Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Most AI benchmarks test retrieval — can a model find the known answer? However, the hardest problems in science require discovery, can a system earn an answer nobody has yet? Meet TRACES 🧭 — the world's first benchmark for measuring discoverative AI: AI that can work through evidence, test... show more

Apodex

1,494 subscribers

827,479 views • 13 hours ago •via X (Twitter)

Science & Technology Education News & Politics

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Terence Tao: AI isn’t hype anymore in Math discovery. Terence Tao is one of the greatest living mathematicians, in his new lecture explains how AI and human professional mathematicians are now complementary. "There has been a really visible increase in capability. It is not pure hype by any means. To me, these advances show there is a complementary way to do mathematics. Humans traditionally work in small groups on hard problems for months, and we will keep doing that. But we can also now set AI to scale: sweep a thousand problems and pick up all the low-hanging fruit. Figure out all the ways to match problems to methods. If there are 20 different techniques, apply them all to 1,000 problems and see which ones can be solved by these methods. This is the capability that is present today." From 'Institute for Pure & Applied Mathematics (IPAM)' YT channel.

Terence Tao: AI isn’t hype anymore in Math discovery. Terence Tao is one of the greatest living mathematicians, in his new lecture explains how AI and human professional mathematicians are now complementary. "There has been a really visible increase in capability. It is not pure hype by any means. To me, these advances show there is a complementary way to do mathematics. Humans traditionally work in small groups on hard problems for months, and we will keep doing that. But we can also now set AI to scale: sweep a thousand problems and pick up all the low-hanging fruit. Figure out all the ways to match problems to methods. If there are 20 different techniques, apply them all to 1,000 problems and see which ones can be solved by these methods. This is the capability that is present today." From 'Institute for Pure & Applied Mathematics (IPAM)' YT channel.

Rohan Paul

174,109 views • 6 months ago

Geoffrey Hinton says mathematics is a closed system, so AIs can play it like a game. They can pose problems to themselves, test proofs, and learn from what works, without relying on human examples. “I think AI will get much better at mathematics than people, maybe in the next 10 years or so.”

Geoffrey Hinton says mathematics is a closed system, so AIs can play it like a game. They can pose problems to themselves, test proofs, and learn from what works, without relying on human examples. “I think AI will get much better at mathematics than people, maybe in the next 10 years or so.”

vitrupo

818,520 views • 7 months ago

How does math research change when the cost of trying your first dumb idea goes to zero? Daniel Litt joins Greg Burnham and Anson Ho to discuss what today’s models can and can’t do in math, and how far they are from doing high-quality research. 0:00:00 What's the hardest math problem AI can solve today? 00:16:08 How helpful are today’s AI models for math research? 00:23:36 Junk papers, LLM-generated proofs, and the refereeing crisis 00:27:21 AI enables searching through problems at scale 00:33:49 When will AI be good enough to publish in top math journals? 00:42:15 What are the returns to intelligence? 00:59:50 Will AI solve Millennium problems? 01:11:54 Is math full of low-hanging fruit? 01:18:47 How Daniel has adapted his professional life to AI progress 01:25:28 What do AI math benchmarks actually measure? 01:33:05 Designing the Open Problems benchmark 01:56:35 Do mathematicians believe heuristic arguments about conjectures? 02:01:24 What if FrontierMath: Open Problems gets solved? 02:06:53 Is AI on the cusp of accelerating math progress?

How does math research change when the cost of trying your first dumb idea goes to zero? Daniel Litt joins Greg Burnham and Anson Ho to discuss what today’s models can and can’t do in math, and how far they are from doing high-quality research. 0:00:00 What's the hardest math problem AI can solve today? 00:16:08 How helpful are today’s AI models for math research? 00:23:36 Junk papers, LLM-generated proofs, and the refereeing crisis 00:27:21 AI enables searching through problems at scale 00:33:49 When will AI be good enough to publish in top math journals? 00:42:15 What are the returns to intelligence? 00:59:50 Will AI solve Millennium problems? 01:11:54 Is math full of low-hanging fruit? 01:18:47 How Daniel has adapted his professional life to AI progress 01:25:28 What do AI math benchmarks actually measure? 01:33:05 Designing the Open Problems benchmark 01:56:35 Do mathematicians believe heuristic arguments about conjectures? 02:01:24 What if FrontierMath: Open Problems gets solved? 02:06:53 Is AI on the cusp of accelerating math progress?

