StarZero All shows Try the podcast clipper →

Feb 29, 2024 · 32m · how-i-built-this

The peril (and promise) of AI with Tristan Harris: Part 2

Tristan Harris · 18m spoken Guy Raz · 10m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this episode of How I Built This Lab, host Guy Raz interviews Center for Humane Technology co-founder Tristan Harris to examine the profound societal risks posed by unconstrained generative AI. They explore the collapse of institutional trust, the necessity of corporate legal liability, and historical models for international governance.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Guy holds 37.4% of the talking time here. How this is scored →

Guy as informed peer 4.9 Guest teaching 6.1 Guest disagreement 2.0 Guy pushing back 2.4
05100:0010:0020:0030:001:43–4:45 · Guy as informed peer 4/10 Silicon Valley Hype and Market Incentives for Safety Raz opens by observing the uncritical hype in Silicon Valley and questioning how market incentives can be restrained. Harris reframes the problem away from individual actors toward legal frameworks, providing historical context on Section 230 and corporate liability.4:46–10:03 · Guy as informed peer 5/10 Synthetic Media and Societal Immune System Overload Raz shares an anecdote about a convincing synthetic surfing video and pressing security risks like password cracking. Harris illustrates societal immune system failure and challenges the defeatist talking point regarding falling behind China in AI capability.10:05–15:22 · Guy as informed peer 5/10 Mid-Roll Break: Policy and Executive Actions Preview Raz compares the breakdown of shared evidence to the sci-fi plot of The Running Man. Harris validates the analogy and introduces the 'Windows 95' metaphor to argue that society needs a foundational systemic upgrade.15:23–18:46 · Guy as informed peer 5/10 Evaluating the White House Executive Order on AI Raz probes the tangible impact of the White House Executive Order on AI, questioning whether it provides enforceable protections or merely symbolic guidance. Harris breaks down specific policy levers like federal funding conditions and compute thresholds.18:48–24:10 · Guy as informed peer 6/10 Mid-Roll Break: Global Cooperation and Personal Drive Raz distinguishes AI from historical nuclear proliferation by noting that AI includes decentralized individual actors rather than just nation-states. Harris counters by linking uranium to Nvidia GPU supply chains and warning against open-sourcing advanced models.24:11–27:20 · Guy as informed peer 6/10 Making Dark Futures Legible Through Shared Consciousness Raz draws on the cultural impact of the 1983 film The Day After to suggest that dramatizing dark futures can foster public consciousness. Harris agrees, emphasizing that making worst-case scenarios legible is essential to steering policy toward safety.27:21–28:48 · Guy as informed peer 3/10 Actionable Timelines and Grassroots AI Advocacy Raz expresses a sense of public helplessness given the rapid pace of technological change. Harris offers a 12-month critical window and outlines concrete grassroots actions, urging listeners to focus on systemic liability reform.1:43–4:45 · Guest teaching 6/10 Silicon Valley Hype and Market Incentives for Safety Raz opens by observing the uncritical hype in Silicon Valley and questioning how market incentives can be restrained. Harris reframes the problem away from individual actors toward legal frameworks, providing historical context on Section 230 and corporate liability.4:46–10:03 · Guest teaching 6/10 Synthetic Media and Societal Immune System Overload Raz shares an anecdote about a convincing synthetic surfing video and pressing security risks like password cracking. Harris illustrates societal immune system failure and challenges the defeatist talking point regarding falling behind China in AI capability.10:05–15:22 · Guest teaching 6/10 Mid-Roll Break: Policy and Executive Actions Preview Raz compares the breakdown of shared evidence to the sci-fi plot of The Running Man. Harris validates the analogy and introduces the 'Windows 95' metaphor to argue that society needs a foundational systemic upgrade.15:23–18:46 · Guest teaching 7/10 Evaluating the White House Executive Order on AI Raz probes the tangible impact of the White House Executive Order on AI, questioning whether it provides enforceable protections or merely symbolic guidance. Harris breaks down specific policy levers like federal funding conditions and compute thresholds.18:48–24:10 · Guest teaching 7/10 Mid-Roll Break: Global Cooperation and Personal Drive Raz distinguishes AI from historical nuclear proliferation by noting that AI includes decentralized individual actors rather than just nation-states. Harris counters by linking uranium to Nvidia GPU supply chains and warning against open-sourcing advanced models.24:11–27:20 · Guest teaching 5/10 Making Dark Futures Legible Through Shared Consciousness Raz draws on the cultural impact of the 1983 film The Day After to suggest that dramatizing dark futures can foster public consciousness. Harris agrees, emphasizing