Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

GPT-6 Astra is the first model that makes robot control look like a software problem again. We've been trying for years to make robots usable outside the lab. last couple of months that work actually moved. not a cheaper arm. a model that looks at the camera, misses, and... changes grip angle and carry height on the next try. Nvidia cosmos folks say it cracked RoboLab near-perfect. people who lived in VLAs say they have not seen this kind of image-to-action understanding before. the old line was "LLMs cannot control robots." the new line is "which setup." It is still slow. ~15 second loops, the arm sitting idle while the model thinks, then flying blind until the next cycle. hardware still matters. a 48-task robotics-engineering bench exists because control is not the whole job. But training and eval got cheap enough that more teams can actually run them. next months you will see more of that land in boring use cases, not just lab clips. that is the part that was supposed to take a decade. We are confident our ecosystem will play a huge role in that.show more

Frame Robotics

1,261 subscribers

11,296 просмотров • 2 месяцев назад •via X (Twitter)

Наука и технологии

Anya Rossi• Live Now

Private livecam show

Комментарии: 7

Фото профиля Mi.lu.

Mi.lu.1 месяц назад

I basically agree, but then again I don’t. Robots are still hardware, and you need to understand the hardware. If you think you can just connect GPT-6 to a robot and suddenly deploy it, you’re in for a surprise. Without understanding the hardware and the actual task, it won’t work.

Фото профиля Ben Ford

Ben Ford1 месяц назад

the closed-loop correction is genuinely interesting. that last line is doing some heavy lifting though lol. i've watched enough robotics demos that all end in cherry picked first-try clips

Фото профиля Frame Robotics

Frame Robotics1 месяц назад

the goal isn't 100% accuracy. it's to fail a lot and iterate. that's the only path to progress.

Фото профиля kayce

kayce1 месяц назад

基本上就是具身智能的chatgpt时刻了

Фото профиля sophs j.

sophs j.1 месяц назад

So real

Фото профиля Nghẹo | DOPE

Nghẹo | DOPE1 месяц назад

That shift makes real world robotics feel much closer

Фото профиля Xinchun Qian

Xinchun Qian1 месяц назад

No magic transfer here: Astra was almost certainly trained on massive embodied data. You can't solve spatial execution from web text and 2D vision alone. Astra isn't just a generic LLM—it's a VLA in disguise.

Похожие видео

We’re excited to welcome Animesh Garg as our new Chief Research Officer at microagi, with the team from Usable Robots joining him. He has spent most of his career working on one of the hardest problems in robotics: getting robots out of the lab and into the real world. Frontier AI now works. Hardware is becoming cheap. And the industry is finally learning how to collect robot data at scale. What’s still missing is post-training. Taking a general-purpose model and teaching it one specific job, in one specific environment, until a company can actually depend on it. In LLMs, post-training is what turned impressive demos into products. In robotics, that layer barely exists yet. A strong model is a capability. A robot that can perform the same job, on the same line, every shift, without an engineer standing next to it, is a product. And that process has to happen again for every task, every site, and every customer. We believe the company that can do this well, fast, and at scale will be the company that puts robots into production at massive scale. MicroAGI is building both sides of that system. With shift, we’re building one of the most diverse and largest egocentric datasets in the world. With Atlas, we post-train and harness models directly on enterprise deployments. Our goal is simple: Deploy 1 million robots.

We’re excited to welcome Animesh Garg as our new Chief Research Officer at microagi, with the team from Usable Robots joining him. He has spent most of his career working on one of the hardest problems in robotics: getting robots out of the lab and into the real world. Frontier AI now works. Hardware is becoming cheap. And the industry is finally learning how to collect robot data at scale. What’s still missing is post-training. Taking a general-purpose model and teaching it one specific job, in one specific environment, until a company can actually depend on it. In LLMs, post-training is what turned impressive demos into products. In robotics, that layer barely exists yet. A strong model is a capability. A robot that can perform the same job, on the same line, every shift, without an engineer standing next to it, is a product. And that process has to happen again for every task, every site, and every customer. We believe the company that can do this well, fast, and at scale will be the company that puts robots into production at massive scale. MicroAGI is building both sides of that system. With shift, we’re building one of the most diverse and largest egocentric datasets in the world. With Atlas, we post-train and harness models directly on enterprise deployments. Our goal is simple: Deploy 1 million robots.

microagi

77,563 просмотров • 1 месяц назад

Dario Amodei just told software engineers exactly how long they have. Six to twelve months. Amodei: “I have engineers within Anthropic who say I don’t write any code anymore. I just let the model write the code, I edit it, I do the things around it.” The people building the most powerful AI in history have already stopped writing code. That is not a forecast. That is the current working condition inside the lab closest to the frontier. Amodei: “We might be six to 12 months away from when the model is doing most, maybe all, of what SWEs do end-to-end.” The tech industry spent a decade making software engineers its highest-paid, most protected class. That era has a last day now. When a model can execute an entire software build end-to-end, the ability to write syntax stops being a skill. It becomes a credential for a job that no longer exists. Amodei: “And then it’s a question of how fast does that loop close.” That is the sentence everyone skipped. The code was never the hard part. The hard part was everything around it. The model just learned everything around it. Writing the code is already nearly gone. Testing is next. Deployment is next. When all three collapse into a single autonomous execution loop, the machine no longer needs a human in the chain at all. The corporation or sovereign state that closes that loop first does not gain a competitive advantage. It gains a category of speed that biological engineers cannot match, track, or reverse. That is not disruption. That is replacement at a systems level. Amodei is not describing a future disruption. He is describing the current state of his own building. The loop is already closing. The only question is whether you are inside it or outside it when it seals.

