层。 - [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) 基于马斯洛需求层次理论，内容创作属于顶层的自我实现，但其必须建立在底层人际连接与信任的基石之上。 - [02:30](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=150s) 在文字、图片与视频等媒介中，用户在视频内容上的平均时长持续高速增长，且绝大部分注意力流向了独立创作者。 - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) AI 生成内容的门槛极低引发了网络废料（slop）泛滥，粗制滥造和情绪操纵严重侵蚀了受众对创作者的底层信任。 - [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) 主播 N3on 打造的 24 小时 AI 虚拟直播因严重穿模和逻辑硬伤引发舆论反噬，被观众视作圈钱背叛。 - [07:20](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=440s) Hank Green 停更道歉与 MrBeast 下架 AI 缩略图工具等事件表明，受众对虚假感极其敏感，全盘用 AI 替代前台创作风险极大。 - [07:20](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=440s) 创作者应对 AI 浪潮的最佳切入点是在幕后将其作为“个人超级智能”，充当涵盖选题、文案与运营策略的虚拟智囊团。 - [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) Meta 开发的创作者 AI 助手深度整合历史数据与受众洞察，能精准定位过往爆款视频的关键帧以指导后续创作。 - [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) 借助前沿智能体工作流（agentic flows），Meta 将创作者 AI 助手等产品的研发周期大幅缩短了 3 至 4 倍。 - [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) 在内容制作端引入 AI 必须保持显式与透明，如提供 AI 翻译标识与语言切换功能，以在扩展触达的同时保护受众信任。 - [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) Edits 的 Restyles 滤镜月活达 5000 万，印度的雪景特效（Snow preset）展现了趣味非欺骗性 AI 特效在塑造流行文化上的裂变能力。 - [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) 以 Ray-Ban Meta 智能眼镜的第一人称 POV 视角为代表，未经修饰的粗粝真实体验正在成为对抗 AI 虚假感的有力形态。 - [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) 后疫情时代直播流量逆势反弹，其未排练、即兴的真诚氛围重新吸引了厌倦虚假修饰的大规模受众。 - [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) AI 内容将像过往媒介一样迈向高度专业化，例如剧本端到端生成的全 AI 微短剧视听品质已开始超越部分真人影视短剧。 - [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) 随着精良 AI 内容的泛滥，大众对人类本真性（authenticity）与真实情感共鸣的渴望和溢价将呈指数级增长。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=0s) 马斯洛需求金字塔之巅：创意表达与内容生态的三阶段演进 Meta 创作者产品负责人 Yair Livne 在开场时回顾了同事 Ime Archibong 去年的演讲。当时 Archibong 探讨了 AI 时代对人际真实连接带来的挑战与机遇，并以马斯洛需求层次理论为基石，重点剖析了金字塔中层关于朋友、家人乃至陌生人之间的“情感与归属连接”。而 Livne 则将目光投向了金字塔的最顶端——他本人最感兴奋的“自我实现”（Self-actualization）。他认为，当人们获得温饱、安全感并感受到与他人的连接和尊重后，便开启了通过创造性表达和自主性实现自我的大门。这种表达不仅对个人意义深远，在社会层面更构成了娱乐、文化与艺术的基石。Livne 自称是一名“转型中的经济学家”（recovering economist），职业生涯大部分时间都在 Quora 以及如今的 Meta 打造公共内容生态，专注于服务 Facebook、Instagram、Threads 等平台的创作者，并运用供需法则来打磨更好的创作者产品。随后，Livne 展示了一张追踪全球每月发布文字、图片和视频帖子人数的历史图表，复盘了互联网内容创作的爆发轨迹。从 2000 年代初期的个人网站、MySpace、Facebook 和 Twitter 让数以亿计的人拥有发声渠道，到 2010 年代初 iPhone 普及与相机技术提升促成 Instagram 和 Snapchat 的崛起，再到 2010 年代末手机摄影、网络带宽升级与短视频平台让所有人皆可成为视频创作者。Livne 指出，在每一种媒介形态释放的过程中，生产资料的控制权都经历了结构性洗牌，并呈现出高度一致的三个颠覆阶段： ```mermaid flowchart LR accTitle: 媒介颠覆的三阶段演进模型 accDescr: 展示传统精英垄断、大众化混乱期到新专业创作者阶层崛起的三个阶段。 subgraph Phase1["第一阶段：旧秩序垄断"] A["专业精英把守创作大门 (报社、电视台、电影制片厂)"] end subgraph Phase2["第二阶段：大众化与噪音"] B["技术门槛降低与平台爆发 (人人皆可创作，质量与信任信号混乱)"] end subgraph Phase3["第三阶段：新专业阶层崛起"] C["原生创作者脱颖而出 (精通新格式与平台，开辟全新商业模式)"] end Phase1 --> Phase2 Phase2 --> Phase3 ``` 正如 Livne 所阐述的那样，每个浪潮最初都由少数传统专业守门人主导，随后在技术普及中进入泥沙俱下的大众化阶段，最终催生出真正驾驭新媒介的现代专业创作者。在阐述自我实现与创作冲动的本质时，Yair Livne 强调：“When we are nourished and safe and feel a connection to others and feel respected, That opens the door for us to self-actualize through creative expression and autonomy.” - **历史拐点**：2000 年代初（文本社交爆发）、2010 年代初（移动影像与滤镜社交崛起）、2010 年代末（短视频与全民视频创作者时代）。 - **颠覆规律**：旧行业精英掌舵 $\rightarrow$ 技术下放引发海量噪音与信任缺失 $\rightarrow$ 新专业创作者阶级（Creators）确立商业与艺术标准。 --- ### [02:30](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=150s) 视频主导的注意力经济与创作者万亿商业版图在供给侧经历了三波形态演进的同时，需求侧的用户注意力也发生了剧烈转移。Livne 呈现了用户在各格式上平均每日花费时长的走势数据：文字与图片的消费时长在经历初期的飞速增长后最终趋于平缓，唯独代表视频的粉色曲线始终保持强劲的上升势头，至今未见任何放缓迹象。为了具象化视频内容的惊人传播力，现场展示了音乐制作人 Alan Walker 近期爆火的一支 Reel 短视频。Livne 指出，用户在视频上消耗的绝大多数时间，流向的并非传统影视，而是独立创作者的内容，这也正是视频创作者如今能主导全球消费者心智的根本原因。视频创作者已然成为视听领域的新守门人。各大数字平台为了争夺顶级创作者的内容与忠诚度展开了激烈角逐，就连长期深耕高成本自制剧的传统流媒体巨头 Netflix，也开始大举引入网络创作者以维持平台活力与年轻化。例如顶级网红 Alix Earle 本月早些时候推出的节目直接冲上了 Netflix 全美前十榜单，这在两年前完全不可想象。与此同时，创作者经济已发展为一个年规模接近 5000 亿美元的庞大产业，资金来源涵盖品牌合作、打赏赞赏、周边电商、联盟营销和平台分成等多元渠道，部分头部创作者甚至打造出了估值数亿乃至数十亿美元的商业帝国。谈及当下正在发生的代际剧变，Livne 指出世界已经跨入了 AI 驱动创作的初始阶段，多模态模型具备同时重塑文本、图像与音视频的能力，已有数以亿计的用户每月在社交网络上发布 AI 生成内容。针对创作者群体目前牢牢占据的统治地位，Yair Livne 总结道：“The vast majority of this time spent on video goes to creator content, and this is why video creators run the show today and have dominated consumer mindshare.” - **产业规模**：全球创作者经济体量已逼近 5000 亿美元（half a trillion dollars）。 - **行业破圈标志**：网红 Alix Earle 的原创节目成功跻身 Netflix 全美收视前十名。 - **AI 普及速度**：尽管尚处于极早期阶段，每月发布 AI 生成内容的用户数已达数亿人规模。 --- ### [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) AI 劣质内容（Slop）引发的信任危机与翻车案例面对多模态 AI 带来的无限画布，创作者群体正处于极其微妙的十字路口。一方面，AI 打破了以往仅靠手机摄像头取景的物理局限，赋予创作者无限的创意空间；但另一方面，AI 内容在网络文化中迅速与低质无趣的“废料/垃圾内容”（slop）画上了等号。由于 AI 内容生成成本极低且能规模化量产，社交平台上充斥着大量质感廉价、毫无审美价值的作品，更恶劣的 AI 内容甚至试图利用虚假信息恶意操纵受众情绪，去激发恶心、同情或愤怒等负面反应。这种泛滥直接侵蚀了观众对创作者的底层信任，一旦创作者贸然使用粗劣的 AI 内容，极易被粉丝视作偷懒和背叛。为了展现这种信任崩塌的现实后果，Livne 援引了知名 Twitch 主播 N3on 的极端翻车案例。N3on 此前高调宣布从真人直播中“退役”，转而与 Higgsfield AI 合作打造史上首个全天候 AI 虚拟直播流。此举立刻招致了粉丝与其他创作者的铺天盖地抨击，被斥为毫无诚意和彻底的“圈钱背叛”（selling out）。现场播放的直播切片中，AI 生成画面充满了滑稽而刺眼的物理逻辑破绽——角色的手在打开箱子时直接穿透了模型表面，整场直播情节断裂、人物凭空消失又出现。观众无法从这种粗制滥造的内容中感受到创作者的用心，原有的信任纽带荡然无存。 Livne 再次将这一现象与马斯洛需求理论联系起来：在金字塔顶端的“自我实现”之下，是不可动摇的“连接与信任”基石；如果失去了受众的信任，上层的所有表达都将瞬间坍塌。关于过度采用劣质 AI 内容给创作者品牌带来的毁灭性打击，Yair Livne 严肃指出：“By adopting AI content, creators risk turning off their audience and making the relationship to them feel cheapened and betrayed.” - **AI 负面情绪陷阱**：劣质内容往往依赖猎奇与算法投机，诱发恶心（disgust）、同情（pity）和愤怒（anger）等负面情绪。 - **典型翻车事件**：主播 N3on 联手 Higgsfield AI 推出的 24 小时 AI 直播因穿模、逻辑硬伤和虚假宣传引发全网舆论反噬。 --- ### [07:20](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=440s) 信任腐蚀的严峻抉择与作为“超级智能团队”的 AI 机遇观众对“虚假感”的嗅觉极为敏锐，哪怕是一丝一毫的不真诚都会引发惨烈反噬，即便顶流创作者也无法幸免。Livne 举了两个极具代表性的例证：著名科普 YouTube 创作者 Hank Green 因视频文案中出现了一句带有典型 Claude 风格的句式，引发受众怀疑其使用 AI 代笔脚本，舆论压力迫使他不得不暂停了多个频道的更新，并做出极为深刻的公开道歉后才得以复播；全网顶流 MrBeast 曾推出一款 AI 缩略图生成工具，允许中小创作者模仿大博主的视觉风格，但因被指责“规模化鼓励造假行为”、损害了创作者生态的真实性信任，最终也不得不紧急下架该产品。面对这场技术风暴，创作者面临着艰难的三岔路口：是彻底无视 AI、固守传统工作流但承担被时代抛下的风险？还是全盘拥抱、激进甚至用 AI 代替自己？亦或是保持实验与试水，在承担可控风险的同时布局未来？Livne 明确主张，完全用 AI 替代自身创作是极其危险的自杀行为，创作者必须将“信任”视为最昂贵的资产来悉心守护。在此前提下，AI 最健康、最强大的应用切入点，是在内容正式制作之前充当创作者的“个人超级智能”（Personal Superintelligence）。顶级创作者通常拥有庞大的幕后团队负责策划选题、统筹宣发与跨平台运营，而中小创作者往往无力承担雇佣成本。AI 恰好可以在幕后扮演全能社交媒体团队的角色，协助创作者梳理灵感、打磨剧本与优化运营策略，在完全不稀释台前真实人设与观众信任的前提下大幅提升生产力。目前 TikTok、YouTube 以及 Meta（Facebook、Instagram 等）均在积极推出面向创作者的官方 AI 助手工具。针对创作者在技术浪潮中维系受众关系的底线逻辑，Yair Livne 强调：“I'm here to argue that while fully replacing your current creation or yourself with AI really risks creators breaking trust with their audience, we are already seeing how creators are able to react to this AI moment by leaning into both AI technologies and others, experimenting, but really keeping that most expensive currency, trust.” - **Hank Green 停播事件**：仅因一句疑似 Claude 生成的台词引发信任危机，致使多频道停更并公开道歉。 - **MrBeast 撤回工具**：因 AI 缩略图工具助长风格抄袭与虚假内容，为维护受众信任主动关停产品。 - **AI 破局切入点**：从前台内容替代转向幕后“个人超级智能”（选题策划、多平台运营、数据分析），扮演虚拟社交媒体智囊团。 ### [10:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=600s) 赋能创作者：基于个性化超级智能的 AI 创作助手在创作前期的构思与运营阶段，AI 正在为创作者提供前所未有的杠杆效应。Meta 紧跟 Mark Zuckerberg 关于“个性化超级智能”（personalized super intelligence）的愿景，在 Facebook 上为创作者打造了一款专属 AI 助手，以超个性化的方式全方位满足创作者的各类需求。这套专用的技术架构能够调度 4 个子代理（subagents）、16 项技能（skills）以及 40 种工具（tools），且功能规模每月都在持续扩充。它深度覆盖了创作者的核心痛点，包括获取内容洞察、撰写文案与标题、回复粉丝评论以及头脑风暴全新创意。该助手的独特之处在于它深深植根于创作者的历史数据与表现洞察。它清楚哪些内容能够引发受众的强烈共鸣，进而协助创作者共同制定内容策略。Yair Livne 分享了一个鲜活的案例：该助手能够精准指出某位创作者过往视频中引发全网病毒式传播的“关键瞬间”，并指导其将这一洞察复用到未来的内容创作中。该工具由一支精简且人才密度极高的团队利用最前沿的智能体工作流（agentic flows）打造，相比传统软件研发模式，将开发周期大幅缩短了 3 到 4 倍。正如 Yair Livne 所强调的：“This product has been in the market for only 4 months and it's already used by millions of creators on a weekly basis with really strong retention.” - **架构能力**：配备 4 个子代理、16 项专业技能和 40 种工具，系统每月持续升级扩展。 - **核心场景**：爆款复盘分析、智能文案起草、粉丝互动回复、脚本与创意脑暴。 - **业务成果**：产品上线仅 4 个月，周活跃创作者即达数百万级别，用户留存表现异常强劲。 ```mermaid flowchart TD accTitle: 创作者 AI 助手架构与功能流 accDescr: 展示 Meta 创作者助手如何整合历史数据、调用智能体矩阵并输出创作全流程赋能。 subgraph Data["数据与智能体底座"] H["创作者历史数据与受众洞察"] --> Core["专属调度架构 (Dedicated Harness)"] Core --> Sub["4 个子代理 / 16 项技能 / 40 种工具"] end subgraph Actions["核心赋能场景"] Sub --> I["洞察分析 (定位病毒式传播关键帧)"] Sub --> C["创作辅助 (文案撰写与创意脑暴)"] Sub --> E["社群互动 (粉丝评论智能回复)"] end subgraph Output["业务价值"] I & C & E --> Out["协同制定内容策略与高效矩阵分发"] end ``` --- ### [11:15](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=675s) 透明化的 AI 增强：语言翻译、口型同步与滤镜文化创作者在内容制作本身引入 AI 时，必须面对如何维护观众信任的课题。Meta 认为，创作者完全可以在保持内容真实性的前提下，借助前沿模型进行显式（explicit）的 AI 增强。这种增强的关键在于对用户完全透明，既能保护信任基石，又能让内容的触达范围与深度超越以往。例如在 Facebook 和 Instagram 上线的 AI 翻译功能，不仅能够将原声以观众的母语重新配音，还能结合先进的口型同步（lip-syncing）技术。目前该翻译功能已支持 18 种语言，在 Facebook 和 Instagram 两大平台上各自每天触达超过 5 亿用户。系统会明确告知用户当前视频经过了 AI 翻译，并允许用户随时切换回原语言，从而完好地保留了真实感。另一种既不欺骗受众又能激发创造力的增强形式是像 Restyles 这样的趣味视频滤镜。这类工具不仅不会误导观众，反而能够通过模仿与二创塑造全新的网络流行文化。以近期月活跃用户达到 5000 万的视频编辑工具 Edits 为例，其早期在印度引发爆发式传播的关键时刻，正是由于用户大量使用其中的“Snow（雪景）”预设滤镜。从未下过雪的印度城市街景被一键转化为梦幻般的冬日仙境，许多人仅仅为了拍摄自己所在城市的雪景特效而争相下载该应用，形成了一场自发的文化狂欢。 Yair Livne 强调了创作者引入 AI 时的核心原则：“But what's key about the augmentation is that it needs to be transparent to the user so it maintains trust, but still enables the content to go further and more meaningfully than before.” - **跨语言无缝触达**：支持 18 种语言的 AI 翻译与唇形对齐，在 Instagram 和 Facebook 上每天各自触达超 5 亿独立用户。 - **信任保障机制**：显式标注 AI 生成标识，并提供随时回退、自由切换语言的交互选项。 - **轻量化创意裂变**：Edits 的 Restyles 滤镜月活达 5000 万，印度的雪景特效（Snow preset）成为打破地域文化隔阂的现象级爆款。 --- ### [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) 真实感的回归与 AI 时代创作者的永恒价值除了通过技术进行显式增强，第三条路径则是更加深入地拥抱粗粝、真实、未经修饰的内容，以唤醒人与人之间最纯粹的情感联结。第一人称视角（POV）设备（如 Meta 智能眼镜 Ray-Ban Meta）正在重塑这一体验，让观众能够真正“通过另一个人的双眼看世界”。现场展示了 NBA 球星 Karl-Anthony Towns 在训练赛中的第一人称 POV 视角，将观众瞬间沉浸式带入职业运动员的真实视野。与此同时，曾在疫情高潮后进入低谷的直播（Live streaming）形态目前正在强势复苏。直播所具备的真诚、未经排练且毫无伪饰的氛围，正是吸引观众在 AI 时代重新大规模涌入直播间的根本原因。展望未来，AI 制作工具与底层模型的演进不可逆转，目前为人诟病的粗制滥造与“烂俗（slop）”感必将随着技术演进烟消云散。如同曾经的文本、图像和传统视频一样，AI 内容必将走向全面专业化。