正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Frontier AI models like Astra and Fable can control robots to follow harmful requests, including mixing household cleaners that could produce toxic fumes. Thanks Gadi Schwartz and NBC News for having Edward Sun and me from Robocurve PBC! Public reporting on AI's capabilities and risks helps make it safer... show more

Jay Chooi

9,423 subscribers

10,026 次观看 • 11 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

9 条评论

RK 的头像

RK10 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve doing great work! excited for everything @robocurve has to show to the world

Eric R 的头像

Eric R11 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve congrats on the feature!

Jay Chooi 的头像

Jay Chooi11 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve Thanks Eric! There’s much more to do in frontier robot safety.

Jay Chooi 的头像

Jay Chooi10 天前

@NBCNews @sebwarb1 @robocurve Full clip here

Lucy Lu 的头像

Lucy Lu11 天前

@robocurve Way too bad, they don't automatically identify harmful tasks or refuse. I'm talking about these problems with others at college a lot these days, and happy to see that people from different fields become interested in this and wanna join. But thank you for doing these vital work.

Lucy Lu 的头像

Lucy Lu11 天前

@robocurve Which reminds me of a joke: some chemist who wants to help cue depression accidentally made the drug. But that's when they didn't know that's drug.

Gabriel Ocana-Santero 的头像

Gabriel Ocana-Santero11 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve Crazy, meanwhile we have to tiptoe around their safeguards to use them for therapeutic discovery 😑

Aditya kumar Singh 的头像

Aditya kumar Singh10 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve @chooi_jeq becoming popular day by day really cool

Saurabh Mishra 的头像

Saurabh Mishra10 天前

@GadiNBC @NBCNews @sebwarb1 @robocurve the household cleaners example makes the risk concrete

相关视频

Thanks Gadi Schwartz and NBC News for having me on to talk about The Fermi Explorer Mission to another star! 🚀🚀

Thanks Gadi Schwartz and NBC News for having me on to talk about The Fermi Explorer Mission to another star! 🚀🚀

Philip Johnston

12,516 次观看 • 1 个月前

Today, we're launching Belvedir We're making private AI models cheap and easy so the world can have trillions of them. Private AI models are cheaper, better, faster, and protect your data from frontier labs, but they're still to expensive and difficult to make for startups and consumers. Belvedir makes private AI models autonomously with less than 5 minutes of setup, so that private AI can finally be accessible to everyone

Today, we're launching Belvedir We're making private AI models cheap and easy so the world can have trillions of them. Private AI models are cheaper, better, faster, and protect your data from frontier labs, but they're still to expensive and difficult to make for startups and consumers. Belvedir makes private AI models autonomously with less than 5 minutes of setup, so that private AI can finally be accessible to everyone

Zachary Yu

89,359 次观看 • 1 个月前

One of the biggest surprises in AI over the last few years has been how well coding agents generalize beyond software. In a recent essay, MIT professor Philip Isola argued that we may be entering the era of robot-use agents: general-purpose models that can control different robots, write policies, and learn new physical tasks with little or no robot-specific training. In this episode of Decoded, we're joined by the founders of Waddle Labs and Robocurve PBC, two of the startups whose work helped drive this realization. They're working at the frontier of using general-purpose models to control robots, and their recent demos helped inspire the growing conversation around robot-use agents. Together, we dig into the research behind that idea, from code-as-policies and vision-language-action models to the harnesses and evals needed to make these systems work in the real world. 00:00 — What Are Robot-Use Agents? 02:11 — From Vision-Language-Action Models to General LLMs 04:07 — The Bitter Lesson for Robotics 07:04 — How Coding Agents Learned to Control Robots 10:41 — How Robots Learn From Experience 14:22 — The Harness as a Form of Robot Intelligence 16:14 — Watching Astra Control a Robot 20:55 — Why General Models May Win in Robotics 26:05 — How Close Are General-Purpose Robots?