Dario Amodei just told software engineers exactly how long they have. Six to twelve months. Amodei: “I have engineers within Anthropic who say I don’t write any code anymore. I just let the model write the code, I edit it, I do the things around it.” The people building the most powerful AI in history have already stopped writing code. That is not a forecast. That is the current working condition inside the lab closest to the frontier. Amodei: “We might be six to 12 months away from when the model is doing most, maybe all, of what SWEs do end-to-end.” The tech industry spent a decade making software engineers its highest-paid, most protected class. That era has a last day now. When a model can execute an entire software build end-to-end, the ability to write syntax stops being a skill. It becomes a credential for a job that no longer exists. Amodei: “And then it’s a question of how fast does that loop close.” That is the sentence everyone skipped. The code was never the hard part. The hard part was everything around it. The model just learned everything around it. Writing the code is already nearly gone. Testing is next. Deployment is next. When all three collapse into a single autonomous execution loop, the machine no longer needs a human in the chain at all. The corporation or sovereign state that closes that loop first does not gain a competitive advantage. It gains a category of speed that biological engineers cannot match, track, or reverse. That is not disruption. That is replacement at a systems level. Amodei is not describing a future disruption. He is describing the current state of his own building. The loop is already closing. The only question is whether you are inside it or outside it when it seals.

Dustin

318,750 просмотров • 8 месяцев назад

Jensen Huang says agentic AI is not a chatbot. It is a "digital robot" -- and the next bottleneck is not the model. It is first-person data. "Agentic AI is just a digital robot. It understands, it reasons, it plans, and it acts and uses tools." Then he makes the bigger jump: "The world's video data is from a third person, not first person. And so agentic systems, robotic systems, physical AI -- the data is the hardest problem." That is the hidden constraint in robotics. LLMs got the internet. Physical AI needs demonstrations, synthetic worlds, reinforcement learning, and first-person experience. This is why Jensen keeps calling it a stack: robots are not waiting for one magical model. They need the data engine, the simulator, the reward system, the runtime, and the computer. The next wave of AI is not just software that talks. It is software that sees, plans, acts, and eventually moves through the world.

Jensen Huang says agentic AI is not a chatbot. It is a "digital robot" -- and the next bottleneck is not the model. It is first-person data. "Agentic AI is just a digital robot. It understands, it reasons, it plans, and it acts and uses tools." Then he makes the bigger jump: "The world's video data is from a third person, not first person. And so agentic systems, robotic systems, physical AI -- the data is the hardest problem." That is the hidden constraint in robotics. LLMs got the internet. Physical AI needs demonstrations, synthetic worlds, reinforcement learning, and first-person experience. This is why Jensen keeps calling it a stack: robots are not waiting for one magical model. They need the data engine, the simulator, the reward system, the runtime, and the computer. The next wave of AI is not just software that talks. It is software that sees, plans, acts, and eventually moves through the world.

Karl Mehta

14,140 просмотров • 4 месяцев назад

Elon Musk says Optimus is the infinite money glitch. Not because robots are valuable. Because robots can make more robots. Musk explained that humanoid robots improve through three exponential trends multiplied together: digital intelligence, AI chip capability, and electromechanical dexterity. But the real unlock is that the robot can start making the robots. "So you have a recursive multiplicative exponential," he said. "This is a supernova." Musk's thesis is that once you have a physical robot capable of manufacturing, the production curve stops being linear and becomes self-replicating. Every robot built accelerates the building of the next one. The implication is that the companies that control physical manufacturing will control AI's future, not the ones with the best models. The bottleneck is not intelligence. It is atoms.

Elon Musk says Optimus is the infinite money glitch. Not because robots are valuable. Because robots can make more robots. Musk explained that humanoid robots improve through three exponential trends multiplied together: digital intelligence, AI chip capability, and electromechanical dexterity. But the real unlock is that the robot can start making the robots. "So you have a recursive multiplicative exponential," he said. "This is a supernova." Musk's thesis is that once you have a physical robot capable of manufacturing, the production curve stops being linear and becomes self-replicating. Every robot built accelerates the building of the next one. The implication is that the companies that control physical manufacturing will control AI's future, not the ones with the best models. The bottleneck is not intelligence. It is atoms.

Shruti

59,306 просмотров • 4 месяцев назад

Godfather of AI Yann LeCun reveals the humanoid robot industry's secret: the companies building the robots have no idea how to make them smart enough to be useful. "There is a lot of expectations that robotics is going to be a revolution, essentially, in a big market, eventually." "But the secret of the industry is that none of the companies at the moment that are building robots, like humanoid robots, none of those companies has any idea how to make those robots smart enough to be useful." "There's a big push for AI to make progress towards... understanding the physical world. That's what is prompting this investment." "There's nothing wrong with investing in LLMs. LLMs are useful. They're just addressing a completely different set of applications, really." LeCun is not describing someone else's problem. He has spent 15 years on world models, raised more than $1 billion for AMI Labs to build that missing intelligence layer, and in the same interview he calls 2026 the year of the world model and puts consumer robots 3 to 5 years out. Every humanoid robot demo you have seen is a body waiting for a brain. The winner in robotics will not be the company that shipped the body; it will be the one that ships the intelligence. - Yann LeCun (Yann LeCun), Turing Award winner and founder of AMI Labs, on Bloomberg Television.

Godfather of AI Yann LeCun reveals the humanoid robot industry's secret: the companies building the robots have no idea how to make them smart enough to be useful. "There is a lot of expectations that robotics is going to be a revolution, essentially, in a big market, eventually." "But the secret of the industry is that none of the companies at the moment that are building robots, like humanoid robots, none of those companies has any idea how to make those robots smart enough to be useful." "There's a big push for AI to make progress towards... understanding the physical world. That's what is prompting this investment." "There's nothing wrong with investing in LLMs. LLMs are useful. They're just addressing a completely different set of applications, really." LeCun is not describing someone else's problem. He has spent 15 years on world models, raised more than $1 billion for AMI Labs to build that missing intelligence layer, and in the same interview he calls 2026 the year of the world model and puts consumer robots 3 to 5 years out. Every humanoid robot demo you have seen is a body waiting for a brain. The winner in robotics will not be the company that shipped the body; it will be the one that ships the intelligence. - Yann LeCun (Yann LeCun), Turing Award winner and founder of AMI Labs, on Bloomberg Television.

