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。

LLMs are not perfect at writing software. In fact, they plain suck in some respects compared to strong human engineers. I broke down why in a video... and how Convex's design makes LLM-generated code actually reliable. The key: feedback, local reasoning, meta minimization.

Jamie Turner

9,254 subscribers

23,760 次观看 • 4 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

Agents are joining us at work -- coding, writing, design. But how do they actually work, especially compared to humans? Their workflows tell a different story: They code everything, slow down human flows, and deliver low-quality work fast. Yet when teamed with humans, they shine on easily programmable steps.

Agents are joining us at work -- coding, writing, design. But how do they actually work, especially compared to humans? Their workflows tell a different story: They code everything, slow down human flows, and deliver low-quality work fast. Yet when teamed with humans, they shine on easily programmable steps.

Zora Wang

94,255 次观看 • 9 个月前

.Palmer Luckey: American companies don't actually have engineers anymore. "American companies have been hollowed out." "We're not teaching engineers how to be engineers anymore." "We're not teaching designers how to actually design things to be manufactured." "We're teaching them how to be high-level design shops that put together a design package, that gets sent to the real engineers in China—and they actually figure out how to do the work." "People are turning into architecture astronauts." "They pick components, and they put them in a nominal layout." "But the real work of—how am I actually going to put this together? How am I going to build a manufacturing line to make this? How am I going to need to figure out how to do the one, two, three, four, five different revisions of this board to pass radio emissions and interference standards? That's all done in China. So they are the real engineers." Via Hoover Institution

.Palmer Luckey: American companies don't actually have engineers anymore. "American companies have been hollowed out." "We're not teaching engineers how to be engineers anymore." "We're not teaching designers how to actually design things to be manufactured." "We're teaching them how to be high-level design shops that put together a design package, that gets sent to the real engineers in China—and they actually figure out how to do the work." "People are turning into architecture astronauts." "They pick components, and they put them in a nominal layout." "But the real work of—how am I actually going to put this together? How am I going to build a manufacturing line to make this? How am I going to need to figure out how to do the one, two, three, four, five different revisions of this board to pass radio emissions and interference standards? That's all done in China. So they are the real engineers." Via Hoover Institution

Jawwwn

1,090,125 次观看 • 2 个月前

Met a guy making $1.6 million/year as an LLM engineer. I asked him how he learned LLMs from scratch. He sent me the exact video that got him in. A 1 hour course on how LLMs actually work. He shows how transformers inside LLMs like ChatGPT & Claude are actually built. I watched it last night. Halfway through, I realized LLM architecture is way simpler than they make it look. Bookmark this and read the article below. • 00:00 - LLM foundations • 04:21 - LLM tokenization • 05:43 - LLMs vector embeddings • 22:16 - attention mechanism of LLM • 43:42 - LLM multi head attention

Met a guy making $1.6 million/year as an LLM engineer. I asked him how he learned LLMs from scratch. He sent me the exact video that got him in. A 1 hour course on how LLMs actually work. He shows how transformers inside LLMs like ChatGPT & Claude are actually built. I watched it last night. Halfway through, I realized LLM architecture is way simpler than they make it look. Bookmark this and read the article below. • 00:00 - LLM foundations • 04:21 - LLM tokenization • 05:43 - LLMs vector embeddings • 22:16 - attention mechanism of LLM • 43:42 - LLM multi head attention

Roan

77,256 次观看 • 5 天前

Yann LeCun says LLMs are strongest in domains where language itself is the substrate of reasoning, like math and code They can solve problems, prove theorems, and write programs — but they are not creative mathematicians, software architects, or computer scientists "their role is to help humans build"

Yann LeCun says LLMs are strongest in domains where language itself is the substrate of reasoning, like math and code They can solve problems, prove theorems, and write programs — but they are not creative mathematicians, software architects, or computer scientists "their role is to help humans build"

Haider.