One of the biggest surprises in AI over the last few years has been how well coding agents generalize beyond software. In a recent essay, MIT professor Philip Isola argued that we may be entering the era of robot-use agents: general-purpose models that can control different robots, write policies, and learn new physical tasks with little or no robot-specific training. In this episode of Decoded, we're joined by the founders of Waddle Labs and Robocurve PBC, two of the startups whose work helped drive this realization. They're working at the frontier of using general-purpose models to control robots, and their recent demos helped inspire the growing conversation around robot-use agents. Together, we dig into the research behind that idea, from code-as-policies and vision-language-action models to the harnesses and evals needed to make these systems work in the real world. 00:00 — What Are Robot-Use Agents? 02:11 — From Vision-Language-Action Models to General LLMs 04:07 — The Bitter Lesson for Robotics 07:04 — How Coding Agents Learned to Control Robots 10:41 — How Robots Learn From Experience 14:22 — The Harness as a Form of Robot Intelligence 16:14 — Watching Astra Control a Robot 20:55 — Why General Models May Win in Robotics 26:05 — How Close Are General-Purpose Robots?

Y Combinator

91,311 次观看 • 7 天前

Robots can make building tall structures safer. 🧱 The fastening robot from Raise Robotics helps put up walls on high buildings without putting workers in danger. It uses robots to do the risky jobs people used to do. This means fewer accidents and injuries on construction sites. Workers can stay on the ground and control the robot from a safe place. My fear of height just got a fear of height! 🫣 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Robots can make building tall structures safer. 🧱 The fastening robot from Raise Robotics helps put up walls on high buildings without putting workers in danger. It uses robots to do the risky jobs people used to do. This means fewer accidents and injuries on construction sites. Workers can stay on the ground and control the robot from a safe place. My fear of height just got a fear of height! 🫣 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

33,821 次观看 • 3 天前

Andrea Mitchell shared today that after the inauguration in January, she has decided to shift from her daily anchoring duties for “Andrea Mitchell Reports” on MSNBC to a broader role with NBCU News Group. Andrea Mitchell will remain NBC News’ Chief Foreign Affairs Correspondent and Chief Washington Correspondent, and viewers will continue to see her during major breaking news events and on big political nights. Her reporting will continue to appear across the network’s flagship news broadcasts, including TODAY, NBC Nightly News, and Meet the Press, as well as NBC News NOW, and MSNBC. “Andrea Mitchell Reports” launched in 2008 and is MSNBC’s longest-running dayside program. It became the blueprint for MSNBC’s future “Reports” and dayside programming.

Andrea Mitchell shared today that after the inauguration in January, she has decided to shift from her daily anchoring duties for “Andrea Mitchell Reports” on MSNBC to a broader role with NBCU News Group. Andrea Mitchell will remain NBC News’ Chief Foreign Affairs Correspondent and Chief Washington Correspondent, and viewers will continue to see her during major breaking news events and on big political nights. Her reporting will continue to appear across the network’s flagship news broadcasts, including TODAY, NBC Nightly News, and Meet the Press, as well as NBC News NOW, and MSNBC. “Andrea Mitchell Reports” launched in 2008 and is MSNBC’s longest-running dayside program. It became the blueprint for MSNBC’s future “Reports” and dayside programming.

MS NOW Reports

167,775 次观看 • 1 年前

FOR THE FIRST TIME, THE US GOVERNMENT PULLED A PUBLICLY DEPLOYED AI MODEL OFFLINE. THE MARKET SHOULD PAY ATTENTION. 🇺🇸 On June 12, the Commerce Department ordered Anthropic to block access to its newest models, Fable 5 and Mythos 5, for any foreign national. Anthropic had to disable them for everyone to comply. The stated reason was a security concern. Anthropic disputes it, calling the flagged technique a narrow jailbreak available in other models too. Whatever the truth, here's the macro takeaway: Frontier AI now sits close enough to national security that a government can switch it off overnight. If you don't control your own models and compute, you're one directive away from losing access. That's the case for open and local models, in one headline.

