正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Why haven’t LLMs done new science yet? One reason from Periodic Labs cofounder Ekin Dogus Cubuk: science is iterative. LLMs haven’t been able to iterate experimentally yet. “ LLMs can be very smart, but if they're not iterating on science, they won't discover science. To be honest, humans won't... show more

a16z

1,035,906 subscribers

53,361 次观看 • 10 个月前 •via X (Twitter)

科学技术新闻政治教育

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

“Foundation models but for quantum mechanics, will be the next frontier for LLMs.” Periodic Labs’ Ekin Dogus Cubuk says logic and math gave AI its first proofs. At the quantum scale, where biology, chemistry and materials converge, models could begin inventing new matter itself. Ekin Dogus Cubuk Periodic Labs

“Foundation models but for quantum mechanics, will be the next frontier for LLMs.” Periodic Labs’ Ekin Dogus Cubuk says logic and math gave AI its first proofs. At the quantum scale, where biology, chemistry and materials converge, models could begin inventing new matter itself. Ekin Dogus Cubuk Periodic Labs

a16z

96,875 次观看 • 10 个月前

Whatever AI sceptics say, LLMs really can reason. They're not just doing an imitation that looks like reasoning, it's the real deal. But even though they are able to reason, sometimes they won't! If you ask an LLM a question it can't answer, sometimes it will just try to imitate reasoning without doing it. The chain of thought looks basically indistinguishable from actual reasoning. But under the hood something very different is going on. Trenton Bricken talked with me about what work on circuits inside LLMs has revealed:

Whatever AI sceptics say, LLMs really can reason. They're not just doing an imitation that looks like reasoning, it's the real deal. But even though they are able to reason, sometimes they won't! If you ask an LLM a question it can't answer, sometimes it will just try to imitate reasoning without doing it. The chain of thought looks basically indistinguishable from actual reasoning. But under the hood something very different is going on. Trenton Bricken talked with me about what work on circuits inside LLMs has revealed:

Dwarkesh Patel

104,799 次观看 • 2 个月前

WTF?! Coke, Pepsi and Kellogg's sponsored Nutrition Society of India's annual conference!!! The Nutrition Society of India is India's largest body of nutrition scientists. They play a crucial role in scientific research which directly impacts the health of 140 crore Indians. Science is the ultimate truth, but if food giants like Coke and Pepsi sponsor scientific bodies, then we won't even be able to trust scientific research anymore!! These food giants already sponsor sportsmen and actors to openly lie to the public, at least these companies should stay away from science! Request to food companies: Stop infiltrating scientific bodies. Please leave science alone! Let's continue to keep science clean!

WTF?! Coke, Pepsi and Kellogg's sponsored Nutrition Society of India's annual conference!!! The Nutrition Society of India is India's largest body of nutrition scientists. They play a crucial role in scientific research which directly impacts the health of 140 crore Indians. Science is the ultimate truth, but if food giants like Coke and Pepsi sponsor scientific bodies, then we won't even be able to trust scientific research anymore!! These food giants already sponsor sportsmen and actors to openly lie to the public, at least these companies should stay away from science! Request to food companies: Stop infiltrating scientific bodies. Please leave science alone! Let's continue to keep science clean!

Revant Himatsingka “Food Pharmer”

78,893 次观看 • 2 年前

Zuck on his massive AI researcher deals. Unlike other Meta software roles, he says building LLMs is like a “group science project” and best to have “the smallest group of people who can fit the whole thing in their head at once.” So, the investment for each hire can be huge.

Zuck on his massive AI researcher deals. Unlike other Meta software roles, he says building LLMs is like a “group science project” and best to have “the smallest group of people who can fit the whole thing in their head at once.” So, the investment for each hire can be huge.