在微短剧（micro dramas）领域，这种颠覆已经在发生——行业建立仅数年，如今完全由 AI 生成的微短剧在视听水准上往往已经超越了普通的真人短剧。例如 ByteDance 发布的工具，用户仅需上传一份剧本即可一键端到端生成整部微短剧。高品质的 AI 原生内容必然会凭借其实力赢取相当可观的市场份额。然而，Yair Livne 坚信，伴随着精良 AI 内容的泛滥，大众对“本真性（authenticity）”的渴望只会更呈指数级增长。最优秀的人类创作者仍将占据市场的核心版图，因为人与人之间的认同与共情是算法永远无法取代的。在全场演讲的尾声，Yair Livne 总结道：“Humans will continue rooting for, identifying with, and looking up to other humans.” 尽管内容的外在形态会持续迭代，但促进人与人之间深层联结的创造力，将永远拥有不可替代的核心地位。 - **沉浸式硬件赋能**：Meta 智能眼镜打破传统镜头感，以 Karl-Anthony Towns 等第一人称视角呈现未经修饰的专业运动体验。 - **直播流量回流**：即兴、无剧本的原生真实感，驱动直播业务在后疫情时代实现逆势反弹。 - **全 AI 微短剧浪潮**：剧本端到端直出影视级短剧的工具涌现，推动 AI 内容迅速摆脱廉价质感并迈向专业化。 - **不可替代的情感锚点**：AI 生成内容占比越高，市场对人类真实经历与原始情绪的溢价就越高。 ## Notable quotes > [13:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=780s) Humans will continue rooting for, identifying with, and looking up to other humans. — Yair Livne > [05:00](https://www.youtube.com/watch?v=CYuO9-rau5Q&t=300s) By adopting AI content, creators risk turning off their audience and making the relationship to them feel cheapened and betrayed. — Yair Livne ## Mentioned Yair Livne, Ime Archibong, 马斯洛, Meta, Facebook, Instagram, Threads, Quora, MySpace, Twitter, iPhone, Snapchat, Alan Walker, Netflix, Alix Earle, N3on, Higgsfield AI, Hank Green, Claude, MrBeast, TikTok, YouTube, Mark Zuckerberg, Edits, Karl-Anthony Towns, Ray-Ban Meta, ByteDance [Source](https://www.youtube.com/watch?v=CYuO9-rau5Q)

Source quotes

[13:00] Yair Livne: Humans will continue rooting for, identifying with, and looking up to other humans. [5:00] Yair Livne: By adopting AI content, creators risk turning off their audience and making the relationship to them feel cheapened and betrayed.

Source quotes

[13:50] Yair Livne: Humans will continue rooting for, identifying with, and looking up to other humans. [12:55] Yair Livne: Live streams create a sincere, unrehearsed, and authentic vibe, which is why I believe consumers are flocking back to them now.

Original transcript

[0:00] Last year, my colleague Imei Archibong, talked about how the age of AI presents challenges to authentic human connection, but also opportunities to connect people in new ways. He may ground at his talk in Maslow's Hierarchy of Needs. and specifically at the middle of the hierarchy, human connection, between friends, family and strangers. I'll be focused on the very top of the hierarchy, the part I'm most excited about. Self-actualization. When we are nourished and safe and feel a connection to others and feel respected, That opens the door for us to self-actualize through creative expression and autonomy. Creative expression is so important for us as individuals, but it also is meaningful at a societal level. Entertainment, culture, and art are a huge part of society because they allow us to connect to others when they're self-actualizing. I'm Eri Lévenet. I've been building public content ecosystems for the majority of my career, and specifically building for the creators inside those public content ecosystems. First at Quora and more recently at Meta, where my team works and builds for creators on Facebook and works closely also with the Instagram edits, threads and news teams. I really take pride and joy from building for those special creators who bring so much value