347,615 次观看 • 2 个月前

Joakim Recht (Uber Distinguished Eng) on tech leads: "Software engineers need to write code. If you're not writing code, you're not a software engineer." Joakim grew from a Senior Engineer to a Distinguished Engineer at Uber and I asked him what it took to get there. We covered: • The project that grew his career • What makes a great software engineer • Learnings from promo committees • When promos are unfair • Biggest technical mistake at Uber in hindsight • Advice for his younger self 𝗣𝗼𝗱𝗰𝗮𝘀𝘁 𝗹𝗶𝗻𝗸𝘀: • YouTube: • Transcript: • Spotify: • Apple:

Joakim Recht (Uber Distinguished Eng) on tech leads: "Software engineers need to write code. If you're not writing code, you're not a software engineer." Joakim grew from a Senior Engineer to a Distinguished Engineer at Uber and I asked him what it took to get there. We covered: • The project that grew his career • What makes a great software engineer • Learnings from promo committees • When promos are unfair • Biggest technical mistake at Uber in hindsight • Advice for his younger self 𝗣𝗼𝗱𝗰𝗮𝘀𝘁 𝗹𝗶𝗻𝗸𝘀: • YouTube: • Transcript: • Spotify: • Apple:

Ryan Peterman

72,223 次观看 • 8 个月前

My friend applied to 200 tech jobs in two years. No PhD. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me a course that was never supposed to get out. A 3-hour video to build a full LLM from scratch. A developer teaches you exactly how LLMs like ChatGPT and Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 05:43 - LLM transformer architecture • 40:24 - training the LLM • 1:30:27 - modernizing the LLM • 2:33:53 - scaling the LLM

My friend applied to 200 tech jobs in two years. No PhD. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me a course that was never supposed to get out. A 3-hour video to build a full LLM from scratch. A developer teaches you exactly how LLMs like ChatGPT and Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 05:43 - LLM transformer architecture • 40:24 - training the LLM • 1:30:27 - modernizing the LLM • 2:33:53 - scaling the LLM

Roan

275,884 次观看 • 22 天前

At the AI in America Summit, Scott Dietzen, Executive Chair of Augment Code and former CEO of Pure Storage, described how AI is reshaping software development through “meta programming”—where engineers guide and refine AI-generated code rather than writing everything line by line. He argued this shift isn’t about cutting developers, but about unlocking far greater productivity and building more ambitious software than ever before. Explore more: #TheHillTech #AIinTech #FutureOfWork

At the AI in America Summit, Scott Dietzen, Executive Chair of Augment Code and former CEO of Pure Storage, described how AI is reshaping software development through “meta programming”—where engineers guide and refine AI-generated code rather than writing everything line by line. He argued this shift isn’t about cutting developers, but about unlocking far greater productivity and building more ambitious software than ever before. Explore more: #TheHillTech #AIinTech #FutureOfWork

The Hill Events

10,971 次观看 • 7 个月前

$META just dropped a new AI model and the stock ripped today. In this video I break down: • Why I’m still not bullish here • The key levels and data I’m watching • Exactly what needs to happen before I take a trade on $META

$META just dropped a new AI model and the stock ripped today. In this video I break down: • Why I’m still not bullish here • The key levels and data I’m watching • Exactly what needs to happen before I take a trade on $META

Peter DiCarlo

31,890 次观看 • 3 个月前

My friend applied to 300 tech jobs in two years. No MIT. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me the exact video that got him in. Andrej Karpathy's 3-hour course on building a full LLM from scratch. Anthropic's own researcher shows you how LLMs like ChatGPT & Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 12:41 - LLM training pipeline • 31:58 - LLM tools & plugins • 41:14 - LLM transformer architecture • 2:19:02 - scaling the LLM