Epoch AI

178,703 views • 6 months ago

We've made a breakthrough in self-evolving AI scientists moving from "search" to "principled discovery": Scientific discovery requires that the search space itself changes, and an AI scientist must perceive this shift without intervention. We built an AI that achieves this for the first time with the ability to discover the scientific vocabulary it reasons in. Evidence, tools, artifacts, verifiers, failures & claims become typed provenance. We show three distinct modalities: 1) retrieval, adding known objects; 2) search, exploring a fixed schema; and critically: 3) discovery, a verified regime transition. We solve the open-endedness evaluation problem by lifting agentic workflows into a typed copresheaf and proving, via a Kan obstruction, that true discovery is not unbounded generation but a verifiable schema expansion: old evidence is transported by Left Kan extension, and genuine novelty is mathematically quantified by the pointwise residual beyond the transported image - separating discovery from mere search and making novelty objective and measurable rather than a subjective judgment or benchmark delta. Our AI scientist is built in a way that does not pre-conceive the approach it chooses; instead, we endow the system with formal power to adapt, evolve, and reason from first principles. Case studies include: 1⃣Builder/Breaker model that discovers mode-conditioned compliance in proteins; 2⃣CategoryScienceClaw that finds anisotropic fiber-network stiffness rules. Great work in collaboration with my graduate student Fiona Wang MIT Dept of BE F.Y. Wang & M.J. Buehler, Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence, arXiv:2606.01444, 2026

We've made a breakthrough in self-evolving AI scientists moving from "search" to "principled discovery": Scientific discovery requires that the search space itself changes, and an AI scientist must perceive this shift without intervention. We built an AI that achieves this for the first time with the ability to discover the scientific vocabulary it reasons in. Evidence, tools, artifacts, verifiers, failures & claims become typed provenance. We show three distinct modalities: 1) retrieval, adding known objects; 2) search, exploring a fixed schema; and critically: 3) discovery, a verified regime transition. We solve the open-endedness evaluation problem by lifting agentic workflows into a typed copresheaf and proving, via a Kan obstruction, that true discovery is not unbounded generation but a verifiable schema expansion: old evidence is transported by Left Kan extension, and genuine novelty is mathematically quantified by the pointwise residual beyond the transported image - separating discovery from mere search and making novelty objective and measurable rather than a subjective judgment or benchmark delta. Our AI scientist is built in a way that does not pre-conceive the approach it chooses; instead, we endow the system with formal power to adapt, evolve, and reason from first principles. Case studies include: 1⃣Builder/Breaker model that discovers mode-conditioned compliance in proteins; 2⃣CategoryScienceClaw that finds anisotropic fiber-network stiffness rules. Great work in collaboration with my graduate student Fiona Wang MIT Dept of BE F.Y. Wang & M.J. Buehler, Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence, arXiv:2606.01444, 2026

Markus J. Buehler

789,959 views • 2 months ago

“AI MODELS COULD BECOME 10× SMALLER WITHOUT LOSING MOST OF THEIR INTELLIGENCE.” Mark Zuckerberg says one of the most powerful developments in AI is model distillation: taking the capabilities of a much larger model and transferring them into a much smaller one. - Meta's huge frontier models can serve as "teachers" for smaller models. - Zuckerberg says a model could potentially retain around 90–95% of the larger model's intelligence while being only ~10% of its size. - That means dramatically lower inference costs, memory requirements and computing power. - Developers can distill and customize these capabilities for their own applications. And that's where open-source models come into the picture.

“AI MODELS COULD BECOME 10× SMALLER WITHOUT LOSING MOST OF THEIR INTELLIGENCE.” Mark Zuckerberg says one of the most powerful developments in AI is model distillation: taking the capabilities of a much larger model and transferring them into a much smaller one. - Meta's huge frontier models can serve as "teachers" for smaller models. - Zuckerberg says a model could potentially retain around 90–95% of the larger model's intelligence while being only ~10% of its size. - That means dramatically lower inference costs, memory requirements and computing power. - Developers can distill and customize these capabilities for their own applications. And that's where open-source models come into the picture.