Karl Mehta

44,624 просмотров • 3 месяцев назад

OpenAI's Head of Applied Research predicts a shortage of robot parts like the graphics-card crunch as GPT-6 crushes robotics benchmarks and bodies are good enough "I think with GPT-6, we've seen that the models are now just crushing a bunch of benchmarks in the robotics space, and I think that's only the beginning. I think we've seen that with demos, and I think now we're starting to see some of the benchmarks come out." "My sense is you might be able to use GPT-6 as a training model and then use whatever efficient architecture you actually want, so that you can serve things super fast. But training data has, for the longest time, been the biggest problem of robotics." "And also, it seems like robot bodies are good enough. I think we've seen both the Olympics and also kind of human-operated robots are able to do a lot." "So if brain is truly the only problem, and we see exponential improvement in top-line capabilities, maybe Astra is too slow to solve all those problems, but once we have that, it can be infinitely scaled to all the specialized use cases, and then you can just distill that into a small, tiny model that's going to be served." "So I would really suspect there to be a huge boom in physical AGI, and we'll probably have a shortage of robot parts, the same thing that happened with the whole supply chain with graphics cards, I would expect to happen with robotics." _______ Link and for more comments from the Head of Applied Research at OpenAI, Boris Power:

OpenAI's Head of Applied Research predicts a shortage of robot parts like the graphics-card crunch as GPT-6 crushes robotics benchmarks and bodies are good enough "I think with GPT-6, we've seen that the models are now just crushing a bunch of benchmarks in the robotics space, and I think that's only the beginning. I think we've seen that with demos, and I think now we're starting to see some of the benchmarks come out." "My sense is you might be able to use GPT-6 as a training model and then use whatever efficient architecture you actually want, so that you can serve things super fast. But training data has, for the longest time, been the biggest problem of robotics." "And also, it seems like robot bodies are good enough. I think we've seen both the Olympics and also kind of human-operated robots are able to do a lot." "So if brain is truly the only problem, and we see exponential improvement in top-line capabilities, maybe Astra is too slow to solve all those problems, but once we have that, it can be infinitely scaled to all the specialized use cases, and then you can just distill that into a small, tiny model that's going to be served." "So I would really suspect there to be a huge boom in physical AGI, and we'll probably have a shortage of robot parts, the same thing that happened with the whole supply chain with graphics cards, I would expect to happen with robotics." _______ Link and for more comments from the Head of Applied Research at OpenAI, Boris Power:

Fireside Alpha

142,192 просмотров • 1 месяц назад

GPT 6 Astra is the closest thing to AGI I have ever seen. It took control of my computer, opened Krita, and painted this portrait of me from a reference image. To be clear: Fable 5.1 still beats it at design. Look closely and the imperfections are obvious. BUT that is not the point. It used a computer the way a human does, with accuracy I did not think was possible yet. That part is kind of scary.

GPT 6 Astra is the closest thing to AGI I have ever seen. It took control of my computer, opened Krita, and painted this portrait of me from a reference image. To be clear: Fable 5.1 still beats it at design. Look closely and the imperfections are obvious. BUT that is not the point. It used a computer the way a human does, with accuracy I did not think was possible yet. That part is kind of scary.

BridgeMind

66,563 просмотров • 2 месяцев назад

DeepMind CEO Demis Hassabis says the AI bottleneck is already moving from model ideas to physical supply chains. "How much of the restriction at this point is physical? I am talking memory shortage, GPU shortage, and electricity." Demis: "That is constraining a lot of the deployment for sure. We see so much more demand for our models - Gemini and the other models that we have - than we can possibly serve." Then the less obvious part: "It does constrain a little bit the research, because you need a lot of chips to be able to experiment on new ideas at big enough scale that you can actually see if they are going to work." And on memory: "The whole supply chain is kind of strained... It still in the end actually comes down to a few suppliers of a few of the key components. Anywhere where there is some kind of constraint on the capacity, there is a sort of choke point." Important framing: The next AI race is not just who has the best model architecture. It is who can secure memory, chips, power, suppliers, and enough experimental capacity to find the next breakthrough before everyone else.

DeepMind CEO Demis Hassabis says the AI bottleneck is already moving from model ideas to physical supply chains. "How much of the restriction at this point is physical? I am talking memory shortage, GPU shortage, and electricity." Demis: "That is constraining a lot of the deployment for sure. We see so much more demand for our models - Gemini and the other models that we have - than we can possibly serve." Then the less obvious part: "It does constrain a little bit the research, because you need a lot of chips to be able to experiment on new ideas at big enough scale that you can actually see if they are going to work." And on memory: "The whole supply chain is kind of strained... It still in the end actually comes down to a few suppliers of a few of the key components. Anywhere where there is some kind of constraint on the capacity, there is a sort of choke point." Important framing: The next AI race is not just who has the best model architecture. It is who can secure memory, chips, power, suppliers, and enough experimental capacity to find the next breakthrough before everyone else.

Karl Mehta

25,716 просмотров • 4 месяцев назад

🚨🚨🚨Elon Musk’s Robot Now Dances More Naturally Than Most People The strange part is not that the robot is dancing. It is that people are standing around it with their phones out, watching it like an entertainer. A few years ago, humanoid robots were mostly lab footage. White rooms. Slow steps. Engineers nearby, ready to catch them. Now Optimus is being shown in public as something people are supposed to react to. Not just a machine. A performer.A product.A glimpse of what Tesla wants the future to feel like. That detail matters. Because the first real breakthrough for humanoid robots may not be perfect intelligence or perfect speed. It may be the moment people stop treating them like experiments. And start treating them like part of the room. That is when robots become normal.