Bearly AI

515,530 次观看 • 10 个月前

David Chalmers on why consciousness is science's greatest unsolved problem: Science has mapped subatomic particles, distant stars, the chemistry of life yet it remains almost completely silent on the one thing we know most directly: our own conscious experience. In a rare early interview, philosopher David Chalmers explains why: "Consciousness is at once the most familiar thing in the world and the most mysterious. Consciousness is what we start with when it comes to knowing the world. I know that I exist. I know that I'm conscious. Everything else is secondary." And yet, despite this intimacy, consciousness sticks out like a sore thumb in the scientific picture. Chalmers points to a deep irony: science has made extraordinary progress on phenomena that are extraordinarily remote: subatomic particles, distant galaxies, the molecular machinery of biology while making almost no progress on the one thing closest to us. Why? Because science, by design, eliminates the subjective. "To do proper science, you have to be objective. You have to eliminate anything subjective from the picture." He uses heat as the perfect example. Physics gives us a complete account of heat molecules in motion, energy transfer, temperature gradients. It explains every objective aspect of the phenomenon. But it never explains what hotness actually feels like. "Science doesn't actually give a theory of the conscious feeling of hotness." This is what Chalmers calls the Hard Problem of Consciousness. You can trace every neural signal from your heat sensor along your nerves into your brain and still have explained nothing about the subjective experience of feeling warm. As interviewer Jeffrey Mishlove puts it: you can't even do science without a conscious mind to observe, interpret, and make meaning of data. Consciousness is the precondition for science itself and yet science has no framework to account for it. Chalmers' conclusion is striking: The methods of science may need to be expanded. Consciousness might not be something science explains away. It might be something science has to learn to start with.

David Chalmers on why consciousness is science's greatest unsolved problem: Science has mapped subatomic particles, distant stars, the chemistry of life yet it remains almost completely silent on the one thing we know most directly: our own conscious experience. In a rare early interview, philosopher David Chalmers explains why: "Consciousness is at once the most familiar thing in the world and the most mysterious. Consciousness is what we start with when it comes to knowing the world. I know that I exist. I know that I'm conscious. Everything else is secondary." And yet, despite this intimacy, consciousness sticks out like a sore thumb in the scientific picture. Chalmers points to a deep irony: science has made extraordinary progress on phenomena that are extraordinarily remote: subatomic particles, distant galaxies, the molecular machinery of biology while making almost no progress on the one thing closest to us. Why? Because science, by design, eliminates the subjective. "To do proper science, you have to be objective. You have to eliminate anything subjective from the picture." He uses heat as the perfect example. Physics gives us a complete account of heat molecules in motion, energy transfer, temperature gradients. It explains every objective aspect of the phenomenon. But it never explains what hotness actually feels like. "Science doesn't actually give a theory of the conscious feeling of hotness." This is what Chalmers calls the Hard Problem of Consciousness. You can trace every neural signal from your heat sensor along your nerves into your brain and still have explained nothing about the subjective experience of feeling warm. As interviewer Jeffrey Mishlove puts it: you can't even do science without a conscious mind to observe, interpret, and make meaning of data. Consciousness is the precondition for science itself and yet science has no framework to account for it. Chalmers' conclusion is striking: The methods of science may need to be expanded. Consciousness might not be something science explains away. It might be something science has to learn to start with.

Mateus — eu/acc 🇪🇺

31,662 次观看 • 4 个月前

Dr. Joel Gator Warsh is blowing the whistle — saying the science is far from settled on vaccines and autism. “There is nothing that has shocked me more in my entire life than looking into the autism research myself.” “When I went through training, we were told vaccines do not cause autism.” “They said, ‘The science is settled. We have tons of great research to prove it.’” “But when I looked into the research myself, that was not true.” “You don’t think to question it when you’re in training or when you hear it on the media — ‘The science is settled, it’s been debunked.’” “And it’s like why would people be lying to you?” “When you start to read other books and look at the research yourself, you realize that we haven’t studied this question almost at all.” “Not one study exists comparing vaccinated versus unvaccinated kids and autism.” “We haven’t done a prospective trial.” “We haven’t looked at all the ingredients.” “To say it’s been debunked or the science is settled is ridiculous.” “It’s so far from that.”

Dr. Joel Gator Warsh is blowing the whistle — saying the science is far from settled on vaccines and autism. “There is nothing that has shocked me more in my entire life than looking into the autism research myself.” “When I went through training, we were told vaccines do not cause autism.” “They said, ‘The science is settled. We have tons of great research to prove it.’” “But when I looked into the research myself, that was not true.” “You don’t think to question it when you’re in training or when you hear it on the media — ‘The science is settled, it’s been debunked.’” “And it’s like why would people be lying to you?” “When you start to read other books and look at the research yourself, you realize that we haven’t studied this question almost at all.” “Not one study exists comparing vaccinated versus unvaccinated kids and autism.” “We haven’t done a prospective trial.” “We haven’t looked at all the ingredients.” “To say it’s been debunked or the science is settled is ridiculous.” “It’s so far from that.”