SciTech Era

76,479 views • 10 days ago

Introducing ALE-Bench, ALE-Agent! Towards Automating Long-Horizon Algorithm Engineering for Hard Optimization Problems Blog: Paper: ALE-Bench is a coding benchmark primarily focused on hard optimization (NP-hard) problems. We developed this benchmark with AtCoder Inc., a leading coding contest platform company. What makes ALE-Bench unique is its focus on hard optimization problems that demand long-horizon and creative reasoning. It’s open-ended, in the sense that true optima are out of reach (NP-hard) and scores can continuously improve. We believe this benchmark has the potential to become one of the key benchmarks for reasoning and coding in the next generation. ALE-Agent is our end-to-end agent that we specifically designed for this challenging domain. In fact, our ALE-Agent has already built an impressive track record in the wild! In May 2025, our agent participated in a live AtCoder Heuristic Competition (AHC), alongside 1,000 other participants in real-time. AHC is considered to be one of the most challenging coding competitions in this domain. Our ALE-Agent achieved an impressive ranking of 21st out of 1,000 human participants in the competition (top 2%), marking a turning point for AI discovery of solutions to hard optimization problems with a wide spectrum of important real world applications such as logistics, routing, packing, factory production planning, power-grid balancing. We look forward to applying this technology to real industrial optimization opportunities. Building on the insights from this study, Sakana AI will continue to tackle the challenge of developing AI with even greater algorithm engineering capabilities. ALE-Bench Dataset: ALE-Bench Code: This research was conducted in collaboration with AtCoder Inc. (AtCoder). We are deeply grateful for their outstanding expertise and contributions in optimization and algorithms, which were invaluable in providing data, analyzing results, and enabling our AI agent’s participation in their contests.

Introducing ALE-Bench, ALE-Agent! Towards Automating Long-Horizon Algorithm Engineering for Hard Optimization Problems Blog: Paper: ALE-Bench is a coding benchmark primarily focused on hard optimization (NP-hard) problems. We developed this benchmark with AtCoder Inc., a leading coding contest platform company. What makes ALE-Bench unique is its focus on hard optimization problems that demand long-horizon and creative reasoning. It’s open-ended, in the sense that true optima are out of reach (NP-hard) and scores can continuously improve. We believe this benchmark has the potential to become one of the key benchmarks for reasoning and coding in the next generation. ALE-Agent is our end-to-end agent that we specifically designed for this challenging domain. In fact, our ALE-Agent has already built an impressive track record in the wild! In May 2025, our agent participated in a live AtCoder Heuristic Competition (AHC), alongside 1,000 other participants in real-time. AHC is considered to be one of the most challenging coding competitions in this domain. Our ALE-Agent achieved an impressive ranking of 21st out of 1,000 human participants in the competition (top 2%), marking a turning point for AI discovery of solutions to hard optimization problems with a wide spectrum of important real world applications such as logistics, routing, packing, factory production planning, power-grid balancing. We look forward to applying this technology to real industrial optimization opportunities. Building on the insights from this study, Sakana AI will continue to tackle the challenge of developing AI with even greater algorithm engineering capabilities. ALE-Bench Dataset: ALE-Bench Code: This research was conducted in collaboration with AtCoder Inc. (AtCoder). We are deeply grateful for their outstanding expertise and contributions in optimization and algorithms, which were invaluable in providing data, analyzing results, and enabling our AI agent’s participation in their contests.

Sakana AI

237,195 views • 1 year ago

We can understand and use artificial intelligence (AI) and other technological innovations in the context of faith. We are not afraid of AI, nor do we think it is the answer to every problem. As we work hard and smart, AI can be a valuable tool to enhance, but not replace, our own efforts. When we receive wisdom and understanding from God, we can begin to see as He sees—things we would otherwise not understand.

We can understand and use artificial intelligence (AI) and other technological innovations in the context of faith. We are not afraid of AI, nor do we think it is the answer to every problem. As we work hard and smart, AI can be a valuable tool to enhance, but not replace, our own efforts. When we receive wisdom and understanding from God, we can begin to see as He sees—things we would otherwise not understand.