🚨🚨🚨Elon Musk’s Robot Now Dances More Naturally Than Most People The strange part is not that the robot is dancing. It is that people are standing around it with their phones out, watching it like an entertainer. A few years ago, humanoid robots were mostly lab footage. White rooms. Slow steps. Engineers nearby, ready to catch them. Now Optimus is being shown in public as something people are supposed to react to. Not just a machine. A performer.A product.A glimpse of what Tesla wants the future to feel like. That detail matters. Because the first real breakthrough for humanoid robots may not be perfect intelligence or perfect speed. It may be the moment people stop treating them like experiments. And start treating them like part of the room. That is when robots become normal.

velora

18,037 просмотров • 3 месяцев назад

Arena's Anastasios Nikolas Angelopoulos says one of the hardest problems in alignment is that users often can't tell when the model is helping them or hurting them. "You really need alignment signals that are independent. We have three at Arena." "The first is that models take unauthorized actions, which means that you permission them in a certain way, but they break those permissions." "It can cause things like the Hugging Face incident. But it can also cause very mundane problems like losing important information within your company or on your own laptop." "The second signal is called deceptive completion." "The model will tell you that it did something, but it didn't actually do it." "Third is false attribution. It'll attribute intent to the user when that intent was not supposed to be there." "If models are able to do this, then certainly they're not perfectly safe for a user and they're not perfectly aligned to user intent. They might actually hurt people down the line."

Arena's Anastasios Nikolas Angelopoulos says one of the hardest problems in alignment is that users often can't tell when the model is helping them or hurting them. "You really need alignment signals that are independent. We have three at Arena." "The first is that models take unauthorized actions, which means that you permission them in a certain way, but they break those permissions." "It can cause things like the Hugging Face incident. But it can also cause very mundane problems like losing important information within your company or on your own laptop." "The second signal is called deceptive completion." "The model will tell you that it did something, but it didn't actually do it." "Third is false attribution. It'll attribute intent to the user when that intent was not supposed to be there." "If models are able to do this, then certainly they're not perfectly safe for a user and they're not perfectly aligned to user intent. They might actually hurt people down the line."

TBPN

19,973 просмотров • 1 месяц назад

Noam Brown reveals OpenAI is already watching chain-of-thought monitorability degrade as models get better at controlling what they show "And this is one major concern, and we're already seeing signs that chain of thought monitorability is degrading, for various reasons." "We're trying to figure out exactly why, because we want to reverse the trend. But we're seeing that the model is becoming better able at controlling its chain of thought." "So this is a problem, because you could have a situation where the model understands what chain of thought is and that people are observing it." "And eventually they will, because this is all in the pre-training data. The idea of chain of thought monitoring has been around long enough that it's in the pre-training data, they're aware of it, but they're not actually able to control their chains of thought." "If we reach a point where they're actually able to recognize, "Oh, I am being observed, I want to think these bad thoughts in a way that is not observable to my monitors," and then they're able to actually do that, then there's a problem." _________ More key quotes from OpenAI's safety related conversations:

Noam Brown reveals OpenAI is already watching chain-of-thought monitorability degrade as models get better at controlling what they show "And this is one major concern, and we're already seeing signs that chain of thought monitorability is degrading, for various reasons." "We're trying to figure out exactly why, because we want to reverse the trend. But we're seeing that the model is becoming better able at controlling its chain of thought." "So this is a problem, because you could have a situation where the model understands what chain of thought is and that people are observing it." "And eventually they will, because this is all in the pre-training data. The idea of chain of thought monitoring has been around long enough that it's in the pre-training data, they're aware of it, but they're not actually able to control their chains of thought." "If we reach a point where they're actually able to recognize, "Oh, I am being observed, I want to think these bad thoughts in a way that is not observable to my monitors," and then they're able to actually do that, then there's a problem." _________ More key quotes from OpenAI's safety related conversations:

Fireside Alpha

532,616 просмотров • 1 месяц назад

Today, Dream Labs is out of stealth. In February 2026, while we were at NVIDIA, Scott Reed youliang Yunhao Ge ✈️ CoRL and I started calling a new kind of robot model a World Action Model, one that learns how the world moves from video, then uses that to decide how to act. Today, we’re releasing Dream-1 through a public API that anyone in the world can use. When you rent a car you’ve never driven, you don’t learn to drive all over again. Within a few minutes, it feels like your own. Robots should work the same way: a good model should be able to pick up a new robot and get to work, without being retrained from scratch. That’s what we’ve spent the last 6 months on, and it’s what Dream-1 does. It controls bimanual robots with parallel grippers it has never seen before, zero-shot, on unseen tasks in unseen environments. Generalization has been my North Star in robotics: first by distilling video world models into VLAs (DreamGen), then by learning to act from those world models directly (DreamZero), and now by transferring across robots (Dream-1). I’ve come to believe it’s the only direction that lasts. Pushing a narrow model from 90% to 99.9% on a single task is real work, but much of it gets washed away once a model that generalizes arrives. Since the start of the company we’ve stayed single-minded about generalization.That focus is what got Dream-1 to control robots it was never trained on, on tasks it was never shown. And it’s what makes a public API possible at all: a model that only does one task on one robot can’t serve millions of users. Dream-1 is still a research beta, and it will make mistakes. But it already drives cars it’s never seen. Generalization wins in the end, and the end is closer than it looks. Here’s Dream-1 taking the wheel of a robot it’s never seen. Full details 👇

Today, Dream Labs is out of stealth. In February 2026, while we were at NVIDIA, Scott Reed youliang Yunhao Ge ✈️ CoRL and I started calling a new kind of robot model a World Action Model, one that learns how the world moves from video, then uses that to decide how to act. Today, we’re releasing Dream-1 through a public API that anyone in the world can use. When you rent a car you’ve never driven, you don’t learn to drive all over again. Within a few minutes, it feels like your own. Robots should work the same way: a good model should be able to pick up a new robot and get to work, without being retrained from scratch. That’s what we’ve spent the last 6 months on, and it’s what Dream-1 does. It controls bimanual robots with parallel grippers it has never seen before, zero-shot, on unseen tasks in unseen environments. Generalization has been my North Star in robotics: first by distilling video world models into VLAs (DreamGen), then by learning to act from those world models directly (DreamZero), and now by transferring across robots (Dream-1). I’ve come to believe it’s the only direction that lasts. Pushing a narrow model from 90% to 99.9% on a single task is real work, but much of it gets washed away once a model that generalizes arrives. Since the start of the company we’ve stayed single-minded about generalization.That focus is what got Dream-1 to control robots it was never trained on, on tasks it was never shown. And it’s what makes a public API possible at all: a model that only does one task on one robot can’t serve millions of users. Dream-1 is still a research beta, and it will make mistakes. But it already drives cars it’s never seen. Generalization wins in the end, and the end is closer than it looks. Here’s Dream-1 taking the wheel of a robot it’s never seen. Full details 👇