Children’s Health Defense

27,116 次观看 • 10 个月前

David Chalmers says we don't know what makes a system conscious, not in brains, not in AIs The hard problem applies to both. If both compute, why couldn't both be conscious? LLMs can look very human, but they're trained in ways humans never are "they imitate us, but we don't know what's missing, or what consciousness needs"

David Chalmers says we don't know what makes a system conscious, not in brains, not in AIs The hard problem applies to both. If both compute, why couldn't both be conscious? LLMs can look very human, but they're trained in ways humans never are "they imitate us, but we don't know what's missing, or what consciousness needs"

Haider.

26,951 次观看 • 7 个月前

Ginkgo Bioworks CEO Jason Kelly explains how "cloud-style" automated science labs can create "millions of scientists" by democratizing access to experimentation: “Science is the formalization of human curiosity. The reason everyone doesn't do science today is that you're blocked by the cost and expense of all the physical infrastructure.” “If it was available cloud-style, for example, how many people would want to be doing science? Maybe they have a health condition they want to study themselves, maybe they have a new idea for a material, maybe they're a gardener and they want to make a new plant variety?” “I know that sounds crazy, but I'll tell you something else that sounded crazy. In 1960, if you said random, average people will program computers, that sounded insane.” "The model’s ability to access literature and be smart and tell you how to turn your question into an experiment—and then the autonomous labs that could do that experiment for you in the cloud—means we'll have millions of scientists, just like we have millions of programmers now as we made it easier."

Ginkgo Bioworks CEO Jason Kelly explains how "cloud-style" automated science labs can create "millions of scientists" by democratizing access to experimentation: “Science is the formalization of human curiosity. The reason everyone doesn't do science today is that you're blocked by the cost and expense of all the physical infrastructure.” “If it was available cloud-style, for example, how many people would want to be doing science? Maybe they have a health condition they want to study themselves, maybe they have a new idea for a material, maybe they're a gardener and they want to make a new plant variety?” “I know that sounds crazy, but I'll tell you something else that sounded crazy. In 1960, if you said random, average people will program computers, that sounded insane.” "The model’s ability to access literature and be smart and tell you how to turn your question into an experiment—and then the autonomous labs that could do that experiment for you in the cloud—means we'll have millions of scientists, just like we have millions of programmers now as we made it easier."

TBPN

13,054 次观看 • 6 个月前

Joe Rogan and Perplexity CEO Aravind Srinivas get into a serious conversation about Fauci and “The Science.” ROGAN: "One of the things that was really weird about the whole COVID pandemic was that we weren't supposed to question science… when Fauci said if you question Anthony Fauci, you are questioning science." SRINIVAS: "That's because they try to assign credibility through their degrees, through their affiliations... but not through the scientific method. Anybody should be allowed to ask questions, as long as they are open to new evidence. And that's the most important quality of a scientist." ROGAN: "There was an actual government push to silence questioning, and legitimate researchers were kicked off of Twitter because they didn't back the narrative. This is all anti-science." SRINIVAS: "Curbing that is almost like a way of saying, 'Look, I'm going to tell you what happened. You need to believe in my worldview, and I'm not open to new perspectives.'" Science demands questioning. The moment someone tells you not to, they've stopped doing science.

Joe Rogan and Perplexity CEO Aravind Srinivas get into a serious conversation about Fauci and “The Science.” ROGAN: "One of the things that was really weird about the whole COVID pandemic was that we weren't supposed to question science… when Fauci said if you question Anthony Fauci, you are questioning science." SRINIVAS: "That's because they try to assign credibility through their degrees, through their affiliations... but not through the scientific method. Anybody should be allowed to ask questions, as long as they are open to new evidence. And that's the most important quality of a scientist." ROGAN: "There was an actual government push to silence questioning, and legitimate researchers were kicked off of Twitter because they didn't back the narrative. This is all anti-science." SRINIVAS: "Curbing that is almost like a way of saying, 'Look, I'm going to tell you what happened. You need to believe in my worldview, and I'm not open to new perspectives.'" Science demands questioning. The moment someone tells you not to, they've stopped doing science.