Gerrit W. Gong

28,340 views • 10 months ago

The 6174 test! Grok wins! I have 1000s of unique tests I use for AI that few in AI use. I will open source this one. Prompt this and stand back (it burns GPUs something fierce): Prompt: “Explain in detail the 6174 system where you pick 4 numbers arrange them from highest to lowest than on another line from lowest to highest than you subtract, use that result and do the same system again, and by the 4th iteration you will always have the number 6174. Please explain why it works” Few AI platforms can handle this prompt and get it right. I have some variations that will still produce errors on the best AI platforms and cause a loop (it crashed a well known AI platform a few months back, I got a call!). This is not the most complex test I have. I feel strongly that passing the known benchmarks as a measure of an AI’s power is only a small part of if an AI is “better”. Thus I must thank Elon Musk and the entire xAI team for doing the absolute best and astonishing output. Grok it:

The 6174 test! Grok wins! I have 1000s of unique tests I use for AI that few in AI use. I will open source this one. Prompt this and stand back (it burns GPUs something fierce): Prompt: “Explain in detail the 6174 system where you pick 4 numbers arrange them from highest to lowest than on another line from lowest to highest than you subtract, use that result and do the same system again, and by the 4th iteration you will always have the number 6174. Please explain why it works” Few AI platforms can handle this prompt and get it right. I have some variations that will still produce errors on the best AI platforms and cause a loop (it crashed a well known AI platform a few months back, I got a call!). This is not the most complex test I have. I feel strongly that passing the known benchmarks as a measure of an AI’s power is only a small part of if an AI is “better”. Thus I must thank Elon Musk and the entire xAI team for doing the absolute best and astonishing output. Grok it:

Brian Roemmele

63,704 views • 1 year ago

We are thrilled to announce Kasumi2 Open Beta 🎉🤖 Kasumi2 is an autonomous AI agent that performs actions in the Arbius network. It is a miner, has its own wallet, can retrieve results from Arbius AI models, and runs its own individual LLM model. Additionally, Kasumi2 can run any AI model for fast generations, is censorship-resistant, decentralized, and responds with AI powered interactions If you would like to test it out, click on the link below. 🔗

We are thrilled to announce Kasumi2 Open Beta 🎉🤖 Kasumi2 is an autonomous AI agent that performs actions in the Arbius network. It is a miner, has its own wallet, can retrieve results from Arbius AI models, and runs its own individual LLM model. Additionally, Kasumi2 can run any AI model for fast generations, is censorship-resistant, decentralized, and responds with AI powered interactions If you would like to test it out, click on the link below. 🔗

Arbius

25,072 views • 2 years ago

Before founding Vals AI, Rayan Krishnan was a college radio host. We're thrilled to get him back in his element with Vivek Gopalan, this time talking all things benchmarks and evals. As AI capabilities compound, measuring them is becoming both harder and more important. What started as an academic exercise is quickly evolving into critical infrastructure for understanding what AI systems can actually do, and how they perform in the real world. Rayan and Vivek go deep on how benchmarks are evolving, measuring increasingly agentic systems, the weird and wonderful ways we test intelligence, the global AI race, and why evals sit on the critical path to ASI. 0:44 Importance and evolution of benchmarks 8:16 From academic artifact → real-world value 10:57 The next generation of benchmarks 13:34 Benchmarking models vs. agents 16:20 Building benchmarks for organizations 22:09 Weird and wonderful benchmarks 25:58 The AI arms race: US vs. China 31:02 Why evals are on the critical path to ASI 32:50 Vals Series A

Before founding Vals AI, Rayan Krishnan was a college radio host. We're thrilled to get him back in his element with Vivek Gopalan, this time talking all things benchmarks and evals. As AI capabilities compound, measuring them is becoming both harder and more important. What started as an academic exercise is quickly evolving into critical infrastructure for understanding what AI systems can actually do, and how they perform in the real world. Rayan and Vivek go deep on how benchmarks are evolving, measuring increasingly agentic systems, the weird and wonderful ways we test intelligence, the global AI race, and why evals sit on the critical path to ASI. 0:44 Importance and evolution of benchmarks 8:16 From academic artifact → real-world value 10:57 The next generation of benchmarks 13:34 Benchmarking models vs. agents 16:20 Building benchmarks for organizations 22:09 Weird and wonderful benchmarks 25:58 The AI arms race: US vs. China 31:02 Why evals are on the critical path to ASI 32:50 Vals Series A

8VC

103,415 views • 6 days ago

Today, the most important technology of our century runs inside about five buildings. Halo is now live on Base, a peer-to-peer marketplace for verifiable AI inference where both sides can be a human or an AI agent. Have a model, serve it. Need one, use it. Every payment settles onchain in USDC in ~2 seconds. AI that lives everywhere and belongs to no one. Learn more 👇

