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。

Google's Jeff Dean says current pre-training is passive: initialize a model, stream the internet past it, let it observe But models need to learn not just from data, but by acting, predicting, and choosing what to learn from next "we have this artificial distinction now between pre and post-training,... show more

Haider.

67,338 subscribers

55,694 次观看 • 3 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

"We’ve moved past it, but we have to learn from it. … We need some guys to dig in and stack some consistent games together." Hear from head coach Rick Tocchet following today's practice at Flyers Training Center.

"We’ve moved past it, but we have to learn from it. … We need some guys to dig in and stack some consistent games together." Hear from head coach Rick Tocchet following today's practice at Flyers Training Center.

Philadelphia Flyers

23,259 次观看 • 5 个月前

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:

Andrew Ng

125,146 次观看 • 1 年前

Most people have heard of big model smell, but what about pristine pre-training smell? Evan and Derry seem to think it exists.

Most people have heard of big model smell, but what about pristine pre-training smell? Evan and Derry seem to think it exists.

Arena.ai

12,704 次观看 • 3 个月前

In case you missed it, we recently launched "Post-training of LLMs," a short course where you'll: ✅ Understand when and why to use post-training methods like Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning. ✅ Learn the concepts underlying the three post-training methods of SFT, DPO, and Online RL, their common use-cases, and how to curate high-quality data to effectively train a model using each method. ✅ Download a pre-trained model and implement post-training pipelines to turn a base model into an instruct model, change the identity of a chat assistant, and improve a model’s math capabilities. Learn more and enroll for free:

In case you missed it, we recently launched "Post-training of LLMs," a short course where you'll: ✅ Understand when and why to use post-training methods like Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning. ✅ Learn the concepts underlying the three post-training methods of SFT, DPO, and Online RL, their common use-cases, and how to curate high-quality data to effectively train a model using each method. ✅ Download a pre-trained model and implement post-training pipelines to turn a base model into an instruct model, change the identity of a chat assistant, and improve a model’s math capabilities. Learn more and enroll for free:

DeepLearning.AI

16,771 次观看 • 1 年前

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

Physical Intelligence

1,182,649 次观看 • 6 个月前

Jensen just said he was reading a text from DeepMind’s Demis confirming it: pre-training + post-training scaling laws are fully intact, and Gemini 3 is the proof. A huge jump in quality exactly what Blackwell and Rubin were built for. This next curve is going to be wild. $NVDA

Jensen just said he was reading a text from DeepMind’s Demis confirming it: pre-training + post-training scaling laws are fully intact, and Gemini 3 is the proof. A huge jump in quality exactly what Blackwell and Rubin were built for. This next curve is going to be wild. $NVDA

Ben Pouladian

327,245 次观看 • 7 个月前

Ilya Sutskever says that data is the fossil fuel of AI and "we have achieved peak data" because "we have but one internet", so the age of pre-training AI on bigger and bigger compute clusters "will unquestionably end"

Ilya Sutskever says that data is the fossil fuel of AI and "we have achieved peak data" because "we have but one internet", so the age of pre-training AI on bigger and bigger compute clusters "will unquestionably end"

Tsarathustra

127,069 次观看 • 1 年前

Our course recommendation of the day is “Post-training of LLMs, ” where you’ll learn how to customize pre-trained language models using Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL). You'll learn when to use each method, how to curate training data, and implement them in code to shape model behavior effectively. Enroll here:

Our course recommendation of the day is “Post-training of LLMs, ” where you’ll learn how to customize pre-trained language models using Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL). You'll learn when to use each method, how to curate training data, and implement them in code to shape model behavior effectively. Enroll here:

DeepLearning.AI

29,369 次观看 • 9 个月前

Introducing the OpenRouter MCP, live model intelligence right inside your agent Your agent builds and ships, but when it comes to choosing the right model for the right job, it guesses from 6 month old training data Watch it pick, price, and test the right model:

