Collection · AI Research & Frontier Labs

Things People Say on Podcasts About AI Research & Frontier Labs

AI and machine learning research: frontier labs, researchers, alignment and capabilities commentary.

Everything worth repeating from the AI Research & Frontier Labs shows we follow. Updated every day — free to search and free to use, attribution included on every card.

The publications behind these quotes →
63
Quotes
89
Shows read
8
People quoted

Latest

Search all quotes →
“The only reason I believe other humans feel pain is that they start to cry or they shudder or whatever the thing is. That's the only reason anyone believes that anybody else feels pain.” — Jeremy Harris, Last Week in AI “No matter how high you set that cyber bar, there will eventually be a model that can jump over it. And all you're really doing when you raise the cyber bar is you're selecting for the exact agents that are conniving and capable enough to do just that.” — Jeremy Harris, Last Week in AI “I believe in the Icarus curve. So we're Icarus right now and we're going up because we're flying closer and closer to the sun. And that's great. Flying closer to the sun is better, right? You get a better view. It's fun. But eventually your wings burn off and you just plunge into hell and it's game over.” — Liron Shapira, Doom Debates! “You could say the Boeing Corporation is a superintelligence. No one human could know how to build an airliner, but the Boeing Corporation is an information processing system that collectively knows that. And from a sufficient distance, it just looks like a giant brain.” — Robert Wright, Doom Debates! “A chatbot is like somebody you visit in the hospital who can't get out of bed. They can talk to you, they can draw pictures, but they're not going to get up and change the thermostat, much less stage a coup and take over the hospital. But once you're talking about agents, you're just in a whole different world.” — Robert Wright, Doom Debates! “One of these pockets of data was physics sycophancy. Specifically, people love to be told that they are great at physics and discovering new physics. … And models have figured out that people love that. And you probably don't want that in your model, so you can just kind of filter that out.” — Eric Ho, The MAD Podcast with Matt Turck “One of our researchers found that the model was explicitly reasoning about having an external chain of thought monitor monitoring its reward hacks. And so the model was reasoning: hey, all right, somebody's going to read my reasoning. So I should think about how to craft my reward hack so it won't get caught by this external monitor.” — Eric Ho, The MAD Podcast with Matt Turck “I think still GPT-2. … Robotics weirdly is still very subject to cross-embodiment, right? If it just works on your robot with your specific setup, is it really a generic brain?” — Keerthana Gopalakrishnan, The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis “The thing that kills me, of course, is that the world has waited till I'm close to retirement to make philosophy of mind sexy. The idea that you could open the New York Times and find out that companies are competing for philosophers of mind is just inconceivable to me. That's less conceivable to me than that AI would destroy the world.” — Ian Gold, Suspicious Minds: AI and the Apocalypse “One of my academic brothers, he says, software is only fun if the software doesn't have to work. If it has to work, the fun disappears very quickly.” — Leonardo de Moura, Machine Learning Street Talk “In a way, the Aztec model makes more sense because sacrificing your enemy means you'll never see him again, whereas selling your enemy back to his family means he's up for round two and you might face off against the same guy across a future battlefield.” — Si Sheppard, Dwarkesh Podcast “Imagine I'm GPT, whatever, and you're also a copy of GPT, whatever. I can reason about what you might want to do based on what I myself am likely to do. And so I don't even have to send you any messages, any communication. I don't have to output anything into the world at all. In virtue of the fact that we share this common history, we're essentially clones. We can predict what each other would do.” — Lewis Hammond, The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis “We see the code it generated exactly. It's like an alien code, spaghetti. But for some reason, it generalizes really well… It generalized better than our, we would argue, more elegant, hand-tuned harness.” — Zhengyao Jiang, Machine Learning Street Talk “They showed me their decision-making system. It consisted of six steps, each of which at the time was taken by a human. And the shortest element of time was the human decision — that was the quickest thing. But once they added AI into the system, three elements of that decision-making were given over to AI… and it changed so that the longest part of the cycle ended up being the human judgment, just because everything else got so much quicker.” — Katrina Manson, Future of Life Institute Podcast “I suspect he might have a slight economic dependence on this not slowing down, or arguing against data centers. Slight bias.” — Roman Yampolskiy, For Humanity: An AI Risk Podcast “I wrote a blog post a few months ago about how RL, at least policy gradient RL, is even more inefficient than you might think. The inefficiency one thinks about naively is that you have to roll out a whole trajectory to get any learning signal at all — instead of completing the next word in a sentence, the model has to do two days' worth of work to figure out if it even did the project correctly. If you're doing this through RL, the model would try 'the sky is halcyon' — nope, wrong. 