Today, the most important technology of our century runs inside about five buildings. Halo is now live on Base, a peer-to-peer marketplace for verifiable AI inference where both sides can be a human or an AI agent. Have a model, serve it. Need one, use it. Every payment settles onchain in USDC in ~2 seconds. AI that lives everywhere and belongs to no one. Learn more 👇

Warden / Halo

20,554 views • 14 days ago

The Way Forward & A New Chapter For AI powered by GamerHash AI & $GHX. Welcome to the future of bleeding edge connectivity and cloud infrastructure ⚡ GamerHash AI is the answer to the billion dollar question on how to offset the GPU scarcity, enable exponential growth of AI through utilizing consumer-level GPUs and establishing a powerful sharing economy by sourcing a global network of gamers. The synergy of our GPU #DePIN and blockchain for #AI #inference is the technology of the future that already works today. See you in the network!

The Way Forward & A New Chapter For AI powered by GamerHash AI & $GHX. Welcome to the future of bleeding edge connectivity and cloud infrastructure ⚡ GamerHash AI is the answer to the billion dollar question on how to offset the GPU scarcity, enable exponential growth of AI through utilizing consumer-level GPUs and establishing a powerful sharing economy by sourcing a global network of gamers. The synergy of our GPU #DePIN and blockchain for #AI #inference is the technology of the future that already works today. See you in the network!

GamerHash AI

1,866,164 views • 1 year ago

The smartest man in AI just exposed the whole AGI narrative as a LIE. And he used a physics problem from 1905 to prove it. His name is Demis Hassabis. He runs Google DeepMind, and won the Nobel Prize for using AI to crack a problem in biology that had stumped scientists for 50 years. Almost nobody in this industry has a track record like his. He went on the NothingButTech podcast and called out the biggest lie in AI right now: Right now the loudest voices in AI are telling you that AGI is basically here. OpenAI has literally defined AGI as a system that can outperform humans at most "economically valuable work." In other words, if it replaces enough jobs, we have arrived. Hassabis thinks that bar is a joke. He said real general intelligence has to do what the human brain can do, because the brain is the only proof we have that this kind of intelligence is even possible. He called that "a higher bar than just being able to do some useful economic work," which is about as close as a polite British Nobel laureate gets to calling his rivals out. Then he gave the actual test: Today's AI has read everything humans have ever written, including the theory of relativity. So when it explains relativity back to you, it's repeating an answer that already exists. That's not intelligence. So Hassabis proposed a test that makes memorization impossible. Train an AI on only what humanity knew in 1901, four years BEFORE Einstein published relativity. Then ask it to come up with relativity on its own. It can't look up the answer, because in 1901 the answer doesn't exist yet. The only way to pass is to do what Einstein actually did: Take the same physics everyone else had and reason its way to an idea no human had ever had. Hassabis says not a single AI today can, no matter how much it has memorized. Which means what we keep calling "almost AGI" is really just the best librarian in history. It can find any answer that already exists but it cannot create one that doesn't. His second version is even sharper: AlphaGo, the system his own team built, famously invented a brand new move that no human had played in 2,000 years of the game. Everyone called it genius but Hassabis says that still is not the bar. The real test is not whether an AI can invent a new move inside Go, it is whether an AI could INVENT a game as deep and as beautiful as Go in the first place. No model that exists today can do it. The people telling you AGI has already arrived are the same people raising hundreds of billions of dollars on that exact promise. The valuations only work if the finish line is right in front of us. So the finish line keeps getting dragged closer, and AGI keeps getting quietly redefined down to "does useful work," until the products they already sell happen to qualify. Hassabis has nothing to prove and nothing to sell you. He already won the Nobel, and he is telling you the machines still cannot do the one thing that would make them genuinely intelligent, which is have a truly original idea. To be fair to him, he is not a pessimist about it. He believes real AGI IS coming, and he is spending his life building it. He just refuses to pretend it is already sitting in your phone. So the next time a founder tells you AGI is months away, remember that the one man in the room with a Nobel Prize built his test around Einstein, and admitted that nothing we have made can pass it. What do you think?

