Dec 13, 2024 · 49m · american-optimist
The Terrifying Truth AI Researchers Don’t Understand · Joe Lonsdale
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Host Joe Lonsdale interviews tech entrepreneur and Cosmos Institute founder Brendan McCord about the urgent need to ground artificial intelligence development in Western philosophical principles. Together, they analyze competing AI ideologies, the risks of over-automation and techno-authoritarianism, and how cultivating 'philosopher builders' can safeguard human freedom and flourishing.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Joe holds 29.2% of the talking time here. How this is scored →
Joe as informed peer 5.9 Guest teaching 4.9 Guest disagreement 1.2 Joe pushing back 1.1
speaking balance: gold is Joe, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 22:35 McCord dismantles EA utilitarian premise
McCord forcefully rejects the foundational premise of the AI safety and effective altruism movement, arguing that human morality cannot be reduced to a utilitarian math equation.
Hardest push from Joe ▶ 44:02 Lonsdale challenges McCord's Oxford strategyLonsdale pushes back on McCord embedding his lab at Oxford, calling the institution a cynical hub for doomsayers and the epicenter of an intellectual rot.
Biggest teaching moment ▶ 31:04 McCord explains ex-ante vs ex-post jurisprudenceMcCord delivers a clear legal and philosophical lecture distinguishing between preventative statutory regulation and adaptive common law Scrabble-board evolution.
Joe holds their own ▶ 36:08 Lonsdale articulates Texas transparency policy proposalLonsdale demonstrates deep domain knowledge and policy expertise, presenting a detailed legislative roadmap for state governments to mandate disclosure and check federal NGO pressure.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Joe as informed peer | Guest teaching | Guest disagreement | Joe pushing back | Why |
|---|---|---|---|---|---|---|
| Brendan McCord's Background: From Submarines to AI | 5 | 2 | 0 | 0 | Lonsdale walks McCord through his background from submarine service to Project Maven and venture creation. The dynamic is conversational and collaborative, with Lonsdale chiming in with industry familiarity regarding Eric Schmidt, Palantir, and Thomas Tull. | |
| Philosophical Awakening and the Cosmos Fellowship | 6 | 6 | 1 | 1 | McCord explains how his philosophical study inspired the Cosmos Institute, drawing parallels across technological history from Franklin to Berners-Lee, as well as Tocqueville and the Strauss-Kojeve debate. Lonsdale engages with his own passion for Enlightenment philosophy and antipathy toward centralized global governance. | |
| The Control Layer Dilemma: Over-Automation & Moral Outsourcing | 6 | 5 | 1 | 1 | Lonsdale queries whether the emergent AI control layer in society is inherently authoritarian. McCord quotes Alfred North Whitehead on automation while warning against moral outsourcing using historical examples of concentration camp guards. | |
| "Autocomplete for Life" and Algorithmic Echo Chambers | 6 | 4 | 1 | 1 | McCord introduces the concept of avoiding 'autocomplete for life' and scripts for human vulnerability. Lonsdale agrees, citing J.S. Mill and explaining why he restricts his children's screen exposure due to addictive dopamine loops. | |
| Provocateur AI and Decision Quality | 6 | 4 | 1 | 2 | McCord outlines provocateur AI designed to challenge comfort zones. Lonsdale pushes back slightly on consumer adoption, prompting McCord to explain how professional domains like Bloomberg terminals prioritize decision quality over mere engagement. | |
| The AI Doomsayers: Effective Altruism and Rationalism | 5 | 6 | 2 | 1 | McCord deconstructs the AI doomsayer ideology combining EA, rationalism, and longtermism, rejecting the premise that incommensurable moral goods can be reduced to utilitarian calculations. Lonsdale facilitates and connects this mindset to tech figures like Elon Musk. | |
| Accelerationism vs. Human Flourishing | 6 | 6 | 2 | 1 | McCord critiques pure accelerationism for confusing thermodynamic means with human ends, advocating Aristotelian flourishing rooted in reason, autonomy, and decentralization. Lonsdale engages with the philosophical hierarchy of means versus ends for their own sake. | |
| Decentralized AI vs. The Regulatory Threat | 6 | 5 | 2 | 2 | The conversation examines centralized vs decentralized AI architectures, comparing current assumptions to early telecom copper monopoly theories. Lonsdale highlights Peter Thiel's pro-authoritarian framing of AI and stresses the need to check both big government and corporate power. | |
