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。

1/ Can molecular AI move past hard-coded Graph Neural Networks and embrace scalable Transformers that discover molecular structure on their own? We demonstrate that you can train a 1B parameter Transformer model without any graph priors or physical inductive biases. And surprisingly, not only can it maintain competitive performance... under equal compute on the Open Molecules 2025 dataset… it’s faster than a 6M parameter equivariant GNN, and exhibits scaling laws that don’t saturate. We use this as a starting point to investigate emergent internal representations, and find that it adaptively discovers molecular structure! Check out the interactive demo on our website: And our paper: In collaboration with Toby Kreiman, Yutong Bai, Fadi, Elizabeth, and Eric Qu. Here’s a video showing how the Transformer learns distance-aware attention patterns (purple gradient) that adapt to atomic environments 👇show more

Aditi Krishnapriyan

1,069 subscribers

39,282 次观看 • 9 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Transformer-based neural networks achieve impressive performance on coding, math & reasoning tasks that require keeping track of variables and their values. But how can they do that without explicit memory? 📄 Our new ICML paper investigates this in a synthetic setting! 🧵 1/13

Transformer-based neural networks achieve impressive performance on coding, math & reasoning tasks that require keeping track of variables and their values. But how can they do that without explicit memory? 📄 Our new ICML paper investigates this in a synthetic setting! 🧵 1/13

Raphaël Millière

73,151 次观看 • 1 年前

Introducing Oasis: the first playable AI-generated game. We partnered with Decart to build a real-time, interactive world model that runs >10x faster on Sohu. We're open-sourcing the model architecture, weights, and research. Here's how it works (and a demo you can play!):

Introducing Oasis: the first playable AI-generated game. We partnered with Decart to build a real-time, interactive world model that runs >10x faster on Sohu. We're open-sourcing the model architecture, weights, and research. Here's how it works (and a demo you can play!):

Etched

24,648,196 次观看 • 1 年前

Announcing the newest releases from Meta FAIR. We’re releasing new groundbreaking models, benchmarks, and datasets that will transform the way researchers approach molecular property prediction, language processing, and neuroscience. 1️⃣ Open Molecules 2025 (OMol25): A dataset for molecular discovery with simulations of large atomic systems. 2️⃣ Universal Model for Atoms: A machine learning interatomic potential for modeling atom interactions across a wide range of materials and molecules. 3️⃣ Adjoint Sampling: A scalable algorithm for training generative models based on scalar rewards. 4️⃣ FAIR and the Rothschild Foundation Hospital partnered on a large-scale study that reveals striking parallels between language development in humans and LLMs. Read more ➡️

Announcing the newest releases from Meta FAIR. We’re releasing new groundbreaking models, benchmarks, and datasets that will transform the way researchers approach molecular property prediction, language processing, and neuroscience. 1️⃣ Open Molecules 2025 (OMol25): A dataset for molecular discovery with simulations of large atomic systems. 2️⃣ Universal Model for Atoms: A machine learning interatomic potential for modeling atom interactions across a wide range of materials and molecules. 3️⃣ Adjoint Sampling: A scalable algorithm for training generative models based on scalar rewards. 4️⃣ FAIR and the Rothschild Foundation Hospital partnered on a large-scale study that reveals striking parallels between language development in humans and LLMs. Read more ➡️

AI at Meta

899,456 次观看 • 1 年前

Here's another video of emergent behaviour from Instant Policy, which we released yesterday. You will see that when a demo is provided only for Box 1, the robot fails to generalise well to Boxes 2 and 3. But then we add in a demo for Box 4, and the robot can now successfully generalise to Boxes 2 and 3. One of the exciting features of Instant Policy is that, since we encode everything in a graph neural network, we can keep on adding as many demonstrations as we like until we have covered the required task diversity.

Here's another video of emergent behaviour from Instant Policy, which we released yesterday. You will see that when a demo is provided only for Box 1, the robot fails to generalise well to Boxes 2 and 3. But then we add in a demo for Box 4, and the robot can now successfully generalise to Boxes 2 and 3. One of the exciting features of Instant Policy is that, since we encode everything in a graph neural network, we can keep on adding as many demonstrations as we like until we have covered the required task diversity.