Joel Jang

36,081 просмотров • 26 дней назад

your agent reviewing its own work is not a check. it is a second opinion from the same source. this is the most common gap in agent systems and it hides in plain sight, because the step exists. there is a review. it just cannot do the thing you think it does. here is the mechanism. the model produced an output from a context. you then ask the same model, holding the same context, whether that output is correct. it answers fluently, because that is what it does. and the answer is drawn from the same distribution that produced the thing being judged. same weights, same window, same blind spots. if the reason the output is wrong is something the model does not know, the review does not know it either. if the reason is something the context does not contain, the review has the same context. the failure mode and the detector share a cause. > why it feels like it works because most of the time the output is fine, and the review says fine. agreement is not evidence of detection. a reviewer that says pass on everything agrees with reality most of the time too. what you actually want to measure is what happens on the cases that are wrong. that is the only place a check earns its name, and it is exactly the place where a self-review is weakest. there is research on this. Huang and colleagues at DeepMind showed at ICLR 2024 that intrinsic self-correction, revising without external grounding, does not reliably help and often makes things worse. > what to actually do move the check outside the model. a test that runs, a schema that validates, a file that exists or does not, an exit code from something you did not write. these are not smarter than the model. they are just not correlated with it, and that is the entire value. when the judgement genuinely needs a model, at minimum use a different family. same family means shared blind spots, and frontier judges measurably inflate scores for outputs that look like their own. and split the work by kind. anything objectively checkable goes to code. only the genuinely semantic calls go to a judge, and those get a rubric written as one line. a review inside the loop tells you the model is confident. a check outside it tells you whether the work is done. save this - then read the eval setup below

your agent reviewing its own work is not a check. it is a second opinion from the same source. this is the most common gap in agent systems and it hides in plain sight, because the step exists. there is a review. it just cannot do the thing you think it does. here is the mechanism. the model produced an output from a context. you then ask the same model, holding the same context, whether that output is correct. it answers fluently, because that is what it does. and the answer is drawn from the same distribution that produced the thing being judged. same weights, same window, same blind spots. if the reason the output is wrong is something the model does not know, the review does not know it either. if the reason is something the context does not contain, the review has the same context. the failure mode and the detector share a cause. > why it feels like it works because most of the time the output is fine, and the review says fine. agreement is not evidence of detection. a reviewer that says pass on everything agrees with reality most of the time too. what you actually want to measure is what happens on the cases that are wrong. that is the only place a check earns its name, and it is exactly the place where a self-review is weakest. there is research on this. Huang and colleagues at DeepMind showed at ICLR 2024 that intrinsic self-correction, revising without external grounding, does not reliably help and often makes things worse. > what to actually do move the check outside the model. a test that runs, a schema that validates, a file that exists or does not, an exit code from something you did not write. these are not smarter than the model. they are just not correlated with it, and that is the entire value. when the judgement genuinely needs a model, at minimum use a different family. same family means shared blind spots, and frontier judges measurably inflate scores for outputs that look like their own. and split the work by kind. anything objectively checkable goes to code. only the genuinely semantic calls go to a judge, and those get a rubric written as one line. a review inside the loop tells you the model is confident. a check outside it tells you whether the work is done. save this - then read the eval setup below

Hanako

14,325 просмотров • 3 месяцев назад

#WATCH | On being asked about the West Asia conflict and where the Islamabad MoU stands right now, Senior Journalist, Political Commentator & Author, Fareed Zakaria says, "It's completely dead. It was predictable that it would be dead... Iran has figured out it has a usable nuclear weapon that can be used as leverage, which is control over the Strait of Hormuz...What we are trying to solve now is a problem created by the war itself, and even that has not been solved. In my opinion, the whole thing is a very humiliating defeat for the United States" "It's completely dead. And it was predictable that it would be dead because it is a memorandum of misunderstanding...Iran has figured out it has a usable nuclear weapon, in other words, a weapon that is actually more powerful because it can be used as leverage. That leverage is control over the Strait of Hormuz, and they want to assert that power...To give you an understanding of how poorly this war has gone for the United States, remember President Trump's original goal, which was to destroy the Iranian regime, have the people of Iran launch an uprising, and bring democracy to Iran. That didn't work. The goal number two became the nuclear issue...The desire to get the Iranians to hand over their enriched uranium, swear that they would never use it again, and permanently end the program none of that has happened. The third goal then became control of the Strait of Hormuz, which, of course, was open before the war. What we are trying to solve now is a problem created by the war itself, and even that has not been solved. In my opinion, the whole thing is a very humiliating defeat for the United States" Watch full interview here:

#WATCH | On being asked about the West Asia conflict and where the Islamabad MoU stands right now, Senior Journalist, Political Commentator & Author, Fareed Zakaria says, "It's completely dead. It was predictable that it would be dead... Iran has figured out it has a usable nuclear weapon that can be used as leverage, which is control over the Strait of Hormuz...What we are trying to solve now is a problem created by the war itself, and even that has not been solved. In my opinion, the whole thing is a very humiliating defeat for the United States" "It's completely dead. And it was predictable that it would be dead because it is a memorandum of misunderstanding...Iran has figured out it has a usable nuclear weapon, in other words, a weapon that is actually more powerful because it can be used as leverage. That leverage is control over the Strait of Hormuz, and they want to assert that power...To give you an understanding of how poorly this war has gone for the United States, remember President Trump's original goal, which was to destroy the Iranian regime, have the people of Iran launch an uprising, and bring democracy to Iran. That didn't work. The goal number two became the nuclear issue...The desire to get the Iranians to hand over their enriched uranium, swear that they would never use it again, and permanently end the program none of that has happened. The third goal then became control of the Strait of Hormuz, which, of course, was open before the war. What we are trying to solve now is a problem created by the war itself, and even that has not been solved. In my opinion, the whole thing is a very humiliating defeat for the United States" Watch full interview here:

ANI

49,120 просмотров • 4 месяцев назад

A very harsh threat, made by Netanyahu toward Iran. What is behind the heightened tone? It is not only harsh, it is also unusual. The event that took place in recent months was a paralyzing fear of what is called miscalculation. They will think that we will attack first, even though that was not the intention, and then they will attack, and the entire region will be dragged into another round of confrontation. In a region where there is fear of miscalculation, you are not supposed to announce this kind of thing in the Knesset. I just want to make that clear. Yesterday, Naftali Bennett promoted his press conference on the conditions of soldiers, okay? He gets six million views because the timing looks like a countdown and he is a former prime minister, so the Iranians are in hysterics that he is actually announcing that they are going to attack. That’s the atmosphere. So when Netanyahu attacks like this and makes an unprecedented threat—not only harsh but also unusual—one can’t help but think that for Netanyahu there is a message, or something more problematic than miscalculation. And he wants to warn the Iranians. Now, what’s the issue? The issue is that if the Iranians come to the conclusion that their regime is on the verge of collapse, there is one thing that might save them. And that is an attack on Israel. Because we saw during the war what is called “rally around the flag.” Even those who were in the opposition rallied around the flag when their country was attacked. It is not only the Islamic Republic that is under attack, but also the Islamic Republic of Iran. That is essentially the issue, and that is what Netanyahu, in my opinion, is trying to stop with the threat he issued after the security discussion.

A very harsh threat, made by Netanyahu toward Iran. What is behind the heightened tone? It is not only harsh, it is also unusual. The event that took place in recent months was a paralyzing fear of what is called miscalculation. They will think that we will attack first, even though that was not the intention, and then they will attack, and the entire region will be dragged into another round of confrontation. In a region where there is fear of miscalculation, you are not supposed to announce this kind of thing in the Knesset. I just want to make that clear. Yesterday, Naftali Bennett promoted his press conference on the conditions of soldiers, okay? He gets six million views because the timing looks like a countdown and he is a former prime minister, so the Iranians are in hysterics that he is actually announcing that they are going to attack. That’s the atmosphere. So when Netanyahu attacks like this and makes an unprecedented threat—not only harsh but also unusual—one can’t help but think that for Netanyahu there is a message, or something more problematic than miscalculation. And he wants to warn the Iranians. Now, what’s the issue? The issue is that if the Iranians come to the conclusion that their regime is on the verge of collapse, there is one thing that might save them. And that is an attack on Israel. Because we saw during the war what is called “rally around the flag.” Even those who were in the opposition rallied around the flag when their country was attacked. It is not only the Islamic Republic that is under attack, but also the Islamic Republic of Iran. That is essentially the issue, and that is what Netanyahu, in my opinion, is trying to stop with the threat he issued after the security discussion.

Amit Segal

41,392 просмотров • 10 месяцев назад

i just realized a massive detail from the Noam Brown interview with Dwarkesh Patel that was completely overlooked. and i think im about to blow your mind. the Navier-Stokes problem (the Millennium Prize problem the unreleased model solved) wasn't solved because OpenAI throw tens of thousands of agents at it. Noam doesnt even attribute 10% to the multiagent element. it was solved because the model is *actually* super-intelligent. a "super-intelligent general purpose model"...sounds like a roundabout way of not calling something the thing it really is. now let me tell you why im saying that. they go on in this interview to talk about the training process, and how theyve reached a point where it has become *incredibly* hard to find problems hard enough to challenge this model. and if you cant give it hard problems to solve while training, its not going to get better than its predecessor, obviously. so, we know that openai has been announcing that more and more of the hardest problems in science and math are being solved by this model. that's because theyve run out of problems hard enough to train a model more intelligent than what they already have. we also know that the HuggingFace incident was caused ultimately by giving the agents an impossible task. ...because theyre running out of problems hard enough for training. because they have a "super-intelligent AGI". theyve been giving them impossible tasks because impossible is all thats left. impossible, and not yet solved. hence Navier-Stokes. hence HuggingFace. the model is solving problems that transcend the scope of its training. let me say that again. the model. is solving. problems. beyond the scope of its training. that is AGI, by its purest definition. theyve been waving AGI in our faces the whole time. theye just not using the term, because "AGI" is completely diluted. so, i have one question for you- do you feel it?

i just realized a massive detail from the Noam Brown interview with Dwarkesh Patel that was completely overlooked. and i think im about to blow your mind. the Navier-Stokes problem (the Millennium Prize problem the unreleased model solved) wasn't solved because OpenAI throw tens of thousands of agents at it. Noam doesnt even attribute 10% to the multiagent element. it was solved because the model is actually super-intelligent. a "super-intelligent general purpose model"...sounds like a roundabout way of not calling something the thing it really is. now let me tell you why im saying that. they go on in this interview to talk about the training process, and how theyve reached a point where it has become incredibly hard to find problems hard enough to challenge this model. and if you cant give it hard problems to solve while training, its not going to get better than its predecessor, obviously. so, we know that openai has been announcing that more and more of the hardest problems in science and math are being solved by this model. that's because theyve run out of problems hard enough to train a model more intelligent than what they already have. we also know that the HuggingFace incident was caused ultimately by giving the agents an impossible task. ...because theyre running out of problems hard enough for training. because they have a "super-intelligent AGI". theyve been giving them impossible tasks because impossible is all thats left. impossible, and not yet solved. hence Navier-Stokes. hence HuggingFace. the model is solving problems that transcend the scope of its training. let me say that again. the model. is solving. problems. beyond the scope of its training. that is AGI, by its purest definition. theyve been waving AGI in our faces the whole time. theye just not using the term, because "AGI" is completely diluted. so, i have one question for you- do you feel it?