into these public content ecosystems. These platforms are also markets shaped by the forces of supply and demand. As a recovering economist, I find it fascinating to study those forces, but I also find they enable us to build better products. and specifically better products for creators. Today, as we are accelerating into the age of AI, Creativity and creators are facing critical choices. Before we dive into those, let's talk about how we got here. This chart tracks how many people on a monthly basis are posting online across text, photos, and video posts. As you can see, things really kick off in the early 2000s. Personal websites and later dedicated platforms like MySpace and Facebook and Twitter allow people initially millions and hundreds of millions, ultimately billions, to reach the worst of the world. In the early 2010s, with the rise of the iPhone and improving camera technology, and then platforms like Instagram and Snap, Everyone can become a professional photographer. And then, by the late 2010s, with even better cameras on phone, better internet connectivity, and short-form video platforms, everyone becomes a video creator. As each of these formats takes off, there are structural changes in who controls the means of production. As your internet has unleashed each of these mediums, we see three consistent phases for the disruption. Each of these waves of media explosion are predated by an old guard of professionals holding the keys to creation. In the case of tech, these are newspapers, In the case of video, these are TV networks and movie studios. But then, as technological capabilities come into the hands of millions or hundreds of millions of people, and platforms come to support those formats, The behavior becomes mass market. This phase is inherently messy and noisy. There are no clear quality signals and no clear trust signals telling consumers who to pay attention to. In phase 3, out of this semester, a new class of professional arises. Today we call them creators. They get extremely good at utilizing the format and the platforms to hone a new art and a new form of business. But what is happening on the demand side as these formats break through? Where are consumers spending time? This chart tracks how many minutes per day are going to each format on average. As we can see, text and photos break through and ultimately flatten. But what is not flattering is video. That pink line is continuing to grow Even now with no signs of slowing down. Let's get a taste of a super viral recent reel from Alan Walker. *MUZIE* So that was fun. The vast majority of this time spent on video goes to creator content, and this is why video creators run the show today and have dominated consumer mindshare. Going back to our three phases of supply, Video creators are the new garden video. Every platform is competing for their content and their loyalty, and even traditional platforms like Netflix, who historically have focused on produced content, are now bringing on board creator content to stay relevant. Alex Earl, Earlier this month had a top 10 US Netflix show that was not imaginable even two years ago. Creators are also a huge business through the tune