Edward Johns

12,418 次观看 • 1 年前

Can we scale 4D pretraining to learn general space-time representations that reconstruct an object from a few views at any time to any view at any other time? Introducing 4D-LRM: a Large Space-Time Reconstruction Model that ... 🔹 Predicts 4D Gaussian primitives directly from multi-view tokens (no motion vectors, no HexPlane); 🔹 Uses a clean, minimal Transformer backbone; 🔹 Generalizes with fast, high-quality feedforward rendering at any view and infinite frame rate. Check out more interactive demos and scaling behaviors on our homepage/paper. 👉Website: 👉Paper:

Can we scale 4D pretraining to learn general space-time representations that reconstruct an object from a few views at any time to any view at any other time? Introducing 4D-LRM: a Large Space-Time Reconstruction Model that ... 🔹 Predicts 4D Gaussian primitives directly from multi-view tokens (no motion vectors, no HexPlane); 🔹 Uses a clean, minimal Transformer backbone; 🔹 Generalizes with fast, high-quality feedforward rendering at any view and infinite frame rate. Check out more interactive demos and scaling behaviors on our homepage/paper. 👉Website: 👉Paper:

Martin Ziqiao Ma

21,811 次观看 • 1 年前

Arteta on towels: “You cannot use someone else’s towel? I didn’t know. Like somebody else’s drink - I’ve done it as a player. You see a drink from the doctor and nicked it and had a drink and carry on. I don’t know. I think we need to clarity on that on what we can and cannot do and the time we can consume on that probably. On bringing a special sway day towel: “It depends where you leave it, so you turn and the towel goes away so maybe we have to carry on or design something that we can that with the ball. Let’s see.”

Arteta on towels: “You cannot use someone else’s towel? I didn’t know. Like somebody else’s drink - I’ve done it as a player. You see a drink from the doctor and nicked it and had a drink and carry on. I don’t know. I think we need to clarity on that on what we can and cannot do and the time we can consume on that probably. On bringing a special sway day towel: “It depends where you leave it, so you turn and the towel goes away so maybe we have to carry on or design something that we can that with the ball. Let’s see.”

Connor Humm

323,369 次观看 • 8 个月前

Dr. Masaru Emoto demonstrated that human thoughts, emotions and intentions alter the molecular structure of water, proving that human consciousness has an effect on physical matter.

Dr. Masaru Emoto demonstrated that human thoughts, emotions and intentions alter the molecular structure of water, proving that human consciousness has an effect on physical matter.

Red Pill Dispenser

17,387 次观看 • 1 年前

made a big breakthrough on tinynodes! this is a huge .blend file with 679 nodes, running 100% in the browser... and it's fully dynamic! at the end of the video, you can see the raw geometry and the node graph that describes how to turn it into this complex wooden structure

made a big breakthrough on tinynodes! this is a huge .blend file with 679 nodes, running 100% in the browser... and it's fully dynamic! at the end of the video, you can see the raw geometry and the node graph that describes how to turn it into this complex wooden structure

Seth Thompson

36,110 次观看 • 1 个月前

On net, is our biological hardware a help or hindrance to our intelligence? Are GPUs are to brains what fingers are to robot motors, and there’s a bunch of cognitive dexterity that our brain is uniquely good at implementing? Ala Morevac’s paradox, evolution has spent way more time optimizing nanotech than the semiconductor industry has. Or maybe it's something structural: the brain can do unstructured sparsity, and can colocate memory and compute. Or is it the other way around, because with digital computers, we can get rid of all these biological handicaps: The brain operates on a 20 W budget (and consequently runs at only 200 Hz speed in order to accommodate small voltage gaps). And it needs to do all this crazy intense molecular machinery in order to do things that take trivial amounts of code in computers: for example, update this parameter to this value. Net net, does current biological or digital computing hardware win out? I asked Adam Marblestone:

