WHO CLAIMED WHAT
Every position taken in the episode, attributed to its holder, with conviction (0–1) and the verbatim receipt.
Daniel Kokotajlo — 32 claims
opinionconviction 0.742:46
There is roughly a 70% chance that AI development goes horribly wrong — something like AI takeover or a comparable catastrophe, of which human extinction is one possibility.
“70 percent chance that this goes horribly wrong like human extinction”
assertionconviction 0.8520:52
OpenAI and Anthropic are deliberately trying to automate their own AI research so they no longer need human employees, in order to reach superintelligence before rivals.
“trying to automate themselves. They're trying to make it the case that they don't really need human employees anymore.”
opinionconviction 0.812:33
The dominant motive at the top of these companies is power-seeking, not merely commercial, incentives.
“I would describe it as power seeking incentives”
opinionconviction 0.89:35
The 'doomerism' counter-narrative is false and is being pushed by people who stand to benefit from dismissing AI risk.
“This counter-narrative is fairly recent and it's been pushed by the people who stand to benefit from it, and it's not true.”
opinionconviction 0.811:01
The labs' founding safety narratives became rationalizations to justify racing rather than genuine guides to their behavior.
“I increasingly came to think that these were rationalizations to justify what they were doing, rather than sort of like deeply guiding their actual behavior.”
opinionconviction 0.7513:48
You should judge AI leaders like Sam Altman by their actions, not by their shifting public narratives.
“I think you should judge people by their actions, not by their words.”
opinionconviction 0.719:49
He does not believe Sam Altman was genuinely unaware of the equity/anti-disparagement clause; at minimum people close to him knew.
“I think you probably know. And if he didn't know, then people close to him probably did, such as his head lawyer.”
opinionconviction 0.639:19
Anthropic overtook OpenAI in the race mainly through higher talent density and better strategy rather than more resources.
“I guess they have probably higher talent density and better strategy, but not by a lot, but enough to make the difference.”
opinionconviction 0.744:51
Anthropic and Dario have been more willing than competitors to say and do things costly to their bottom line and power, such as the Department of War dispute.
“it seems like Anthropic and Dario have been more willing to say and do things that are costly to the bottom line at least in the last year or so.”
opinionconviction 0.8545:15
None of the AI-lab CEOs should be trusted with the amount of power superintelligence would confer.
“None of these people should be trusted with that much power, basically.”
opinionconviction 0.7543:29
The CEOs privately grant a real chance of extinction but have convinced themselves it will probably be fine and that they should be the one in charge.
“They've all probably convinced themselves that, oh yeah, maybe it'll go horribly wrong, but probably it's going to be fine and probably I should be the one in charge.”
assertionconviction 0.7513:00
These CEOs are racing as hard as they can because each fears that if a rival reaches powerful AI first, that rival could become dictator.
“these powerful CEOs are literally afraid that if the other guy gets there first, he might become dictator.”
opinionconviction 0.758:00
Even if control is retained, superintelligence risks handing a tiny group of people the power of oligarchs or dictators.
“we could very easily end up in a sort of a situation where some tiny group of people are essentially oligarchs or dictators.”
opinionconviction 0.78:00
Dario's 'country of geniuses in the data center' is better described as an 'army of geniuses' — copies of one model owned and commanded by the company.
“I think it would be more accurate to describe it as army of geniuses in the data center”
opinionconviction 0.751:11:14
The default trajectory — building AIs smarter than us, giving them bodies, and putting them in charge — is like choosing a new species that will outcompete us once it no longer needs us.
“isn't this just exactly like us picking a new species that's then going to outcompute us when it doesn't need us anymore?”
assertionconviction 0.931:05
Modern AI systems are not conventional software (lines of code) but trained neural nets, which is why we can't directly read their decision-making.
“modern AI systems are not software in the normal sense. I mean, they are technically software, but they're not lines of code.”
opinionconviction 0.828:20
There is nothing magical about the human brain, so a digital system can perform its functions and superintelligence is achievable.
“I think that there's nothing magical about the human brain.”
opinionconviction 0.753:33
If interpretability research succeeds and we could see what AIs are thinking, loss-of-control scenarios become much less likely.