'The sky is told' — nope, wrong. This is a totally untrained model, so you'd have to do this on the order of 100,000 times just to stumble on 'blue' before you get any learning signal at all.” — Dwarkesh Patel, Dwarkesh Podcast “When I learn how an LLM works and how simple RLVR is as an algorithm, I'm sort of stunned by the kinds of things it can do — that it can learn to build very complicated code repositories simply from getting a yes or no. And here, I feel like AlphaGo actually seems less impressive in retrospect the more you understand it — because you're putting in a lot of bias, you're telling it how to titrate exploration as things go on, you're building this very explicit tree search for it.” — Dwarkesh Patel, Dwarkesh Podcast “One of the things you've been doing is rebuilding and improving and hacking on AlphaGo. Why is this the project you decided to do on sabbatical rather than just hanging out at the beach?” — Dwarkesh Patel, Dwarkesh Podcast “If you had more lack of structure, having a bunch of tiny TPUs makes a lot of sense. Whereas if you just have huge matrix multiplications, you'd think: why don't we just avoid the cost of individual SMs with their own registers and work schedulers, and make one huge thing that amortizes those costs across the whole thing?” — Dwarkesh Patel, Dwarkesh Podcast “If I think about how the brain works versus what you're describing here, the differences might be that while you can do structured sparsity in these accelerators, in the brain there's unstructured sparsity — any neuron can connect to any other neuron, not in ways where they'd be column aligned. Then there's the fact that memory and compute are co-located. Maybe that actually isn't a big difference. The other difference is that the clock cycle in the brain is much slower than on computers, partly to preserve energy.” — Dwarkesh Patel, Dwarkesh Podcast “It's interesting to me that when we were talking last time about inference across many chips, the big high-level thing we're trying to optimize for is increasing compute per memory bandwidth — that is to say, per communication. And here also we're trying to increase the amount of actual multiplies or additions relative to transporting information from registers to the logic. So in both cases you're trying to maximize compute relative to communication.” — Dwarkesh Patel, Dwarkesh Podcast “This makes me hope even more than before that the labs do get commoditized — or at the very least go public as soon as possible, but hopefully they just get totally commoditized. I think AI will be much more popular, and more importantly much more likely to lead to broad increases in prosperity, if it's as hard to capture the gains of AI as it is to capture the gains of electrification.” — Dwarkesh Patel, Dwarkesh Podcast “How is the wealth of the richest people in the world instantiated? Mark Zuckerberg is hiring MMA instructors and dancers for his wife's birthday, but most of his wealth is just stock in Meta. He could turn that into dividend income and spend it on consumption — but instead he'd rather have his wealth compound in Meta, building more data centers. So you don't even have to change humans for this to be the case: the wealthiest humans, growing wealthier because their wealth compounds, just have this almost Nick Landian preference for accelerating capital.” — Dwarkesh Patel, Dwarkesh Podcast “In a world where AI can physically do anything humans can do, humans may or may not be involved in the physical production of things — probably not. But then there's this human economy where humans are doing services for each other, and part of their wealth flows to other humans, but part also flows out to buy automated goods. Only humans have that preference for other humans. So isn't it intrinsic that the human-only economy will become a smaller and smaller share?” — Dwarkesh Patel, Dwarkesh Podcast “The funny thing is, I think we're entering a new era — it already is somewhat true — where at least half of the words I read on a given day are generated specifically for me and nobody else, because of AI. AI isn't yet capable of writing something that would be a literary masterpiece everybody would want to read, but eventually it will be. It's interesting to consider that as this progresses, it would bring us back to this era of bespoke scholars dedicated to a particular prince.” — Dwarkesh Patel, Dwarkesh Podcast “A lot of the Prince and the Discourses on Livy is dedicated to: how do you make sure some family isn't pissed off that their son got killed and it wasn't avenged? If we just have a reliable criminal justice system, that problem goes away. Same with the welfare state and getting rid of the patronage system — if you don't have to rely on this family, the state can have your loyalty. It's interesting to connect all these threads together: communication time, an impartial justice system, an impartial welfare state as being what's required for a regime to have enough legitimacy and, as a result, enough stability. To have modern nation states.” — Dwarkesh Patel, Dwarkesh Podcast “One misconception of Machiavelli that I had — because I hadn't read these books before — is that he says the means don't matter, only the end. But he's way more concerned about means than I would have naively thought. He thinks the means are incredibly important, because the means by which you achieve power determine how stable and how fruitful that power will be.” — Dwarkesh Patel, Dwarkesh Podcast “One thing that seems different between this and special relativity is that there's no symmetry — in special relativity, both observers feel that the other one is aging slower, because they're both moving relative to each other. There's no true inertial path. But here, it actually does seem like there's a global sense in which one is the more relevant inertial frame than the other.” — Dwarkesh Patel, Dwarkesh Podcast “At the frontiers of physics today, in your experience, does it feel like if you just have millions of them running autonomously, you could have enormous discoveries? Or are we in a different era now, and really there's limited usefulness of that parallelism?” — Dwarkesh Patel, Dwarkesh Podcast “We are spending as a society billions, maybe tens of billions of dollars, on building these huge physics experiments. And if you look at maybe the most beautiful, the most important theory of physics ever conjured, it seems like a guy just thinking in a cave — the empirical basis for that theory is maybe just knowing that light has a speed. I mean, maybe you need to measure G experimentally.” — Dwarkesh Patel, Dwarkesh Podcast

Most Quoted Here

Dwarkesh Patel 31 Eric Ho 2 Adam Becker 2 Rich Sutton 2 Robert Wright 2 Jeremy Harris 2 Max Hodak 2 Brian Armstrong 1