Introducing the OpenRouter MCP, live model intelligence right inside your agent Your agent builds and ships, but when it comes to choosing the right model for the right job, it guesses from 6 month old training data Watch it pick, price, and test the right model:

OpenRouter

261,609 次观看 • 15 天前

We introduce Tagger Phase 2, bringing manual #data labeling and #data review into the #DeCorp ecosystem. With this iteration, we have completed the final components necessary for a fully integrated and decentralized cycle: 1) Data processing from scratch 2) Peer-to-peer data validation 3) Automated pre-training of AI copilot models 4) Quasi-autonomous active learning It is not long now before you may directly engage with this unprecedented and comprehensive system. See you soon.

We introduce Tagger Phase 2, bringing manual #data labeling and #data review into the #DeCorp ecosystem. With this iteration, we have completed the final components necessary for a fully integrated and decentralized cycle: 1) Data processing from scratch 2) Peer-to-peer data validation 3) Automated pre-training of AI copilot models 4) Quasi-autonomous active learning It is not long now before you may directly engage with this unprecedented and comprehensive system. See you soon.

Tagger

104,265 次观看 • 1 年前

We need to destroy the data centers and start over from a pre internet era!! MARC REBILLET

We need to destroy the data centers and start over from a pre internet era!! MARC REBILLET

SubwayTakes with Kareem Rahma

55,344 次观看 • 8 个月前

"We found a way to win the game but we need to learn from it" Keith Andrews reflects on Brentford's 4-3 win over Burnley on Saturday and what his side can learn from their performance.

"We found a way to win the game but we need to learn from it" Keith Andrews reflects on Brentford's 4-3 win over Burnley on Saturday and what his side can learn from their performance.

Sky Sports News

74,785 次观看 • 4 个月前

Demis Hassabis says the next AI leap is continual learning. Not train once and freeze, but learn from real experience, constantly. AlphaGo did it in games. Now comes the messy real world. Would you let an AI learn from you? 👀

Demis Hassabis says the next AI leap is continual learning. Not train once and freeze, but learn from real experience, constantly. AlphaGo did it in games. Now comes the messy real world. Would you let an AI learn from you? 👀

VraserX e/acc

19,202 次观看 • 5 个月前

The stuff nobody explains, this is the Freepik Creative Showcase 01 A space to learn from top creatives: their projects, their clients, their workflows. Not just what they did, but how and why it worked Who made it, how long it took, and full tutorial in thread 👇🏼

The stuff nobody explains, this is the Freepik Creative Showcase 01 A space to learn from top creatives: their projects, their clients, their workflows. Not just what they did, but how and why it worked Who made it, how long it took, and full tutorial in thread 👇🏼

Freepik

19,426 次观看 • 11 个月前

🚨 Jensen Huang says everyone panicked about the AI data when MOST training data was never REAL to begin with. Ilya Sutskever told the industry pre-training was over. "Ilya said, 'We're out of data,' or something like that. 'Pre-training is over,' or something like that," Huang says. "The industry panicked, you know, that this is the end of AI." "And of course, of course that's obviously not true. We're gonna keep on scaling the amount of data that we have to train with." "A lot of that data is probably gonna be synthetic." That's where the panic came from — synthetic data sounds like cheating. "Most of the data that we are training, that we teach each other with, inform each other with, is synthetic." "It's synthetic because it didn't come out of nature." "You created it. I'm consuming it. I modify it, augment it, I regenerate it, somebody else consumes it." The textbook in your hand is synthetic. The post you're reading is synthetic. The lecture you took is synthetic. Nature didn't make any of it. Humans did. AI just learned to do the same thing — faster. "Training is now limited by compute," Huang says. "Data is now limited by compute." The data wall wasn't a wall. It was a mirror. If you're new here, follow @AiEvolutio for the latest on ChatGPT, Claude, and the AI tools shaping how we work and create. — Jensen Huang ( NVIDIA ), NVIDIA CEO, on Lex Fridman's ( Lex Fridman ) podcast