FOR THE FIRST TIME, THE US GOVERNMENT PULLED A PUBLICLY DEPLOYED AI MODEL OFFLINE. THE MARKET SHOULD PAY ATTENTION. 🇺🇸 On June 12, the Commerce Department ordered Anthropic to block access to its newest models, Fable 5 and Mythos 5, for any foreign national. Anthropic had to disable them for everyone to comply. The stated reason was a security concern. Anthropic disputes it, calling the flagged technique a narrow jailbreak available in other models too. Whatever the truth, here's the macro takeaway: Frontier AI now sits close enough to national security that a government can switch it off overnight. If you don't control your own models and compute, you're one directive away from losing access. That's the case for open and local models, in one headline.

Merlijn The Trader

48,010 次观看 • 3 个月前

We asked the CEO of HuggingFace clem 🤗 what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. "Six, seven years ago, at the time it was GPT-2, and there was already a lot of people saying that it was too dangerous to release in open source." "Mythos, when it was announced was crazy dangerous... In a few weeks or a few months, everyone is gonna be using Mythos, and not destroy the world as a result." "For cybersecurity, the biggest risk is that a few players have capabilities that other people don't have... If you make it more open, it's usually easier for defenders to react and make the whole system safer." "The idea of restricting a technology like AI based on risks is like saying, 'Some people can punch other people, so let's tie down everybody's hands.'" "Otherwise you slow down progress, you create massive gaps in terms of controls, in terms of capabilities, and you create actually additional risks."

We asked the CEO of HuggingFace clem 🤗 what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. "Six, seven years ago, at the time it was GPT-2, and there was already a lot of people saying that it was too dangerous to release in open source." "Mythos, when it was announced was crazy dangerous... In a few weeks or a few months, everyone is gonna be using Mythos, and not destroy the world as a result." "For cybersecurity, the biggest risk is that a few players have capabilities that other people don't have... If you make it more open, it's usually easier for defenders to react and make the whole system safer." "The idea of restricting a technology like AI based on risks is like saying, 'Some people can punch other people, so let's tie down everybody's hands.'" "Otherwise you slow down progress, you create massive gaps in terms of controls, in terms of capabilities, and you create actually additional risks."

MTS

318,827 次观看 • 4 个月前

The biggest AI companies may be using safety to lock everyone else out of the future (Save this). David Sacks believes that warnings about catastrophic AI risks will build public support for a powerful federal regulator. That regulator may never explicitly ban open source AI. Instead, it could require every advanced model to remain continuously monitored, centrally controlled and capable of being withdrawn. Closed models such as Claude could satisfy those requirements because they operate on company controlled servers. Open weight models would struggle to comply because anyone can download, copy and modify their underlying weights. Once those weights are publicly released, the developer cannot recall every copy, monitor every user or guarantee that its original safeguards remain intact. Anthropic identifies this irreversibility as a legitimate security concern. Applying identical rules to open and closed models could therefore produce very unequal consequences. Anthropic and OpenAI could afford expensive testing, licensing and monitoring requirements, while startups and independent developers might be unable to comply. The eventual result could be a government protected oligopoly dominated by a few closed model companies. There is evidence supporting part of Sacks’ concern. Anthropic advocates mandatory pre-release testing for every sufficiently powerful model, whether it is open or closed, with evaluations focused on cyber, biological and alignment risks. However, Anthropic explicitly denies supporting a blanket ban on open weight models. The company describes open models without dangerous capabilities as a public good and argues that regulation should depend on demonstrated capabilities rather than whether a model is open or closed. Sacks’ strongest argument is therefore about regulatory consequences because safety organizations could receive stronger protections, politicians could acquire greater authority and dominant AI companies could gain an expensive compliance moat. And open source competitors could gradually be eliminated without the government ever formally announcing a ban.