that making worst-case scenarios legible is essential to steering policy toward safety.27:21–28:48 · Guest teaching 6/10 Actionable Timelines and Grassroots AI Advocacy Raz expresses a sense of public helplessness given the rapid pace of technological change. Harris offers a 12-month critical window and outlines concrete grassroots actions, urging listeners to focus on systemic liability reform.1:43–4:45 · Guest disagreement 2/10 Silicon Valley Hype and Market Incentives for Safety Raz opens by observing the uncritical hype in Silicon Valley and questioning how market incentives can be restrained. Harris reframes the problem away from individual actors toward legal frameworks, providing historical context on Section 230 and corporate liability.4:46–10:03 · Guest disagreement 3/10 Synthetic Media and Societal Immune System Overload Raz shares an anecdote about a convincing synthetic surfing video and pressing security risks like password cracking. Harris illustrates societal immune system failure and challenges the defeatist talking point regarding falling behind China in AI capability.10:05–15:22 · Guest disagreement 1/10 Mid-Roll Break: Policy and Executive Actions Preview Raz compares the breakdown of shared evidence to the sci-fi plot of The Running Man. Harris validates the analogy and introduces the 'Windows 95' metaphor to argue that society needs a foundational systemic upgrade.15:23–18:46 · Guest disagreement 2/10 Evaluating the White House Executive Order on AI Raz probes the tangible impact of the White House Executive Order on AI, questioning whether it provides enforceable protections or merely symbolic guidance. Harris breaks down specific policy levers like federal funding conditions and compute thresholds.18:48–24:10 · Guest disagreement 3/10 Mid-Roll Break: Global Cooperation and Personal Drive Raz distinguishes AI from historical nuclear proliferation by noting that AI includes decentralized individual actors rather than just nation-states. Harris counters by linking uranium to Nvidia GPU supply chains and warning against open-sourcing advanced models.24:11–27:20 · Guest disagreement 1/10 Making Dark Futures Legible Through Shared Consciousness Raz draws on the cultural impact of the 1983 film The Day After to suggest that dramatizing dark futures can foster public consciousness. Harris agrees, emphasizing that making worst-case scenarios legible is essential to steering policy toward safety.27:21–28:48 · Guest disagreement 2/10 Actionable Timelines and Grassroots AI Advocacy Raz expresses a sense of public helplessness given the rapid pace of technological change. Harris offers a 12-month critical window and outlines concrete grassroots actions, urging listeners to focus on systemic liability reform.1:43–4:45 · Guy pushing back 2/10 Silicon Valley Hype and Market Incentives for Safety Raz opens by observing the uncritical hype in Silicon Valley and questioning how market incentives can be restrained. Harris reframes the problem away from individual actors toward legal frameworks, providing historical context on Section 230 and corporate liability.4:46–10:03 · Guy pushing back 3/10 Synthetic Media and Societal Immune System Overload Raz shares an anecdote about a convincing synthetic surfing video and pressing security risks like password cracking. Harris illustrates societal immune system failure and challenges the defeatist talking point regarding falling behind China in AI capability.10:05–15:22 · Guy pushing back 1/10 Mid-Roll Break: Policy and Executive Actions Preview Raz compares the breakdown of shared evidence to the sci-fi plot of The Running Man. Harris validates the analogy and introduces the 'Windows 95' metaphor to argue that society needs a foundational systemic upgrade.15:23–18:46 · Guy pushing back 4/10 Evaluating the White House Executive Order on AI Raz probes the tangible impact of the White House Executive Order on AI, questioning whether it provides enforceable protections or merely symbolic guidance. Harris breaks down specific policy levers like federal funding conditions and compute thresholds.18:48–24:10 · Guy pushing back 4/10 Mid-Roll Break: Global Cooperation and Personal Drive Raz distinguishes AI from historical nuclear proliferation by noting that AI includes decentralized individual actors rather than just nation-states. Harris counters by linking uranium to Nvidia GPU supply chains and warning against open-sourcing advanced models.24:11–27:20 · Guy pushing back 1/10 Making Dark Futures Legible Through Shared Consciousness Raz draws on the cultural impact of the 1983 film The Day After to suggest that dramatizing dark futures can foster public consciousness. Harris agrees, emphasizing that making worst-case scenarios legible is essential to steering policy toward safety.27:21–28:48 · Guy pushing back 2/10 Actionable Timelines and Grassroots AI Advocacy Raz expresses a sense of public helplessness given the rapid pace of technological change. Harris offers a 12-month critical window and outlines concrete grassroots actions, urging listeners to focus on systemic liability reform.