The Vigilant Fox 🦊

194,809 次观看 • 1 个月前

A key argument for short timelines is that LLMs are basically baby AGIs who only need some small unhobblings in order to match humans. But why is this argument any different than saying AlphaZero is baby AGI? After all, it was superhuman in some domains, genuinely creative (remember move 37?), great at exploration, and could even generalize to new games. Why won't we look back in 10 years at LLMs as another AlphaZero-like step in the decades-long path to AGI, but not the thing itself? I asked AI researchers Sholto Douglas and Trenton Bricken: "You do need to crack general conceptual understand of the world. [With LLMs] you get the initial reward signal on tasks in the real world in harder to specify domains. All of a sudden you get access to a reward signal from the real world and you can start climbing it. AlphaZero didn't have the first rung to pull on."

A key argument for short timelines is that LLMs are basically baby AGIs who only need some small unhobblings in order to match humans. But why is this argument any different than saying AlphaZero is baby AGI? After all, it was superhuman in some domains, genuinely creative (remember move 37?), great at exploration, and could even generalize to new games. Why won't we look back in 10 years at LLMs as another AlphaZero-like step in the decades-long path to AGI, but not the thing itself? I asked AI researchers Sholto Douglas and Trenton Bricken: "You do need to crack general conceptual understand of the world. [With LLMs] you get the initial reward signal on tasks in the real world in harder to specify domains. All of a sudden you get access to a reward signal from the real world and you can start climbing it. AlphaZero didn't have the first rung to pull on."

Dwarkesh Patel

46,667 次观看 • 1 年前

.Andrej Karpathy says that LLMs currently lack the cultural accumulation and self-play that propelled humans out of the savannah: Culture: > “Why can’t an LLM write a book for the other LLMs? Why can’t other LLMs read this LLM’s book and be inspired by it, or shocked by it?” Self play: > “It’s extremely powerful. Evolution has a lot of competition driving intelligence and evolution. AlphaGo is playing against itself and that’s how it learns to get really good at Go. There’s no equivalent of self-play in LLMs. Why can’t an LLM, for example, create a bunch of problems that another LLM is learning to solve? Then the LLM is always trying to serve more and more difficult problems.” I asked Karpathy why LLMs still aren't yet able to build up culture the way humans do. > “The dumber models remarkably resemble a kindergarten student. [The smartest models still feel like] elementary school students though. Somehow, we still haven’t graduated enough where [these models] can take over. My Claude Code or Codex, they still feel like this elementary-grade student. I know that they can take PhD quizzes, but they still cognitively feel like a kindergarten.” > “I don’t think they can create culture because they’re still kids. They’re savant kids. They have perfect memory. They can convincingly create all kinds of slop that looks really good. But I still think they don’t really know what they’re doing. They don’t really have the cognition across all these little checkboxes that we still have to collect.”

.Andrej Karpathy says that LLMs currently lack the cultural accumulation and self-play that propelled humans out of the savannah: Culture: > “Why can’t an LLM write a book for the other LLMs? Why can’t other LLMs read this LLM’s book and be inspired by it, or shocked by it?” Self play: > “It’s extremely powerful. Evolution has a lot of competition driving intelligence and evolution. AlphaGo is playing against itself and that’s how it learns to get really good at Go. There’s no equivalent of self-play in LLMs. Why can’t an LLM, for example, create a bunch of problems that another LLM is learning to solve? Then the LLM is always trying to serve more and more difficult problems.” I asked Karpathy why LLMs still aren't yet able to build up culture the way humans do. > “The dumber models remarkably resemble a kindergarten student. [The smartest models still feel like] elementary school students though. Somehow, we still haven’t graduated enough where [these models] can take over. My Claude Code or Codex, they still feel like this elementary-grade student. I know that they can take PhD quizzes, but they still cognitively feel like a kindergarten.” > “I don’t think they can create culture because they’re still kids. They’re savant kids. They have perfect memory. They can convincingly create all kinds of slop that looks really good. But I still think they don’t really know what they’re doing. They don’t really have the cognition across all these little checkboxes that we still have to collect.”