The biggest AI companies may be using safety to lock everyone else out of the future (Save this). David Sacks believes that warnings about catastrophic AI risks will build public support for a powerful federal regulator. That regulator may never explicitly ban open source AI. Instead, it could require every advanced model to remain continuously monitored, centrally controlled and capable of being withdrawn. Closed models such as Claude could satisfy those requirements because they operate on company controlled servers. Open weight models would struggle to comply because anyone can download, copy and modify their underlying weights. Once those weights are publicly released, the developer cannot recall every copy, monitor every user or guarantee that its original safeguards remain intact. Anthropic identifies this irreversibility as a legitimate security concern. Applying identical rules to open and closed models could therefore produce very unequal consequences. Anthropic and OpenAI could afford expensive testing, licensing and monitoring requirements, while startups and independent developers might be unable to comply. The eventual result could be a government protected oligopoly dominated by a few closed model companies. There is evidence supporting part of Sacks’ concern. Anthropic advocates mandatory pre-release testing for every sufficiently powerful model, whether it is open or closed, with evaluations focused on cyber, biological and alignment risks. However, Anthropic explicitly denies supporting a blanket ban on open weight models. The company describes open models without dangerous capabilities as a public good and argues that regulation should depend on demonstrated capabilities rather than whether a model is open or closed. Sacks’ strongest argument is therefore about regulatory consequences because safety organizations could receive stronger protections, politicians could acquire greater authority and dominant AI companies could gain an expensive compliance moat. And open source competitors could gradually be eliminated without the government ever formally announcing a ban.

Milk Road AI

57,220 次观看 • 21 天前

I’ve been working on GTA6: Shenzhen Lite lately. Mainly using GPT-6 Astra for development, with some help from Fable 5.1. For assets like cars, tanks, and maid characters, I generate images with Image-2.5 first, then turn them into 3D models with Tripo AI. I’ve spent 100+ hours on it so far, and it’s turning out pretty well. I posted it on Douyin, Rednote, and Bilibili, and it blew up on all three. There’s still a lot left to do, but I’ve open-sourced it anyway. Could use some help from the community #TripoAstra #Tripo3D #TripoAmbassador

I’ve been working on GTA6: Shenzhen Lite lately. Mainly using GPT-6 Astra for development, with some help from Fable 5.1. For assets like cars, tanks, and maid characters, I generate images with Image-2.5 first, then turn them into 3D models with Tripo AI. I’ve spent 100+ hours on it so far, and it’s turning out pretty well. I posted it on Douyin, Rednote, and Bilibili, and it blew up on all three. There’s still a lot left to do, but I’ve open-sourced it anyway. Could use some help from the community #TripoAstra #Tripo3D #TripoAmbassador

地球人

32,491 次观看 • 16 天前

When seconds matter, perspective is everything. 🚓 Aerial support helps us understand what is happening in real time. It helps guide officers, coordinate response, and find people quickly. Using cameras with Infrared (IR) Thermal Imaging, we can spot things that are hard to see from the ground. Making things safer for both officers and the public. These tools help us make better decisions, reduce risks and respond faster. It’s just one of the many ways we are working to keep our community safe. For more information, visit 🛩️

When seconds matter, perspective is everything. 🚓 Aerial support helps us understand what is happening in real time. It helps guide officers, coordinate response, and find people quickly. Using cameras with Infrared (IR) Thermal Imaging, we can spot things that are hard to see from the ground. Making things safer for both officers and the public. These tools help us make better decisions, reduce risks and respond faster. It’s just one of the many ways we are working to keep our community safe. For more information, visit 🛩️

Ottawa Police

10,459 次观看 • 5 个月前

For a long time, I said no to interviews. I don’t particularly enjoy being on camera and, if I’m honest, I’ve always been much happier staying anonymous and letting the footage speak for itself. But things changed after what I’ve witnessed on the French coast. I’ve realised that having a person there to anchor the footage gives it a human side and helps explain what I’m seeing on the ground. Thanks again to Patrick Christys and GB News for having me on. Hopefully plenty more footage coming daily from the French coast 👍🏻

For a long time, I said no to interviews. I don’t particularly enjoy being on camera and, if I’m honest, I’ve always been much happier staying anonymous and letting the footage speak for itself. But things changed after what I’ve witnessed on the French coast. I’ve realised that having a person there to anchor the footage gives it a human side and helps explain what I’m seeing on the ground. Thanks again to Patrick Christys and GB News for having me on. Hopefully plenty more footage coming daily from the French coast 👍🏻

The Nowhere Photographer

65,998 次观看 • 1 个月前

A round of applause is needed for this excellent video from NBC News on Samidoun. I was more than happy to contribute some of my recordings when they reached out to me. Reporting like this will hopefully educate the public on insidious non-profits and organizations harming people worldwide. Please go like and comment on their Youtube video as well. This was so well done, and I am happy to be a small part of it.