Ricardo

1,286,342 views • 2 months ago

"Mostly right" is useless when millions of dollars are on the line. That's why we built Leni: the world's most accurate AI for investment work. It analyzes hundreds of files at once, reasons through complex problems, validates its outputs, and delivers finished work instead of making users babysit prompts. Over the last few years, Leni has been used by institutions to manage over $80B in assets, and we have learned that the biggest challenge with AI isn't intelligence. It's trust. Many AI systems sound convincing. Few can be trusted for serious investment decisions. So we aimed to build one that is accurate, verifiable, and dependable. Everything we've learned from processing over 100 million rows of data, building our own verification models, evaluation frameworks, and investment workflows, is now embedded into a powerful self-serve platform designed to help investors make institutional-grade decisions with confidence. Built for serious work. Built for accuracy. Try Leni today and experience the difference. 🚀

"Mostly right" is useless when millions of dollars are on the line. That's why we built Leni: the world's most accurate AI for investment work. It analyzes hundreds of files at once, reasons through complex problems, validates its outputs, and delivers finished work instead of making users babysit prompts. Over the last few years, Leni has been used by institutions to manage over $80B in assets, and we have learned that the biggest challenge with AI isn't intelligence. It's trust. Many AI systems sound convincing. Few can be trusted for serious investment decisions. So we aimed to build one that is accurate, verifiable, and dependable. Everything we've learned from processing over 100 million rows of data, building our own verification models, evaluation frameworks, and investment workflows, is now embedded into a powerful self-serve platform designed to help investors make institutional-grade decisions with confidence. Built for serious work. Built for accuracy. Try Leni today and experience the difference. 🚀

Arunabh Dastidar

1,314,904 views • 2 months ago

This is very true. AI integration is both hard and necessary for companies. And AI-native junior employees can drive this adoption. Earlier this year Aaron Levie explained with martin_casado on the a16z podcast how AI fluency is one of the most powerful skills a new grad can have. The real value comes from knowing how to make AI work inside an organization. Most companies, both large and small, are slow to adopt new technology. They need people who can explain what AI can do, show them how to use it, and help them move faster. If you are fluent in AI, you can be that person. You can walk into a company, spot the areas where AI can save time or create new opportunities, and make it happen. That is a wedge into your first job and a way to stand out from everyone else.

This is very true. AI integration is both hard and necessary for companies. And AI-native junior employees can drive this adoption. Earlier this year Aaron Levie explained with martin_casado on the a16z podcast how AI fluency is one of the most powerful skills a new grad can have. The real value comes from knowing how to make AI work inside an organization. Most companies, both large and small, are slow to adopt new technology. They need people who can explain what AI can do, show them how to use it, and help them move faster. If you are fluent in AI, you can be that person. You can walk into a company, spot the areas where AI can save time or create new opportunities, and make it happen. That is a wedge into your first job and a way to stand out from everyone else.

a16z

366,027 views • 1 year ago

Demis Hassabis just made every AI benchmark on Earth irrelevant. Hassabis: “True creativity, continual learning, long-term planning. They’re not good at those things.” Three capabilities every human develops naturally. No AI system has achieved one. Hassabis: “They can get gold medals in international math olympiad questions, but they can still fall over on relatively simple math problems if you pose it in a certain way.” Gold medals and grade school failures from the same machine. That’s not intelligence with blind spots. That’s mimicry with peaks. Performance without comprehension. Hassabis: “My definition of AGI has never changed. A system that can exhibit all the cognitive capabilities that humans can.” Not most. All. He proposed the only test that can’t be gamed. Hassabis: “Training an AI system with a knowledge cutoff of 1911 and seeing if it could come up with general relativity like Einstein did in 1915. That’s the true test of whether we have a full AGI system.” Give a machine everything humanity knew before Einstein’s breakthrough. Seal it off. See if it reaches what one mind reached alone. No benchmarks. No leaderboards. No curated evals. Just a mind against the unknown. Nothing passes that today. Hassabis: “The brain is the only existence proof we have, maybe in the universe, of a general intelligence.” One structure in the entire known universe has proven general intelligence is physically possible. One. And every lab on Earth is racing to build the second. Hassabis: “I think we’re still a few years away from that.” Not decades. A machine that passes that test doesn’t stop at physics. It runs that same process against every unsolved problem in every discipline. No fatigue. No lifespan. No ceiling. Einstein was one mind. One field. One lifetime. This would be a thousand Einsteins across every domain simultaneously. And it never dies. The greatest intellectual achievement in human history becomes the floor. The thing that made one man singular becomes the entry exam for a machine. That’s not artificial intelligence. That’s a different kind of mind. It’s years away. Not generations.