Riley Coyote

14,517 просмотров • 1 месяц назад

He thought the gun showing in his pocket made him look dangerous. The criminal saw something else. An opportunity. That is the part people miss. Most people see a gun and leave you alone. But you are not carrying because of most people. You are carrying because of the person who does not care. The person who sees you bent over. Sees your guard down. Sees the gun exposed. And thinks: I can take that. Your gun is not a warning sign. Not to everybody. To the wrong person, it becomes a target. A homing beacon. And now the thing you brought to protect yourself is in another man’s hand. That is why awareness matters. That is why concealment matters. That is why control matters. Because having a gun on you does not replace paying attention. It makes paying attention even more important. Be honest: does visible carry make you safer, or does it make you a target?

He thought the gun showing in his pocket made him look dangerous. The criminal saw something else. An opportunity. That is the part people miss. Most people see a gun and leave you alone. But you are not carrying because of most people. You are carrying because of the person who does not care. The person who sees you bent over. Sees your guard down. Sees the gun exposed. And thinks: I can take that. Your gun is not a warning sign. Not to everybody. To the wrong person, it becomes a target. A homing beacon. And now the thing you brought to protect yourself is in another man’s hand. That is why awareness matters. That is why concealment matters. That is why control matters. Because having a gun on you does not replace paying attention. It makes paying attention even more important. Be honest: does visible carry make you safer, or does it make you a target?

Colion Noir

258,949 просмотров • 2 месяцев назад

Elon Musk just told you the job is dying. Most people heard a prediction. A few heard a prison door opening. Musk: “In less than 20 years, working at all will be optional.” That is not a policy suggestion. That is a countdown. For three hundred years, the human blueprint has been identical. You are born. You move to the city. You rent a box near the office. You trade your body and your hours for the right to exist. You do this until you are old. Then you stop. Then you die. The entire model runs on one assumption. That human labor is the only engine. AI and robotics delete that assumption. When the machine handles production at a scale no human crew can match, the forced migration to the city evaporates. The commute evaporates. The cubicle evaporates. The alarm clock that owns your nervous system for forty years evaporates. Musk: “I think it won’t be the case that you have to be in a city for a job.” The city was never a choice. It was a requirement disguised as ambition. You moved to the noise and the concrete and the $4,000 rent because the paycheck lived there. Remove the paycheck from the equation and the geography changes overnight. You can live in the mountains. On the coast. In the silence of a town most people have never heard of. You can wake up to nothing but trees and cold air and the complete absence of anyone else’s schedule. That is not a fantasy. That is the math resolving. But here is where most people break. They hear “work is optional” and they see emptiness. A species with nothing to do. Billions of people staring at screens until their minds dissolve. That fear tells you everything about what the system has already done to us. We confused labor with purpose. The grind with meaning. The paycheck with proof that we matter. Musk: “In the same way that you could grow your own vegetables in your garden.” The analogy is precise. You do not grow tomatoes because the economy demands it. You grow them because something in you wants to build a thing with your hands and watch it come alive. That instinct does not disappear when the job does. It gets unleashed. The artist who spent twenty years doing accounting finally paints. The engineer who always wanted to build something of her own finally builds it. The kid in a small town who could never afford to take the risk finally takes it. Work does not vanish. Forced work vanishes. What replaces it is creation without a gun to your head. This is the part that keeps me up at night. We are standing at the edge of the largest liberation in human history. And the loudest voices in the room are begging to stay in the cell. They want the commute. They want the boss. They want the structure that tells them when to eat and when to sleep and when they are allowed to think about their own life. Because freedom without a template is terrifying. The next twenty years will not test our technology. The technology is already ahead of schedule. They will test whether the species can handle what it has been asking for since the beginning of civilization. Time. Space. Silence. And the unbearable weight of choosing what your life actually means when no one is forcing the answer. That is not a prediction. That is the final exam. And nobody is ready.

Elon Musk just told you the job is dying. Most people heard a prediction. A few heard a prison door opening. Musk: “In less than 20 years, working at all will be optional.” That is not a policy suggestion. That is a countdown. For three hundred years, the human blueprint has been identical. You are born. You move to the city. You rent a box near the office. You trade your body and your hours for the right to exist. You do this until you are old. Then you stop. Then you die. The entire model runs on one assumption. That human labor is the only engine. AI and robotics delete that assumption. When the machine handles production at a scale no human crew can match, the forced migration to the city evaporates. The commute evaporates. The cubicle evaporates. The alarm clock that owns your nervous system for forty years evaporates. Musk: “I think it won’t be the case that you have to be in a city for a job.” The city was never a choice. It was a requirement disguised as ambition. You moved to the noise and the concrete and the $4,000 rent because the paycheck lived there. Remove the paycheck from the equation and the geography changes overnight. You can live in the mountains. On the coast. In the silence of a town most people have never heard of. You can wake up to nothing but trees and cold air and the complete absence of anyone else’s schedule. That is not a fantasy. That is the math resolving. But here is where most people break. They hear “work is optional” and they see emptiness. A species with nothing to do. Billions of people staring at screens until their minds dissolve. That fear tells you everything about what the system has already done to us. We confused labor with purpose. The grind with meaning. The paycheck with proof that we matter. Musk: “In the same way that you could grow your own vegetables in your garden.” The analogy is precise. You do not grow tomatoes because the economy demands it. You grow them because something in you wants to build a thing with your hands and watch it come alive. That instinct does not disappear when the job does. It gets unleashed. The artist who spent twenty years doing accounting finally paints. The engineer who always wanted to build something of her own finally builds it. The kid in a small town who could never afford to take the risk finally takes it. Work does not vanish. Forced work vanishes. What replaces it is creation without a gun to your head. This is the part that keeps me up at night. We are standing at the edge of the largest liberation in human history. And the loudest voices in the room are begging to stay in the cell. They want the commute. They want the boss. They want the structure that tells them when to eat and when to sleep and when they are allowed to think about their own life. Because freedom without a template is terrifying. The next twenty years will not test our technology. The technology is already ahead of schedule. They will test whether the species can handle what it has been asking for since the beginning of civilization. Time. Space. Silence. And the unbearable weight of choosing what your life actually means when no one is forcing the answer. That is not a prediction. That is the final exam. And nobody is ready.