A round of applause is needed for this excellent video from NBC News on Samidoun. I was more than happy to contribute some of my recordings when they reached out to me. Reporting like this will hopefully educate the public on insidious non-profits and organizations harming people worldwide. Please go like and comment on their Youtube video as well. This was so well done, and I am happy to be a small part of it.

Stu Smith

95,877 次观看 • 2 年前

MICROSOFT 🔥: Early look at Microsoft Scout Agent for Microsoft Frontier users. Scout agent is designed for work use cases and was recently revealed at Microsoft Build 2026. Features 👀 > Models from OpenAI and Anthropic, including GPT-5.5 and Claude Opus 4.7 > Multi-step automations with support for browser use. > Co-Create, an open canvas where users can collaborate with AI and export outputs as documents. > Deep integration with Microsoft Teams > Daily Briefing that summarises work context across connected services. > Skills support, document generation, coding capabilities, and all that. > Both macOS and Windows apps are available. It would be quite a decent tool for Windows users! Yet, it is only for Frontier. Looking forward to testing their super app too.

MICROSOFT 🔥: Early look at Microsoft Scout Agent for Microsoft Frontier users. Scout agent is designed for work use cases and was recently revealed at Microsoft Build 2026. Features 👀 > Models from OpenAI and Anthropic, including GPT-5.5 and Claude Opus 4.7 > Multi-step automations with support for browser use. > Co-Create, an open canvas where users can collaborate with AI and export outputs as documents. > Deep integration with Microsoft Teams > Daily Briefing that summarises work context across connected services. > Skills support, document generation, coding capabilities, and all that. > Both macOS and Windows apps are available. It would be quite a decent tool for Windows users! Yet, it is only for Frontier. Looking forward to testing their super app too.

🚨 AI News | TestingCatalog

38,898 次观看 • 4 个月前

Shoe0nHead says her "career" for a decade has been telling the left that woke is toxic and it drives people away from the left and the Democratic Party. "And for 10 years, I was called a racist, sexist, pick-me, fascist, nazi. Every 'ist' and 'ism' under the sun. And now they're just conceding to that, and acting like it never happened. And you just expect us to, what, move on? I've been like a dog chasing a car, and I finally got the car, and I don't know what to do with it."

Shoe0nHead says her "career" for a decade has been telling the left that woke is toxic and it drives people away from the left and the Democratic Party. "And for 10 years, I was called a racist, sexist, pick-me, fascist, nazi. Every 'ist' and 'ism' under the sun. And now they're just conceding to that, and acting like it never happened. And you just expect us to, what, move on? I've been like a dog chasing a car, and I finally got the car, and I don't know what to do with it."

Post-Left Watch (postleftwatch.bsky.social)

58,237 次观看 • 1 个月前

Sahil Bloom reveals you can make $100,000/month doing AI consulting for businesses “You could be making an enormous amount of money by consulting directly with companies that have no idea how to use AI or implement it into their workflow.” “I think that a young person that understands AI, and like has a pretty decent understanding of some of the models that are out there and the capabilities, can go build a $100,000 a month business doing that.”

Sahil Bloom reveals you can make $100,000/month doing AI consulting for businesses “You could be making an enormous amount of money by consulting directly with companies that have no idea how to use AI or implement it into their workflow.” “I think that a young person that understands AI, and like has a pretty decent understanding of some of the models that are out there and the capabilities, can go build a $100,000 a month business doing that.”

The Iced Coffee Hour

18,669 次观看 • 4 个月前

Emily Blunt says Spielberg offered to use AI for the alien clicking sounds in ‘DISCLOSURE DAY’ "Steven and I discussed it and he said ‘We could do it with AI or you could do it’ and I said ‘I feel confident I can make some weird noises.’ “And I was sending him voice memos of me clicking, humming mixed with morse code sounds… I just tried everything." But Spielberg then said: “I would never use AI even if Emily was unable to produce the sounds that she did produce.”