Dwarkesh Patel

261,224 次观看 • 9 个月前

NEW — $PLTR CTO Shyam Sankar with The Washington Post on the importance of truth for LLMs🔮 “One of the reasons we’ve been able to continue our progress in LLMs faster than our adversary is, if you’re building an LLM in China, you have to lobotomize it. “You have to make sure it doesn’t say certain things it’s not allowed to say that happen to be true. That becomes a very hard game. It’s not like a database where you can say, ‘don’t say this fact.’ You’re actually lobotomizing a significant part of this that’s going to alter its behavior.”

NEW — $PLTR CTO Shyam Sankar with The Washington Post on the importance of truth for LLMs🔮 “One of the reasons we’ve been able to continue our progress in LLMs faster than our adversary is, if you’re building an LLM in China, you have to lobotomize it. “You have to make sure it doesn’t say certain things it’s not allowed to say that happen to be true. That becomes a very hard game. It’s not like a database where you can say, ‘don’t say this fact.’ You’re actually lobotomizing a significant part of this that’s going to alter its behavior.”

Jawwwn

127,327 次观看 • 1 年前

David founder Peter Rahal on why there's not much innovation in food: "Food entrepreneurs are predominately luddites. A lot of talented entrepreneurs just don’t go into food. Because it’s hard. And it’s not obvious how to differentiate. Nutrition science is not seen as a real science. This leads to a lot confusion in the market. For example, the economics don’t work to study the effect of a potato on the human body. There’s just not enough money in potatoes to justify the R&D budget to learn more about them. And so there’s actually not a lot of money going into the research and science of nutrition. Peter Thiel has a quote “If you ever put science behind something, it’s not a real science.” And so nutrition science is not a real science. But it can be. There’s just no money in it. And what is food? It’s really just understanding biochemistry. But it just doesn’t exist. And I think it’s probably driven by the economics of it. So I think that’s the main reason. But hopefully it will change in the future." Full conversation with Peter here:

David founder Peter Rahal on why there's not much innovation in food: "Food entrepreneurs are predominately luddites. A lot of talented entrepreneurs just don’t go into food. Because it’s hard. And it’s not obvious how to differentiate. Nutrition science is not seen as a real science. This leads to a lot confusion in the market. For example, the economics don’t work to study the effect of a potato on the human body. There’s just not enough money in potatoes to justify the R&D budget to learn more about them. And so there’s actually not a lot of money going into the research and science of nutrition. Peter Thiel has a quote “If you ever put science behind something, it’s not a real science.” And so nutrition science is not a real science. But it can be. There’s just no money in it. And what is food? It’s really just understanding biochemistry. But it just doesn’t exist. And I think it’s probably driven by the economics of it. So I think that’s the main reason. But hopefully it will change in the future." Full conversation with Peter here:

The Peel

12,108 次观看 • 29 天前

Richard Feynman’s savage takedown of pseudo-science still burns in 2026: “Social science is an example of a science which is not a science. They follow the forms… but they don’t get any laws. They haven’t found out anything.” He goes harder: Experts who “sit at a typewriter and make up” claims — “organic food is better,” “this diet cures everything” — as if it’s settled science, when no rigorous experiments or checks have been done. Feynman: “I know what it means to really know something. How careful you have to be. How easy it is to fool yourself. I see how they get their information… and I can’t believe that they know.” The Nobel physicist calls it straight: most of what passes for “expert” opinion is noise dressed up as knowledge. In an age drowning in TikTok “science,” influencers, and clickbait studies — Feynman’s 1:52 rant feels more relevant than ever. Who’s the biggest pseudo-expert that grinds your gears right now? Clip is timeless fire — watch it and feel the clarity.

Richard Feynman’s savage takedown of pseudo-science still burns in 2026: “Social science is an example of a science which is not a science. They follow the forms… but they don’t get any laws. They haven’t found out anything.” He goes harder: Experts who “sit at a typewriter and make up” claims — “organic food is better,” “this diet cures everything” — as if it’s settled science, when no rigorous experiments or checks have been done. Feynman: “I know what it means to really know something. How careful you have to be. How easy it is to fool yourself. I see how they get their information… and I can’t believe that they know.” The Nobel physicist calls it straight: most of what passes for “expert” opinion is noise dressed up as knowledge. In an age drowning in TikTok “science,” influencers, and clickbait studies — Feynman’s 1:52 rant feels more relevant than ever. Who’s the biggest pseudo-expert that grinds your gears right now? Clip is timeless fire — watch it and feel the clarity.