On net, is our biological hardware a help or hindrance to our intelligence? Are GPUs are to brains what fingers are to robot motors, and there’s a bunch of cognitive dexterity that our brain is uniquely good at implementing? Ala Morevac’s paradox, evolution has spent way more time optimizing nanotech than the semiconductor industry has. Or maybe it's something structural: the brain can do unstructured sparsity, and can colocate memory and compute. Or is it the other way around, because with digital computers, we can get rid of all these biological handicaps: The brain operates on a 20 W budget (and consequently runs at only 200 Hz speed in order to accommodate small voltage gaps). And it needs to do all this crazy intense molecular machinery in order to do things that take trivial amounts of code in computers: for example, update this parameter to this value. Net net, does current biological or digital computing hardware win out? I asked Adam Marblestone:

Dwarkesh Patel

26,451 次观看 • 6 个月前

Our vision is for AI that uses world models to adapt in new and dynamic environments and efficiently learn new skills. We’re sharing V-JEPA 2, a new world model with state-of-the-art performance in visual understanding and prediction. V-JEPA 2 is a 1.2 billion-parameter model, trained on video, that can enable zero-shot planning in robots—allowing them to plan and execute tasks in unfamiliar environments. Learn more about V-JEPA 2 ➡️ As we continue working toward our goal of achieving advanced machine intelligence (AMI), we’re also releasing three new benchmarks for evaluating how well existing models can reason about the physical world from video. Learn more and download the new benchmarks ➡️

Our vision is for AI that uses world models to adapt in new and dynamic environments and efficiently learn new skills. We’re sharing V-JEPA 2, a new world model with state-of-the-art performance in visual understanding and prediction. V-JEPA 2 is a 1.2 billion-parameter model, trained on video, that can enable zero-shot planning in robots—allowing them to plan and execute tasks in unfamiliar environments. Learn more about V-JEPA 2 ➡️ As we continue working toward our goal of achieving advanced machine intelligence (AMI), we’re also releasing three new benchmarks for evaluating how well existing models can reason about the physical world from video. Learn more and download the new benchmarks ➡️

AI at Meta

309,942 次观看 • 1 年前

Israelis on the left and right are divided on many issues, but at least one thing unites them. The belief that Palestinians hate them and will continue to hate them, and that therefore they cannot be trusted. It’s not just Bibi. The majority, even on the left, are deeply sceptical. This is a tragedy. As Palestinians, we have the power to shatter this stereotype. To show Israelis that trust is possible, that coexistence isn’t a dream, that we don’t harbour any ill will. And once we do that, we will see those sentiments reciprocated. We need a reset. And honestly I don’t think it’s as difficult as people make out. Once we gain each other’s trust, we can move ahead with a political resolution that will end this conflict once and for all. We have a history, I’m well aware. But it needn’t determine our future.

Israelis on the left and right are divided on many issues, but at least one thing unites them. The belief that Palestinians hate them and will continue to hate them, and that therefore they cannot be trusted. It’s not just Bibi. The majority, even on the left, are deeply sceptical. This is a tragedy. As Palestinians, we have the power to shatter this stereotype. To show Israelis that trust is possible, that coexistence isn’t a dream, that we don’t harbour any ill will. And once we do that, we will see those sentiments reciprocated. We need a reset. And honestly I don’t think it’s as difficult as people make out. Once we gain each other’s trust, we can move ahead with a political resolution that will end this conflict once and for all. We have a history, I’m well aware. But it needn’t determine our future.