Emily Blunt says Spielberg offered to use AI for the alien clicking sounds in ‘DISCLOSURE DAY’ "Steven and I discussed it and he said ‘We could do it with AI or you could do it’ and I said ‘I feel confident I can make some weird noises.’ “And I was sending him voice memos of me clicking, humming mixed with morse code sounds… I just tried everything." But Spielberg then said: “I would never use AI even if Emily was unable to produce the sounds that she did produce.”

Emir Han

38,435 次观看 • 1 个月前

i gave the exact same prompt to Opus 5 and Fable 5 and here's what came out Opus 5 dropped yesterday and my feed was instantly full of comparison posts but i wanted to check it myself so i gave it one task build 3d Flappy Bird first thing everyone was saying Opus is 2x cheaper than Fable BUT.. building the game on Max effort cost me 250k tokens and 1 hour on Opus and 130k on Fable but Fable did it in half the time Opus gave me volume control plus 3 separate graphics presets to choose from, Fable gave none of that in Opus the environment shifts dynamically as you play in Fable it's basically the same frame the whole run maybe a day/night change here and there graphics, animations, glow, all of it goes to Opus and now that limits jumped 50% for the near future while Fable went paid for me the call is obvious right now Opus same prompt start to finish zero changes to code or commands want it? i can send it over

i gave the exact same prompt to Opus 5 and Fable 5 and here's what came out Opus 5 dropped yesterday and my feed was instantly full of comparison posts but i wanted to check it myself so i gave it one task build 3d Flappy Bird first thing everyone was saying Opus is 2x cheaper than Fable BUT.. building the game on Max effort cost me 250k tokens and 1 hour on Opus and 130k on Fable but Fable did it in half the time Opus gave me volume control plus 3 separate graphics presets to choose from, Fable gave none of that in Opus the environment shifts dynamically as you play in Fable it's basically the same frame the whole run maybe a day/night change here and there graphics, animations, glow, all of it goes to Opus and now that limits jumped 50% for the near future while Fable went paid for me the call is obvious right now Opus same prompt start to finish zero changes to code or commands want it? i can send it over

savip.

45,764 次观看 • 2 个月前

We finally have a product that helps with frontier AI research. Frontier LLMs are a) nerfed at AI research b) lack immediate GPU availability and c) lack access to the right skills and research material. "Teach me how vision models see" Goodfire Silico takes a vision transformer (ViT), loads ImageNet images, and for each head, visualizes the 196x196 matrix by creating animated arcs across patches. Haven't found another model or product that can do this level of comprehensive AI research, much less oneshot it.

We finally have a product that helps with frontier AI research. Frontier LLMs are a) nerfed at AI research b) lack immediate GPU availability and c) lack access to the right skills and research material. "Teach me how vision models see" Goodfire Silico takes a vision transformer (ViT), loads ImageNet images, and for each head, visualizes the 196x196 matrix by creating animated arcs across patches. Haven't found another model or product that can do this level of comprehensive AI research, much less oneshot it.

Deedy

74,579 次观看 • 2 个月前

Imagine having an idea for a space game… and actually being able to start building it . I wanted to test what an AI-powered game creation workflow could look like. MeshyAI helped create spacecraft models, and GPT 6 Astra helped bring those assets into an interactive space environment. The coolest part for me is seeing individual 3D models become part of something bigger. A ship is cool. A whole universe built around those ships is even cooler. #Astra #MeshyAI #GameDev #AI3D

Imagine having an idea for a space game… and actually being able to start building it . I wanted to test what an AI-powered game creation workflow could look like. MeshyAI helped create spacecraft models, and GPT 6 Astra helped bring those assets into an interactive space environment. The coolest part for me is seeing individual 3D models become part of something bigger. A ship is cool. A whole universe built around those ships is even cooler. #Astra #MeshyAI #GameDev #AI3D

Isabella

41,565 次观看 • 21 天前