| Ex-Post Regulation and Simple Legal Frameworks | 6 | 7 | 1 | 1 | Lonsdale laments Texas legislators attempting heavy-handed AI regulation. McCord educates on the legal distinction between ex-ante civil statutory rules and ex-post adaptive common law frameworks, citing Richard Epstein's Simple Rules for a Complex World. | |
| Techno-Authoritarianism and Countering Government Collusion | 8 | 2 | 1 | 2 | Lonsdale takes charge of the segment, formulating a detailed state-level legislative policy agenda focused on radical disclosure of federal and NGO collusion rather than creating new bureaucratic regulators. McCord expresses interest in the state dynamic. | |
| AI Governance Risks and Psychological Reliance on AI | 5 | 5 | 1 | 1 | McCord references a Wyoming mayoral candidate campaigning as an AI avatar to illustrate human psychological willingness to defer authority to algorithmic judgment. Both discuss the risks of trusting unprincipled models in politics. | |
| Limits of AI Reasoning and Socratic Inquiry | 6 | 7 | 2 | 1 | McCord highlights research showing AI reasoning benchmark drops under minor perturbations because researchers lack a positive definition of reasoning. Lonsdale suggests structural Socratic self-questioning mechanisms. | |
| Cultivating Philosopher Builders: The Cosmos Mission | 6 | 5 | 2 | 2 | McCord details Cosmos Institute's mission to bridge philosophy and code across Oxford and Austin. Lonsdale pushes back on Oxford being the epicentre of EA nihilism, but McCord explains Oxford's decentralized college tradition and the strategic value of entering the intellectual fray. | |
| An Optimistic AI Vision and Final Thoughts | 6 | 4 | 0 | 0 | Lonsdale asks for the best-case optimistic future for AI. McCord outlines breakthroughs in epistemic infrastructure, rapid scientific discovery in cancer cures, and mass liberal education, closing with mutual alignment on philosopher builders. |
Statements from this episode (25)
McCord: 4,000 Google Employees Protested Project Maven Military AI Work
“After I left, after I rolled off the project, we had deployed it in Afghanistan, all that, 4000 people came together and they protested Google's involvement.”
Lonsdale: Wealthy Founders Try Fixing Society While Lacking Enlightenment Philosophical Grounding
“I'm obsessed with, and it's the biggest frustration for me, because our friends will make a lot of money, and they have no idea about these enlightenment values, and then they try to apply things to fix our society. That are completely missing the wisdom of wh…”
McCord: Tech revolutions offer two paths, building for freedom or control
“Whenever you have a technological revolution, you have two choices. You either build for freedom, or you watch as others build for control”
McCord: AI infrastructure spending will reach $1T by early 2030s
“I think, you know, it'll be the biggest infrastructure build that humanity has ever done, I think. I think you could make that argument. I think we will hit a trillion dollars in the early 20 thirties.”
Lonsdale: Computers dictate 20% to 50% of work in many industries
“Already there's a lot of industries
Where like, 20% or even 50% of people's time and action is, is in some ways dictated by the computer.”
McCord: Outsourcing moral deliberation leads ordinary people to commit atrocities
“Normal people, people who don't have great evil you know, genetically or pre, or, you know, they weren't engaging in active malice. Became concentration camp guards. Because they outsourced gradually, you know, bit by bit, pieces of their moral deliberation.”
McCord: Human biological passivity makes AI cognitive offloading uniquely dangerous
“Human brain works on glucose. There's only so much glucose to go around. We're biologically wired to be passive. So AI presents this very interesting, sinister. It's very good in the sense that it can offload plausibly everything, but it's bad in the sense tha…”
Lonsdale restricts children's screen time, calling current AI design a drug
“We don't actually use screens very much, because I'm a little bit afraid of the interactions with them, and like what, you know, I see a lot of this AI design as it is right now, as like, it's a form of drug in some sense, right? Because basically it's trying …”
McCord: AI personalization risks trapping users in self-curated realities
“The other thing that I think is concerning is it could create a hall of mirrors culture where everyone has a kind of a self curated reality.”