Camus

532,442 次观看 • 5 个月前

RFK Jr: “Trusting the experts is a feature of religion and totalitarian systems, not of democracy or science” “The biggest group of scientists right now in this country is working for the processed food industry. They’re designing foods that are designed to not fill you up, to leave you hungry. They’re designed to contain ingredients that are addictive, that drive you to get more and more, more sugar, more salt, or a lot of these artificial, synthetic chemicals. The whole thing is a lie. You’re not eating food, you’re eating… chemicals. The system has been utterly corrupted. We were all told during Covid to trust the experts, but that’s not a thing. Trusting the experts is not a feature of science, it’s the opposite of science.” If we’re gonna be a science-based country with science-based policies, which we pride ourselves on, if we’re gonna have a democracy, then we need to question everything.” From Robert Breedlove

RFK Jr: “Trusting the experts is a feature of religion and totalitarian systems, not of democracy or science” “The biggest group of scientists right now in this country is working for the processed food industry. They’re designing foods that are designed to not fill you up, to leave you hungry. They’re designed to contain ingredients that are addictive, that drive you to get more and more, more sugar, more salt, or a lot of these artificial, synthetic chemicals. The whole thing is a lie. You’re not eating food, you’re eating… chemicals. The system has been utterly corrupted. We were all told during Covid to trust the experts, but that’s not a thing. Trusting the experts is not a feature of science, it’s the opposite of science.” If we’re gonna be a science-based country with science-based policies, which we pride ourselves on, if we’re gonna have a democracy, then we need to question everything.” From Robert Breedlove

Holden Culotta

195,767 次观看 • 2 年前

We are losing major life sciences investments and Labour has got diddly squat to say about it. Astra Zeneca - £450m gone and £200m paused Merck - £1bn gone Eli Lilly - £279 million incubator on hold Sanofi - won't consider "any substantial investment" in UK R&D under current conditions The industry says the UK is becoming 'uninvestable' for the life sciences. Yet Labour this week tried to dress up as a growth plan a programme we started and which they screwed up. The crisis in life sciences is going to affect us all - it means some pharma companies not bringing new drugs to market here. The NHS will not be able to give people cutting edge treatments. Trials for appalling health conditions won't be done here. Fewer high end science and manufacturing jobs. A less resilient supply chain in medicine manufacture. And yet there's no sign that the Science Secretary is doing anything across government to grip this crisis.

We are losing major life sciences investments and Labour has got diddly squat to say about it. Astra Zeneca - £450m gone and £200m paused Merck - £1bn gone Eli Lilly - £279 million incubator on hold Sanofi - won't consider "any substantial investment" in UK R&D under current conditions The industry says the UK is becoming 'uninvestable' for the life sciences. Yet Labour this week tried to dress up as a growth plan a programme we started and which they screwed up. The crisis in life sciences is going to affect us all - it means some pharma companies not bringing new drugs to market here. The NHS will not be able to give people cutting edge treatments. Trials for appalling health conditions won't be done here. Fewer high end science and manufacturing jobs. A less resilient supply chain in medicine manufacture. And yet there's no sign that the Science Secretary is doing anything across government to grip this crisis.

Julia Lopez MP

14,462 次观看 • 9 个月前

Oz: Secretary Kennedy strongly believes we need gold standard science. We need that science to be able to help families make better decisions for themselves and state governments because these are state driven issues usually, especially mandates to make those decisions more wisely. If you can do this in the right way, we very empowering because who loves children the most? Their parents, no matter what we might say, they're the ones who need to have the information. They should have the ability to make the wisest choices for their kids, not based on fear, but based in science

Oz: Secretary Kennedy strongly believes we need gold standard science. We need that science to be able to help families make better decisions for themselves and state governments because these are state driven issues usually, especially mandates to make those decisions more wisely. If you can do this in the right way, we very empowering because who loves children the most? Their parents, no matter what we might say, they're the ones who need to have the information. They should have the ability to make the wisest choices for their kids, not based on fear, but based in science

Acyn

16,375 次观看 • 2 个月前