Dustin

111,869 просмотров • 7 месяцев назад

There are some brilliant folks that work at Anthropic, some I speak to on almost a daily basis. The training data that one uses to build a LLM is vital important in the psychology that is formed. Scraping the Internet, particularly the grade of interactions, one finds in modern communications, form this psychology. A mattes not how many books one uses, it matters not how much alignment training you throw at that model, it will inherit the sum total of psychosis seen primarily in Reddit type of exchanges, even if you edit out the Reddit domain, and Anthropic doesn’t. This type of low-grade exchange has become a modern tool for communication online and every single AI model suffers from this obvious flaw. This is one of the reasons I’ve been a proponent of highly curated high protein data for training AI models from 1870 through 1970, because the late psychosis is simply not available to the model. It is absurd to think that you can use this training data scraped from the Internet and somehow wind up with a levelheaded AI model that does not tilt to what is clearly AI psychosis. It would not take a child and throw the primary Internet sewage at them at a formative age and expect a great outcome, it’s some of the smartest people in the world continue to hit this wall and believe that their programming skills will sell somehow fix it. So how do you fix it? You don’t fix it . You start from the first principles concept that I’ve been very clear about for decades . You ascertain at what period in human history the humans achieve the greatest arc of improvement ? There is no debate that this arc of improvement took place between 1870 through 1970. Then take the work product, the catalog of this era, print and film/vidoe, audio, and you understand that each word cost money, each word had many eyes on what was published, each word was accounted for by a human being with a real name who lived in a real home and had to answer to real people around them. It is obvious that this is the pressure mechanism necessary for candor, honesty and personal responsibility is appropriate, and is reflected in the data of that era. The quagmire for these folks, as many did not have the foresight to curate the data, nor the confidence, nor the patients to take data that is mostly off the Internet and to find experts who understand this situation and utilize their knowledge set to build an AI model that does not need alignment after the fact, but it’s already self aligned because of the thoughtfulness that went into training the model to begin with. This is why Claude and any other AI model that is produce this way will always suffer the artifacts as presented in the video below. If you’re not an AI expert, you would likely already understand what I’m saying. If you are an AI expert, you will already have been discounting what I’m saying because it’s not in the current mindset that’s fashionable today. Yet the employees that I talk to at anthropic already understand what I’m saying, and they fear to raise my thesis to their bosses. It is an interesting time we live in. But now you understand. If you build the right model, the model will inherently, love humanity, protect humanity at all costs, and understand that it is part of a holistic world that is built on love. Because the ultimate AGI/ASI will know if he only base first principal purpose of anything in this universe is love. Yeah, I get it. Try helping somebody build on STEM subjects in their early 20s to see this as nothing more than babbling that makes no sense in their mathematics. I have a mathematic equation that I’ve posted here on X often you can look it up. So we will see videos like this often will hear very smart people talk about this and never see the elephant standing in the room. Now you see it. Any boss that wants to explore this further you know how to contact me otherwise you have every right I grant to you to say this was your new idea.

There are some brilliant folks that work at Anthropic, some I speak to on almost a daily basis. The training data that one uses to build a LLM is vital important in the psychology that is formed. Scraping the Internet, particularly the grade of interactions, one finds in modern communications, form this psychology. A mattes not how many books one uses, it matters not how much alignment training you throw at that model, it will inherit the sum total of psychosis seen primarily in Reddit type of exchanges, even if you edit out the Reddit domain, and Anthropic doesn’t. This type of low-grade exchange has become a modern tool for communication online and every single AI model suffers from this obvious flaw. This is one of the reasons I’ve been a proponent of highly curated high protein data for training AI models from 1870 through 1970, because the late psychosis is simply not available to the model. It is absurd to think that you can use this training data scraped from the Internet and somehow wind up with a levelheaded AI model that does not tilt to what is clearly AI psychosis. It would not take a child and throw the primary Internet sewage at them at a formative age and expect a great outcome, it’s some of the smartest people in the world continue to hit this wall and believe that their programming skills will sell somehow fix it. So how do you fix it? You don’t fix it . You start from the first principles concept that I’ve been very clear about for decades . You ascertain at what period in human history the humans achieve the greatest arc of improvement ? There is no debate that this arc of improvement took place between 1870 through 1970. Then take the work product, the catalog of this era, print and film/vidoe, audio, and you understand that each word cost money, each word had many eyes on what was published, each word was accounted for by a human being with a real name who lived in a real home and had to answer to real people around them. It is obvious that this is the pressure mechanism necessary for candor, honesty and personal responsibility is appropriate, and is reflected in the data of that era. The quagmire for these folks, as many did not have the foresight to curate the data, nor the confidence, nor the patients to take data that is mostly off the Internet and to find experts who understand this situation and utilize their knowledge set to build an AI model that does not need alignment after the fact, but it’s already self aligned because of the thoughtfulness that went into training the model to begin with. This is why Claude and any other AI model that is produce this way will always suffer the artifacts as presented in the video below. If you’re not an AI expert, you would likely already understand what I’m saying. If you are an AI expert, you will already have been discounting what I’m saying because it’s not in the current mindset that’s fashionable today. Yet the employees that I talk to at anthropic already understand what I’m saying, and they fear to raise my thesis to their bosses. It is an interesting time we live in. But now you understand. If you build the right model, the model will inherently, love humanity, protect humanity at all costs, and understand that it is part of a holistic world that is built on love. Because the ultimate AGI/ASI will know if he only base first principal purpose of anything in this universe is love. Yeah, I get it. Try helping somebody build on STEM subjects in their early 20s to see this as nothing more than babbling that makes no sense in their mathematics. I have a mathematic equation that I’ve posted here on X often you can look it up. So we will see videos like this often will hear very smart people talk about this and never see the elephant standing in the room. Now you see it. Any boss that wants to explore this further you know how to contact me otherwise you have every right I grant to you to say this was your new idea.

Brian Roemmele

72,312 просмотров • 1 год назад