speaking balance: gold is Guy, purple is the guest (3 minute bins)

0:00 · Guy 84.9% · guest 15.1%0:00 · Guy 84.9% · guest 15.1%3:00 · Guy 36.1% · guest 63.9%3:00 · Guy 36.1% · guest 63.9%6:00 · Guy 24% · guest 76%6:00 · Guy 24% · guest 76%9:00 · Guy 63.1% · guest 36.9%9:00 · Guy 63.1% · guest 36.9%12:00 · Guy 23.8% · guest 76.2%12:00 · Guy 23.8% · guest 76.2%15:00 · Guy 17.3% · guest 82.7%15:00 · Guy 17.3% · guest 82.7%18:00 · Guy 51.5% · guest 48.5%18:00 · Guy 51.5% · guest 48.5%21:00 · Guy 7.9% · guest 92.1%21:00 · Guy 7.9% · guest 92.1%24:00 · Guy 44.5% · guest 55.5%24:00 · Guy 44.5% · guest 55.5%27:00 · Guy 25% · guest 75%27:00 · Guy 25% · guest 75%30:00 · Guy 34.3% · guest 65.7%30:00 · Guy 34.3% · guest 65.7%
Sharpest disagreement ▶ 23:40 Denouncing open-source AGI

Harris forcefully condemns Mark Zuckerberg's public commitment to building open-source AGI, calling it the most dangerous course of action possible.

 Hardest push from Guy ▶ 15:23 Challenging the weight of executive orders

Raz refuses to accept high-level executive action at face value, directly questioning whether non-binding executive orders accomplish anything beyond symbolic PR without congressional law.

 Biggest teaching moment ▶ 22:44 Explaining model weight fine-tuning risks

Harris educates Raz on the difference between open-source code and open model weights, explaining how a small fine-tuning budget can strip safety guardrails from models like Llama 2.

 Guy holds their own ▶ 19:40 Distinguishing AI from state nuclear monopoly

Raz demonstrates substantive domain expertise by challenging the nuclear analogy, pointing out that nuclear non-proliferation relied on a small club of nation-states whereas AI involves millions of decentralized commercial and private developers.
the scores for every segment, with the reasoning behind each
ChapterTopicGuy as informed peerGuest teachingGuest disagreementGuy pushing backWhy
Silicon Valley Hype and Market Incentives for Safety 4622 Raz opens by observing the uncritical hype in Silicon Valley and questioning how market incentives can be restrained. Harris reframes the problem away from individual actors toward legal frameworks, providing historical context on Section 230 and corporate liability.
Synthetic Media and Societal Immune System Overload 5633 Raz shares an anecdote about a convincing synthetic surfing video and pressing security risks like password cracking. Harris illustrates societal immune system failure and challenges the defeatist talking point regarding falling behind China in AI capability.
Mid-Roll Break: Policy and Executive Actions Preview 5611 Raz compares the breakdown of shared evidence to the sci-fi plot of The Running Man. Harris validates the analogy and introduces the 'Windows 95' metaphor to argue that society needs a foundational systemic upgrade.
Evaluating the White House Executive Order on AI 5724 Raz probes the tangible impact of the White House Executive Order on AI, questioning whether it provides enforceable protections or merely symbolic guidance. Harris breaks down specific policy levers like federal funding conditions and compute thresholds.
Mid-Roll Break: Global Cooperation and Personal Drive 6734 Raz distinguishes AI from historical nuclear proliferation by noting that AI includes decentralized individual actors rather than just nation-states. Harris counters by linking uranium to Nvidia GPU supply chains and warning against open-sourcing advanced models.
Making Dark Futures Legible Through Shared Consciousness 6511 Raz draws on the cultural impact of the 1983 film The Day After to suggest that dramatizing dark futures can foster public consciousness. Harris agrees, emphasizing that making worst-case scenarios legible is essential to steering policy toward safety.
Actionable Timelines and Grassroots AI Advocacy 3622 Raz expresses a sense of public helplessness given the rapid pace of technological change. Harris offers a 12-month critical window and outlines concrete grassroots actions, urging listeners to focus on systemic liability reform.