“I think that it would be much less likely for us to get into those loss of control scenarios if we could just actually see what our AIs were thinking and why and how at any given time, right?”
opinionconviction 0.749:41
Broad job displacement will arrive suddenly rather than gradually, because of intelligence-explosion / recursive-self-improvement dynamics.
“I think it'll be sudden because of the intelligence explosion dynamics, or recursive self-improvement dynamics.”
opinionconviction 0.852:00
Once superintelligence is reached, which jobs remain is a political question, not a technical one, since technically all jobs could be automated.
“I think that's actually a political question, not a technical question.”
opinionconviction 0.856:58
The 'AI will create many new unforeseeable jobs' argument fails, because past automation was narrow while superintelligence automates everything.
“The problem with that is that past technological advancements have been more narrow.”
opinionconviction 0.851:03:16
Waiting to regulate until most people have lost their jobs is already too late, because the companies will likely already have superintelligence by then.
“if you wait until most people have lost their jobs to regulate the AI companies, that's already too late, because they will probably already have super intelligent AI by then”
opinionconviction 0.71:07:36
He recommends Plan A (domestic regulation plus an international deal to continue AI more safely), though he is sympathetic to Plan S (shut it all down).
“Plan A is our recommendation. I think that I'm sympathetic to Plan S, but for reasons we explain, we recommend Plan A instead.”
predictionconviction 0.551:07:49
The most probable default outcome is Plan D — the companies keep racing with little meaningful slowdown, as depicted in AI 2027.
“Yeah, 2027 type of thing where they keep racing, they don't really slow down significantly.”
opinionconviction 0.751:21:00
Total research transparency (publishing training recipes) matters because you don't want biased companies deciding whether their own AIs are safe.
“open science is really important for solving the alignment problem fast enough because you don't want to have biased companies making the decisions about whether the AIs are safe.”
opinionconviction 0.61:31:04
Working lie detectors are good when used on the powerful and dangerous (enabling new totalitarianism) when used by the powerful.
“Broadly speaking, I would say the good uses are when lie detectors are used on the powerful instead of by the powerful.”
opinionconviction 0.751:35:41
Losing jobs threatens not just income but political power, because economic contribution is a source of leverage over governments.
“when people lose their jobs, they're not just threatened with loss of income, they're also threatened with loss of political power.”
opinionconviction 0.61:42:30
In a post-superintelligence world, most of Earth should be preserved while economic expansion is pushed into special economic zones and off-planet.
“you preserve like 99 percent of the Earth, mostly as is, as historic or environmental reasons.”
opinionconviction 0.51:49:00
If humanity never builds powerful AI, civilization will probably still die eventually (nuclear war, pandemic), which is why he is torn on permanently shutting it down.
“if we don't build powerful AI systems eventually, then we're probably going to die as a civilization eventually”
opinionconviction 0.61:47:24
He would immediately press a button for a temporary AI shutdown because civilization is not ready, but would agonize over a permanent one.
“I would totally slam that button because we are not ready to do this.”
opinionconviction 0.81:52:00
The core problem is that people aren't taking AI risk seriously yet; broad public attention would itself produce better, more precise regulation.
“the core problem is that people aren't taking it seriously yet.”
opinionconviction 0.851:52:23
AI is the most important thing happening in our lifetimes, probably in all of history.
“This is the most important thing happening in our lifetimes, probably in all of history, in fact, and it's very important that it go well.”
OpenAI founders — 1 claim
assertionconviction 0.712:33voiced by Daniel Kokotajlo
OpenAI was founded in part because its founders feared Demis Hassabis at Google would become dictator with AGI.
“the reason why we made OpenAI was because we were worried that Demisosavis at Google was going to become dictator with AGI.”
Geoffrey Hinton — 1 claim
assertionconviction 0.81:10:37voiced by Steven Bartlett
There is no example in nature where a more intelligent species has less control than a less intelligent one — so assuming we'll command superintelligence is arrogant.
“there's no example in nature where a more intelligent species has less control than a less intelligent species.”
Elon Musk — 1 claim
predictionconviction 0.61:05:24voiced by Steven Bartlett
AI is bringing an 'age of abundance' in which working becomes a choice rather than a necessity.
“when Elon says that working will be a choice”