McCord: AI founders should target markets where decision quality outweighs engagement
“When you use a Bloomberg terminal, you actually want to optimize for decision quality. When you are scrolling at night, you're kind of in engagement mode, right? So what does that tell an entrepreneur? Well, it tells maybe, like, look at those markets Where li…”
McCord: EA, Rationalism, and Longtermism Dominate the AI Practitioner Community
“That's the most common dominant philosophy in the practitioner community, and it's three philosophies in a trench coat, but they're all very entangled. One is effective altruism, One is rationalism, one is long-termism.”
McCord: Accelerationism confuses tech progress as an end rather than means
“And so it becomes like the ultimate goal is almost the ultimate end of human life is a kind of technological progress, or it's a progress in harnessing energy. I think that completely confuses the means with the ends. I think technology is incredible equipment…”
McCord: Treating AI progress as ultimate end risks abandoning human defense
“To say that that is the ultimate end of human life, that that's what we move towards, gives you these kind of sinister conclusions, which is like, we should pass the baton to this higher, better intelligence if we can. You know, in other words, like. There's n…”
McCord: Prime Intellect's Decentralized Training Run Challenges Centralization Assumptions
“There's a huge decentralized training run that I think prime intellect did recently that, you know, shows that the assumptions everyone has about how AI is just going to be a centralized thing could be wrong.”
Lonsdale: Regulators Are the 'Lowest IQ' Group Stifling AI Innovation
“I think this third area is like the lowest IQ area of these four, which is the regulators basically responding to every challenge with rules that just stifle new things way ahead of time.”
McCord: Complex systems adapt best under simple property and liability rules
“Actually, the best way in complex systems is to have
A very simple uniform set of rules like property rights or contract law or tort liability.
And these things are much better at adapting than trying to come up with a patchwork of systems that just lead to in…”
Lonsdale: European regulators are banning OpenAI
“Such as the Europeans who are banning OpenAI, by the way, which is very funny for the latest launch.”
McCord: AI Regulation Offers Government a Unique Vector for Power Expansion
“AI touches everything, and so AI regulation is a unique vector for, like, government Power.”
McCord: US States Seek Over 700 Different AI Regulations
“The vast majority of states are pretty regulation happy and want to put up, like, 700 plus different versions.”
Lonsdale: Texas Should Mandate Big Tech Transparency Instead of AI Regulators
“Forced transparency on that, and the big tech would hate it, and I think that's the right thing for Texas to do, so don't create a regulator, just like demand they do these things, you know?”
McCord: A Wyoming mayoral candidate ran pledging to defer decisions to ChatGPT
“So this forty-year-old guy in Wyoming ran for mayor. Recently, and he, his name is Victor, and he runs as the humble meat avatar of an, of a chat GPT, of AI system. So he says, like, I'm gonna run, because humans can run from air, AI can't, but really I'm just…”
McCord: Apple paper shows minor perturbations drop AI reasoning scores by 65%
“But then you have Apple come out with a paper that shows that if you do minor perturbations, you drop the reasoning score on these systems by like, 65%.”
McCord: AI researchers do not really know what reasoning is
“Well, I think the reason for that is AI researchers don't really know what reasoning is.”
McCord: Silicon Valley culture is trapped in mimetic ideological bubbles
“There are these, like, strong mimetic bubbles in Silicon Valley, where people don't even really think they're busy, and they're like, all my other friends care about existential risk, I'm in, you know.”
Lonsdale: Musk, Thiel, and Koch are philosopher builders
“A lot of the builders who I admire most, whether it's Elon Musk, Peter Thiel, Charles Koch, they're deeply enmeshed in philosophy. They're philosopher builders, and they've taken the time to go and kind of work on society to make sure our society stays somethi…”