Samer Sinijlawiسامر السنجلاوي

96,705 次观看 • 10 个月前

Data is 👑 Our foundation model learns from globally diverse, unlabelled data. Without telling it how to drive or naming different scene elements, our AI learns patterns and behaviours. Our results this week demonstrate how rapidly our AI can learn new behaviours, finding that with hundreds of hours worth of country-specific data, our foundation model learns new skills, like driving on the right in the US! See how our Wayve AI Driver improves driving in the US with data from 0-10-100-250-500 hours worth of additional US-specific data. 👇 By leveraging diverse data, we are building a highly generalizable AI that enables safe, adaptable, and scalable autonomy—we’ve taken it to the US and Germany but stay tuned for where is next 🌎⁉️

Data is 👑 Our foundation model learns from globally diverse, unlabelled data. Without telling it how to drive or naming different scene elements, our AI learns patterns and behaviours. Our results this week demonstrate how rapidly our AI can learn new behaviours, finding that with hundreds of hours worth of country-specific data, our foundation model learns new skills, like driving on the right in the US! See how our Wayve AI Driver improves driving in the US with data from 0-10-100-250-500 hours worth of additional US-specific data. 👇 By leveraging diverse data, we are building a highly generalizable AI that enables safe, adaptable, and scalable autonomy—we’ve taken it to the US and Germany but stay tuned for where is next 🌎⁉️

Alex Kendall

17,178 次观看 • 1 年前

This dude nails it and explains how AI can only “threaten” or “replace” people without consciousness. And I’m starting to think the people that don’t understand this point are revealing that they may not have consciousness. It’s actually projection and someone arrogant to think that everyone has consciousness just because you do.

This dude nails it and explains how AI can only “threaten” or “replace” people without consciousness. And I’m starting to think the people that don’t understand this point are revealing that they may not have consciousness. It’s actually projection and someone arrogant to think that everyone has consciousness just because you do.

Owen Benjamin 🐻

31,016 次观看 • 11 个月前

Early on, Dame Daphne Sheldrick discovered that the mothering of an orphaned elephant needs to be shared. Elephants can form over-attachments to any one individual and if their presence is missed, and orphan can fall into a deep depression and rapidly lose condition. It's why we have a whole team of caregivers caring for the orphans (that and it takes many hands to feed the orphans). It gives our Keepers extended time at home with their own families, and it means every orphan comes to know every Keeper, so that any one Keeper's absence is understood as a break, not another loss. We believe if we work together, we can have a bigger impact for wildlife. Find out how you can play a role:

Early on, Dame Daphne Sheldrick discovered that the mothering of an orphaned elephant needs to be shared. Elephants can form over-attachments to any one individual and if their presence is missed, and orphan can fall into a deep depression and rapidly lose condition. It's why we have a whole team of caregivers caring for the orphans (that and it takes many hands to feed the orphans). It gives our Keepers extended time at home with their own families, and it means every orphan comes to know every Keeper, so that any one Keeper's absence is understood as a break, not another loss. We believe if we work together, we can have a bigger impact for wildlife. Find out how you can play a role:

Sheldrick Wildlife Trust

22,744 次观看 • 1 个月前

This is Cortical Cloud. Live neural networks that you can interact with and train! Now open to the public. What will you discover? Credits: Frank Yang and a big thank you and acknowledgement to the rest of the Cortical Labs team. Sign up for Cortical Cloud: Learn more about us: Check out our API: Check out our API Docs: Check out our Developer Guide: Join our Discord:

This is Cortical Cloud. Live neural networks that you can interact with and train! Now open to the public. What will you discover? Credits: Frank Yang and a big thank you and acknowledgement to the rest of the Cortical Labs team. Sign up for Cortical Cloud: Learn more about us: Check out our API: Check out our API Docs: Check out our Developer Guide: Join our Discord:

Cortical Labs

224,300 次观看 • 3 个月前

I read a lot of open-source code and that’s mostly how I learn to write better code. But it often takes a while to understand different projects their structure, how things are connected and GitHub’s UX isn’t that great for that. built "Gitvizz" to solve that exact problem it instantly turn your codebase into interactive graph so you can see how everything’s wired Check it out at - gitvizz. com ps: thats a real horse

I read a lot of open-source code and that’s mostly how I learn to write better code. But it often takes a while to understand different projects their structure, how things are connected and GitHub’s UX isn’t that great for that. built "Gitvizz" to solve that exact problem it instantly turn your codebase into interactive graph so you can see how everything’s wired Check it out at - gitvizz. com ps: thats a real horse