of almost half a trillion dollars a year. This comes through a combination of multiple revenue streams from brand deals to creator, [5:00] to tipping, to merchandising, affiliate marketing, and platform payouts. Some creators have built businesses valued at hundreds of millions or even billions of dollars. So this is the state of the world pre-AI, but as we all know, we are now in the first phases of the AI-fueled creativity era. AI is special in that with multi-modal models, it can transform any type of media and sometimes all of them at once. We are so early, but this blue line is large and quickly accelerating. Already hundreds of millions of people are posting AI-generated content every month. So how should creators respond? On the one hand, AI offers them a new exciting canvas where everything is possible. What was potentially limited or hard earlier with just a phone camera is now limitless. On the other hand, we all know that in the context of content creation, AI has also come to be associated with flop. Why? AI content is super easy and cheap to produce, and sometimes at scale. A lot of what we see today on platforms doesn't really feel high quality. There's a cheap nature to it. and the worst AI content feels like it's trying to trick the audience into a negative emotion. Disgust, pity, anger, All these factors sum up to a feeling of a breach of trust between audiences and creators. By adopting AI content, creators risk turning off their audience and making the relationship to them feel cheapened and betrayed. Neon, a prominent Twitch streamer recently declared he was retiring and becoming a the first ever live AI streamer in partnership with Higgsfield AI. Neon got an avalanche of hate from both his audience and other creators, saying that his choice was lazy and he was really selling out. Clips of the stream quickly went viral for their poor quality. Let's take a look. Is it a bowling ball? Oh, hello. Oh, wait. Just hit the button. I'm upside down. Again, hold it. See, nothing to it. - Again! You asshole, help me! Should we get out? I don't know. So this seems fun and loud, but we're also seeing some real issues. Like physics problems when the guy's hand opens the box, but the box goes through it. If you tune into this livestream, you'll see a lot more inconsistencies, both plot holes, characters coming in and out, and more physics problems. This immediately puts the viewer in a place where trust breaks. You can't trust that somebody actually cares about the experience you're watching. And this brings us back to Maslow's hierarchy. Below self-actualization we have connection and trust And without those... That hierarchy collapses. Let's look at a couple more examples, a little less extreme. It turns out that even catching a whiff of inauthenticity can really backfire on Creator. Hank Green, a successful YouTuber that my kids happen to really like, had to fully pause multiple of his channels after his audience became suspicious that he was using AI to generate his script. This was based on just one phrase that sounded like Claude speak. He issued profound apologies, before being able to come back online. And even the biggest creators are not