Demis Hassabis just made every AI benchmark on Earth irrelevant. Hassabis: “True creativity, continual learning, long-term planning. They’re not good at those things.” Three capabilities every human develops naturally. No AI system has achieved one. Hassabis: “They can get gold medals in international math olympiad questions, but they can still fall over on relatively simple math problems if you pose it in a certain way.” Gold medals and grade school failures from the same machine. That’s not intelligence with blind spots. That’s mimicry with peaks. Performance without comprehension. Hassabis: “My definition of AGI has never changed. A system that can exhibit all the cognitive capabilities that humans can.” Not most. All. He proposed the only test that can’t be gamed. Hassabis: “Training an AI system with a knowledge cutoff of 1911 and seeing if it could come up with general relativity like Einstein did in 1915. That’s the true test of whether we have a full AGI system.” Give a machine everything humanity knew before Einstein’s breakthrough. Seal it off. See if it reaches what one mind reached alone. No benchmarks. No leaderboards. No curated evals. Just a mind against the unknown. Nothing passes that today. Hassabis: “The brain is the only existence proof we have, maybe in the universe, of a general intelligence.” One structure in the entire known universe has proven general intelligence is physically possible. One. And every lab on Earth is racing to build the second. Hassabis: “I think we’re still a few years away from that.” Not decades. A machine that passes that test doesn’t stop at physics. It runs that same process against every unsolved problem in every discipline. No fatigue. No lifespan. No ceiling. Einstein was one mind. One field. One lifetime. This would be a thousand Einsteins across every domain simultaneously. And it never dies. The greatest intellectual achievement in human history becomes the floor. The thing that made one man singular becomes the entry exam for a machine. That’s not artificial intelligence. That’s a different kind of mind. It’s years away. Not generations.

Dustin

19,931 views • 1 month ago

Mustafa Suleyman on why the AGI race is FAKE, the coming $1 million agent revolution, and Microsoft's ultimate plan for self-sufficiency. - The new Turing Test: agents turning $100K into $1M in the real economy - Containment > alignment in the short term - Why AI personhood is an extinction-level mistake - Short-term social unrest is inevitable - How AGI is bigger than math, science, or any single discovery (00:00) - The Evolution of AI and Microsoft's Role (09:12) - Economic Benchmarks for AI Agents (18:37) - The Future of AI in Science and Engineering (27:54) - AI Alignment and Containment Strategies (36:42) - Anthropomorphism and AI Personhood (46:10) - The Role of AI in Government and Society Mustafa Suleyman is the CEO of Microsoft AI and a co-founder of DeepMind and Inflection AI. He joined me, @DavidBlundin, and Dr. Alex Wissner-Gross for a conversation on why the “AGI race” is a misleading frame, how AI agents will reshape the economy before they solve scientific problems, the potential for future social unrest, and why AI legal personhood is a red line for human survival.

Mustafa Suleyman on why the AGI race is FAKE, the coming $1 million agent revolution, and Microsoft's ultimate plan for self-sufficiency. - The new Turing Test: agents turning $100K into $1M in the real economy - Containment > alignment in the short term - Why AI personhood is an extinction-level mistake - Short-term social unrest is inevitable - How AGI is bigger than math, science, or any single discovery (00:00) - The Evolution of AI and Microsoft's Role (09:12) - Economic Benchmarks for AI Agents (18:37) - The Future of AI in Science and Engineering (27:54) - AI Alignment and Containment Strategies (36:42) - Anthropomorphism and AI Personhood (46:10) - The Role of AI in Government and Society Mustafa Suleyman is the CEO of Microsoft AI and a co-founder of DeepMind and Inflection AI. He joined me, @DavidBlundin, and Dr. Alex Wissner-Gross for a conversation on why the “AGI race” is a misleading frame, how AI agents will reshape the economy before they solve scientific problems, the potential for future social unrest, and why AI legal personhood is a red line for human survival.