Adithya S K

35,195 次观看 • 8 个月前

Give your voice agents a face. Introducing Lemon Slice-2 – the world's first interactive talking AI video model Think of it as a face layer for your voice agents, available today in production as an API or an embeddable widget. What makes Lemon Slice-2 great: → Easy: All you need is 1 photo → Expressive: hands and full body motion → Any character and style: if it has a face, Lemon Slice-2 can animate it Under the hood: we trained a custom, 20B-parameter video diffusion transformer, streaming at 20fps on a single GPU. Infinite-length video generation with no error accumulation. Talk to the demo avatars on our homepage. Or, add your own interactive avatar to your website. Companies and creators are already building video agents that are: → Customer support agents who screen-share and show you where to click → Lead qualification agents that replace forms with real conversations → AI personal shoppers who try on clothes and guide purchases People don't engage with chat bubbles. They engage with faces. And now you can have a 24/7 spokesperson that’s the face of your brand. We’ve partnered with , Daily, and Modal to make this happen. As part of this launch, we’re giving away a guide: “How to Add a High-Converting Video Agent to Your Website in under 10 Minutes” Retweet and comment "LEMON SLICE" below and I'll send it over 🍋

Lina Colucci

1,160,649 次观看 • 6 个月前

A group of Turning Point USA high schoolers here in Alabama got to their high school early and prayed for the nation. While the left encourages our youth to riot and ditch class, we as Christians encourage them to put their faith in God. It's the only way that we can build strong men and women, and it's the only way that we can save our country. I'm very proud that Alabama gets it.

A group of Turning Point USA high schoolers here in Alabama got to their high school early and prayed for the nation. While the left encourages our youth to riot and ditch class, we as Christians encourage them to put their faith in God. It's the only way that we can build strong men and women, and it's the only way that we can save our country. I'm very proud that Alabama gets it.

Jared Hudson

59,055 次观看 • 4 个月前

This is how you can get programmatic access to any website. You need two things: 1. The URL of the website 2. A prompt specifying what you want to do on the site Mino is a neat platform that uses browser automation and AI-powered navigation to understand your prompt, open the website, and extract the information you want. This opens the doors to unlimited potential automations you could build using data from everyday websites that don't give you an API. Mino uses AI to go way further than you could get by using a regular web scraping tool: • It can handle dynamic JavaScript content • It can handle login walls • It can navigate through interactive booking flows • It adapts to different interfaces and layout changes • It can fill out forms automatically • It supports stealth browser mode In my experience, running Mino on relatively straightforward sites takes about 30-60 seconds to complete. The more times you run it, the faster it gets. I recorded a quick video to show you the platform. You can check their site here: Thanks to the team for the support, onboarding me on their platform, and the collaboration on this post.

This is how you can get programmatic access to any website. You need two things: 1. The URL of the website 2. A prompt specifying what you want to do on the site Mino is a neat platform that uses browser automation and AI-powered navigation to understand your prompt, open the website, and extract the information you want. This opens the doors to unlimited potential automations you could build using data from everyday websites that don't give you an API. Mino uses AI to go way further than you could get by using a regular web scraping tool: • It can handle dynamic JavaScript content • It can handle login walls • It can navigate through interactive booking flows • It adapts to different interfaces and layout changes • It can fill out forms automatically • It supports stealth browser mode In my experience, running Mino on relatively straightforward sites takes about 30-60 seconds to complete. The more times you run it, the faster it gets. I recorded a quick video to show you the platform. You can check their site here: Thanks to the team for the support, onboarding me on their platform, and the collaboration on this post.

Santiago

31,742 次观看 • 6 个月前

It’s absolutely wild that you can take this train to and from an airport, and you can fly and land in the United States and your only option will be rideshare, a bus that comes every 60 minutes or to rent a car. Are we genuinely not embarrassed?

It’s absolutely wild that you can take this train to and from an airport, and you can fly and land in the United States and your only option will be rideshare, a bus that comes every 60 minutes or to rent a car. Are we genuinely not embarrassed?

Hayden

3,605,765 次观看 • 9 个月前