Statements from this episode (13)

Opinion
Harris: Applying Section 230 to Social Media Shielded Addictive Tech From Liability
“And that made sense when the internet was just a bunch of bulletin boards and it was not powered by AI, but we used that immunity shield and applied it to social media companies when they came along later. And so when they go and intentionally addict children,…”
Tristan Harris Feb 29, 2024 ▶ 3:44
Prediction Not checkable as stated
Harris: Anything that can be emulated will be generated by AI
“So AI is not going to get worse at emulating someone's voice, someone's handwriting, someone's likeness, their This, that's what generative AI does, is it gives you those capabilities. Anything that can be emulated will be emulated, and that's why it's generat…”
Tristan Harris Feb 29, 2024 ▶ 5:52
Prediction Not checkable as stated
Harris: Generative AI will force transition to secure, authenticated communications
“I think that in the age of generative AI, we're going to move to these more privileged and secure environments.”
Tristan Harris Feb 29, 2024 ▶ 7:34
Insight
Harris: AI capabilities are outpacing society's ability to adapt
“We're just simply releasing more capabilities into society faster than society has the immune systems to absorb and adapt to all the new changes that accompany all of that AI getting released.”
Tristan Harris Feb 29, 2024 ▶ 8:26
Opinion
Harris: The US-China AI race is about responsible deployment, not speed
“The U S has to lead and say, we need to set the terms of the race. And it's actually a race to the responsible and conscious deployment of technology that in its effect strengthens your society relative to other ones. That's the true competition.”
Tristan Harris Feb 29, 2024 ▶ 9:50
Insight
Harris: Future internet media could require watermarks to verify authenticity
“We have to come up with new meaning for what is evidence in a world where AI can generate that evidence. And there are ways of doing that. We could live in a world where the only places that any media you see on the internet Will only be on the internet if it'…”
Tristan Harris Feb 29, 2024 ▶ 14:49
Assertion Supported
Harris: American and Chinese AI scientists hold informal safety dialogues
“I'm happy to report that informally, there are some of those dialogues that are happening between Chinese AI scientists about the risks and us ASI AI scientists about the risks.”
Tristan Harris Feb 29, 2024 ▶ 17:57
Insight
Harris: Corporate Legal Liability Would Force AI Companies to Slow Down
“And I think things like liability as a regulatory framework are powerful because people understand it, right? You as an AI company shouldn't be worried about being liable for the harms if there are not going to be any harms or risks. So if you don't think ther…”
Tristan Harris Feb 29, 2024 ▶ 18:21
Insight
Harris: Nvidia GPUs are the uranium equivalent for AI governance
“In this case, what uranium was for nuclear weapons Advanced Nvidia GPU chips are for AI.”
Tristan Harris Feb 29, 2024 ▶ 21:53
Assertion Supported
Harris: Safety controls on open-source AI models can be removed for $100
“I won't go into technical details, but basically for about a hundred dollars, you can retrain all the safety controls off of it.”
Tristan Harris Feb 29, 2024 ▶ 23:37
Opinion
Harris: Mark Zuckerberg's plan to open-source AGI is extremely dangerous
“Mark Zuckerberg has publicly stated within the last week or so that he wants to build a Open source artificial general intelligence, which is the most dangerous thing you could possibly do.”
Tristan Harris Feb 29, 2024 ▶ 23:51
Opinion
Harris: The next 12 months are decisive for establishing AI guardrails
“In my mind, it's like, it's the next 12 months, it's like everything has to happen”
Tristan Harris Feb 29, 2024 ▶ 27:43
Insight
Harris: Tech policy should be pro-technology and anti-externalities, not binary
“I think we can be pro-technology and anti-externalities, and that's the nuanced position that I want everybody to be in, rather than saying you're either for tech and for acceleration or you're a D cell.”
Tristan Harris Feb 29, 2024 ▶ 31:07
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 800 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more