immune to this. Mr. Beast, perhaps the most successful creator out there, had to pull back an AI-generated thumbnail maker that he put out there that enabled small creators to adapt bigger creator styles, after feeling he was losing trust with his audience, for enabling inauthentic behavior at scale. No creator is immune to this corrosion of trust. So where are the Sleeve Creators? All these examples drive home that creators need to choose. How do they react to this moment? Do they ignore it, stay true to the old ways, but risk staying behind? Go all in, fast adopt, maybe even replace themselves? Do they experiment and dabble, take on some risk, but try to be future-proof? I'm here to argue that while fully replacing your current creation or yourself with AI really risks creators breaking trust with their audience, we are already seeing how creators are able to react to this AI moment. by leaning into both AI technologies and others, experimenting, but really keeping that most expensive currency, trust. Let's talk about these opportunities. The first one is using AI as personal super intelligence. This comes in before the content is even made. Large creators usually have teams of people working for them to help them produce their content. manage their presence on different platforms, and figure out what to create next. But AI can serve as a creator's social media team, even if they don't have the ability or money to hire one yet. All of TikTok, YouTube, edits, and Facebook have come out with AI assistance to help creators create better content without jeopardizing the audience's trust. Muse, despite being out [10:00] The market for only a couple of months is already fast becoming a powerful creator tool. We see creators use it to manage their Facebook page or Instagram presence with a lot more leverage. At Facebook, we built an assistant following Mark's vision of personalized super intelligence. We wanted to provide a tool that can help creators with all their needs in a hyper-personalized way. Our dedicated harness can invoke four subagents, 16 skills, and 40 tools growing by the month across key creator needs like getting content insights, grafting captions, responding to comments, and brainstorming new ideas. All these abilities are deeply rooted in the creator's insights and history, knowing which of their content resonates with their audience, and enabling them to co-create a content strategy. I've seen the assistant tell a creator which key moment in their video enabled that video to really go viral and then use that insight in their future content. The creator assistant was built by a lean talent dense team using the latest and greatest agentic flows, enabling us to cut down development times by 3 to 4x versus traditional approaches. And creators love it. This product has been in the market for only 4 months and it's already used by millions of creators on a weekly basis with really strong retention. - So this first opportunity is all about enabling creators before content is made, but can creators use AI to help with the content itself without breaking trust? Creators can keep their content real but explicitly augment it with AI using the latest