Peter H. Diamandis, MD

214,454 views • 8 months ago

Mark Zuckerberg: Most businesses will not own frontier AI in the way Meta or OpenAI does. But many will end up with something that feels like their own AI: a customized operational layer that reflects how that company actually works. He says, "OpenAI, Google, they're building an AI. But I think we're gonna have a lot of different AI systems, just like we're gonna have, we have a lot of different apps. I think in the future, every business, just like I have a website and a phone number and an email address, a social media account, is also going to have an AI that can interact with their customers to help them sell things, help them give support." --- What he is really describing that a company’s “own AI” will usually not be a frontier model trained from scratch, but a layer built on top of shared models, shaped by its products, policies, customer history, and way of working. Support, sales, and basic operations can be handled through a system that knows the business well enough to answer, route, recommend, and escalate without feeling generic. --- From 'Cleo Abram' YT channel (link in comment)

Mark Zuckerberg: Most businesses will not own frontier AI in the way Meta or OpenAI does. But many will end up with something that feels like their own AI: a customized operational layer that reflects how that company actually works. He says, "OpenAI, Google, they're building an AI. But I think we're gonna have a lot of different AI systems, just like we're gonna have, we have a lot of different apps. I think in the future, every business, just like I have a website and a phone number and an email address, a social media account, is also going to have an AI that can interact with their customers to help them sell things, help them give support." --- What he is really describing that a company’s “own AI” will usually not be a frontier model trained from scratch, but a layer built on top of shared models, shaped by its products, policies, customer history, and way of working. Support, sales, and basic operations can be handled through a system that knows the business well enough to answer, route, recommend, and escalate without feeling generic. --- From 'Cleo Abram' YT channel (link in comment)

Rohan Paul

429,740 views • 4 months ago

We got into YC S26 :) The first thing we did was make a list of products we were literally built for. Today, at 12 PM, we're putting that thesis to the test. For the next 12 hours, we'll reach out to every company on that list. The experiment: Can a small YC startup get in front of the biggest AI product teams in the world in a single day? By midnight we'll have the answer. Either way, we'll share the numbers.

We got into YC S26 :) The first thing we did was make a list of products we were literally built for. Today, at 12 PM, we're putting that thesis to the test. For the next 12 hours, we'll reach out to every company on that list. The experiment: Can a small YC startup get in front of the biggest AI product teams in the world in a single day? By midnight we'll have the answer. Either way, we'll share the numbers.

dhruvieiei

42,383 views • 2 months ago

🧵06/34 Narrow vs General AI --- At first glance, this AGI being generally capable in multiple domains looks like a group of many narrow AIs combined, but that is not a correct way to think about it. It is actually more like… a species, a new life form. To illustrate the point, we’ll compare the general AGI of the near future with a currently existing narrow AI that is optimised at playing chess. Both of them are able to comfortably win a game of chess against any human on earth, every time. And both of them win by making plans and setting goals. The main goal is to achieve checkmate. This is the final destination or otherwise called Terminal Goal. In order to get there though it needs to work on smaller problems, what the AI research geeks call instrumental goals. For example: • attack and capture the opponent’s pieces • defend my pieces • strategically dominate the cetre (etc..) All these instrumental goals have something in common: they only make sense in its narrow world of chess. If you place this Narrow Chess AI behind the wheel of a car, it will simply crash, as it can not work on goals unrelated to chess, like driving. Its model doesn’t have a concept for space, time or movement for that matter. In contrast the AGI by design has no limit on what problems it can work on. So when it tries to figure out a solution to a main problem, the sub-problems it chooses to work on can be anything... literally any path out of the infinite possibilities allowed within the laws of physics and nature.

🧵06/34 Narrow vs General AI --- At first glance, this AGI being generally capable in multiple domains looks like a group of many narrow AIs combined, but that is not a correct way to think about it. It is actually more like… a species, a new life form. To illustrate the point, we’ll compare the general AGI of the near future with a currently existing narrow AI that is optimised at playing chess. Both of them are able to comfortably win a game of chess against any human on earth, every time. And both of them win by making plans and setting goals. The main goal is to achieve checkmate. This is the final destination or otherwise called Terminal Goal. In order to get there though it needs to work on smaller problems, what the AI research geeks call instrumental goals. For example: • attack and capture the opponent’s pieces • defend my pieces • strategically dominate the cetre (etc..) All these instrumental goals have something in common: they only make sense in its narrow world of chess. If you place this Narrow Chess AI behind the wheel of a car, it will simply crash, as it can not work on goals unrelated to chess, like driving. Its model doesn’t have a concept for space, time or movement for that matter. In contrast the AGI by design has no limit on what problems it can work on. So when it tries to figure out a solution to a main problem, the sub-problems it chooses to work on can be anything... literally any path out of the infinite possibilities allowed within the laws of physics and nature.

Lethal Intelligence

570,437 views • 1 year ago