🚨 Jensen Huang says everyone panicked about the AI data when MOST training data was never REAL to begin with. Ilya Sutskever told the industry pre-training was over. "Ilya said, 'We're out of data,' or something like that. 'Pre-training is over,' or something like that," Huang says. "The industry panicked, you know, that this is the end of AI." "And of course, of course that's obviously not true. We're gonna keep on scaling the amount of data that we have to train with." "A lot of that data is probably gonna be synthetic." That's where the panic came from — synthetic data sounds like cheating. "Most of the data that we are training, that we teach each other with, inform each other with, is synthetic." "It's synthetic because it didn't come out of nature." "You created it. I'm consuming it. I modify it, augment it, I regenerate it, somebody else consumes it." The textbook in your hand is synthetic. The post you're reading is synthetic. The lecture you took is synthetic. Nature didn't make any of it. Humans did. AI just learned to do the same thing — faster. "Training is now limited by compute," Huang says. "Data is now limited by compute." The data wall wasn't a wall. It was a mirror. If you're new here, follow @AiEvolutio for the latest on ChatGPT, Claude, and the AI tools shaping how we work and create. — Jensen Huang ( NVIDIA ), NVIDIA CEO, on Lex Fridman's ( Lex Fridman ) podcast

AI Evolution

15,565 次观看 • 1 个月前

🚨Anthropic CEO was just asked: “Is pre-training dead?” > Claude 4 embodies advances in both pre-training and post-training > pre-training scaling laws work the way they’ve worked before > we are going to continue to scale up > the timeline is accelerating > 1-2 years until these models become peers to us Good. Accelerate.

🚨Anthropic CEO was just asked: “Is pre-training dead?” > Claude 4 embodies advances in both pre-training and post-training > pre-training scaling laws work the way they’ve worked before > we are going to continue to scale up > the timeline is accelerating > 1-2 years until these models become peers to us Good. Accelerate.

NIK

61,775 次观看 • 1 年前

Sari Azout says the term “artificial intelligence” is a branding error from 1955. LLMs aren't artificial. They're collective intelligence: just better ways of aggregating what humans already know and have made. “Changing your vocabulary can change your reality.” Maybe the fear around AI doesn't come from what it does, but from what we've chosen to call it.

Sari Azout says the term “artificial intelligence” is a branding error from 1955. LLMs aren't artificial. They're collective intelligence: just better ways of aggregating what humans already know and have made. “Changing your vocabulary can change your reality.” Maybe the fear around AI doesn't come from what it does, but from what we've chosen to call it.

vitrupo

123,988 次观看 • 1 年前

Ilya Sutskever says the AI scaling era is running out of road because pre-training is about to hit a finite-data wall. "At some point though, pre-training will run out of data. The data is very clearly finite." "Then, okay, what do you do next? Either you do some kind of a souped-up pre-training... or you're doing RL, or maybe something else." "In some sense, we are back to the age of research." His timeline is the part most AI investors and founders should hear: "From 2012 to 2020, it was the age of research. Now from 2020 to 2025, it was the age of scaling... But now the scale is so big." "Is the belief really that... if you had 100x more everything would be so different?" "I don't think that's true. So it's back to the age of research again, just with big computers."

Ilya Sutskever says the AI scaling era is running out of road because pre-training is about to hit a finite-data wall. "At some point though, pre-training will run out of data. The data is very clearly finite." "Then, okay, what do you do next? Either you do some kind of a souped-up pre-training... or you're doing RL, or maybe something else." "In some sense, we are back to the age of research." His timeline is the part most AI investors and founders should hear: "From 2012 to 2020, it was the age of research. Now from 2020 to 2025, it was the age of scaling... But now the scale is so big." "Is the belief really that... if you had 100x more everything would be so different?" "I don't think that's true. So it's back to the age of research again, just with big computers."

Karl Mehta

199,606 次观看 • 18 天前