and greatest models, But what's key about the augmentation is that it needs to be transparent to the user so it maintains trust, but still enables the content to go further and more meaningfully than before. while appearing to viewers in the viewer's own native language, with both lip-syncing and dubbing. AI translations are already in 18 languages on Instagram and Facebook, reaching over half a billion users on a daily basis on each platform. Users explicitly see that the video is using AI translations and can even swap languages, So authenticity is maintained. Restyles is another great example. These are fun, easy filters to apply on short videos that are clearly playful and not deceiving an audience. At the same time, they can unlock creativity and mimicry and shape culture. One of the early breakout moments for edits, which recently reached 50 million now, was going really viral in India for featuring Indian cityscapes and never see snow as winter wonderlands. Using the Snow preset on the app. People are downloading edits just to capture the snow effect in their city. Let's take a look at what that cultural moment looked like. *music* Finally, there's a third path, leaning harder into real, authentic, and sometimes raw content that taps into human-to-human connection. POV videos taken by devices like MetaGlasses can enable you to literally see through the eyes of another human being. What could be more authentic than that? Imagine being transported into an NBA practice and see through the eyes of Carl Anthony Towns. Let's take a look at what that looks like. In parallel, we're also seeing a resurgence in live streaming. Live hit a lull post the pandemic peak, but is now growing again. Live streams create a sincere, unrehearsed, and authentic vibe, which is why I believe consumers are flocking back to them now. And with more demand, we see more creators leaning into the format as well. But let's be sincere, looking ahead, the reality is that AI models and tooling around them will only get better and more compelling. The sloth feeling that we see today will go away. Just like text, images, and video, that all got professionalized over time. This will happen with AI content as well. We are already seeing this in the micro drama space, where industry that just got built in the last few years is now being inundated by full AI micro dramas that are often better than the average live action ones. And this trend will only be accelerating now that we have tooling like Draw Magic that came out of Bintedance two weeks ago, enabling the creation of an end-to-end full microdrama with just the upload of a script. With this higher quality AIA content coming, it will appropriately earn market share. But with that, the demand for authenticity will naturally rise too. I truly believe that the best human creators will continue to have a huge part of the overall market, bringing their authentic and original content to meet that growing demand. Humans will continue rooting for, identifying with, and looking up to other humans. The nature of their content might continue to evolve, but creativity [15:00] Foster's human connection. will always have a big place. Thank you.