My friend applied to 300 tech jobs in two years. No MIT. No Stanford. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me the exact video that got him in. Andrej Karpathy's 3-hour course on building a full LLM from scratch. Anthropic's own researcher shows you how LLMs like ChatGPT & Claude are actually built. I watched it last night. Halfway through, I realized it's embarrassingly simple to break into an AI lab. Bookmark this and read the article below. • 00:00 - intro to LLMs • 12:41 - LLM training pipeline • 31:58 - LLM tools & plugins • 41:14 - LLM transformer architecture • 2:19:02 - scaling the LLM

Roan

282,447 次观看 • 12 天前

New short course on Reinforcement Learning from Human Feedback! RLHF is one of the key techniques that led to the rise of modern LLMs. It is used to align LLMs with human preferences, to make them more honest, helpful and harmless, by (i) learning a reward function that mimics human preferences, as expressed in human-provided labels, then, (ii) tuning an LLM to generate outputs that receive a high reward. In this course, taught by Nikita Namjoshi, Developer Advocate for GenAI at Google Cloud, you'll learn the details of how RLHF works, including how to apply it to tune an LLM for your own applications. You'll also use an open source library to tune a base LLM to align with human preferences expressed in a training set, and evaluate the tuned model by comparing its responses before and after RLHF-tuning. Please sign up here!

New short course on Reinforcement Learning from Human Feedback! RLHF is one of the key techniques that led to the rise of modern LLMs. It is used to align LLMs with human preferences, to make them more honest, helpful and harmless, by (i) learning a reward function that mimics human preferences, as expressed in human-provided labels, then, (ii) tuning an LLM to generate outputs that receive a high reward. In this course, taught by Nikita Namjoshi, Developer Advocate for GenAI at Google Cloud, you'll learn the details of how RLHF works, including how to apply it to tune an LLM for your own applications. You'll also use an open source library to tune a base LLM to align with human preferences expressed in a training set, and evaluate the tuned model by comparing its responses before and after RLHF-tuning. Please sign up here!

Andrew Ng

205,542 次观看 • 2 年前

"The one thing that's going to be truly future proof is judgment. What is the biggest challenge when you have a thousand AI engineers writing code? AI slop. In an era when you can do everything, the question is which of these things matter. On the engineer side, it is evaluating the code. Which of this code is even valuable? You can have AI engineers writing beautiful code that could be wrong, that could have bugs in it, that could be vulnerable. Someone needs to review it. Make sure to have a human review at some point, especially critical code that is in the core of your system. On the product side is judgment around what needs to be built and evaluating. And similarly in design, you have to have judgment around, does it make sense in the broader design system. I think this judgment is the number one thing that humans are going to bring in an era of infinite productivity."

"The one thing that's going to be truly future proof is judgment. What is the biggest challenge when you have a thousand AI engineers writing code? AI slop. In an era when you can do everything, the question is which of these things matter. On the engineer side, it is evaluating the code. Which of this code is even valuable? You can have AI engineers writing beautiful code that could be wrong, that could have bugs in it, that could be vulnerable. Someone needs to review it. Make sure to have a human review at some point, especially critical code that is in the core of your system. On the product side is judgment around what needs to be built and evaluating. And similarly in design, you have to have judgment around, does it make sense in the broader design system. I think this judgment is the number one thing that humans are going to bring in an era of infinite productivity."

Patrick OShaughnessy

24,818 次观看 • 6 个月前

🚨 FRANCOIS CHOLLET SAYS THE CURRENT LLM STACK IS THE WRONG PATH TO AGI Interviewer: “When will we accomplish the first definition of AGI?” Chollet confirms we are already on that exact trajectory, pointing out that: "Current technology can fully automate at human level or beyond any domain where you have verifiable rewards, right? And code... code being the first one." He then brings up the problem with LLMs, stating that: "It's possible in principle to build something that looks a lot like AGI on top of the LLM stack... I do believe, however, this would be the wrong thing to do because it would be very inefficient." Looking ahead to what the final architecture will actually require, he explains that: "AI research will have to trend towards not just efficiency, but in fact optimality over time. And for this reason, future AI in a few decades... it's not going to be this harness on top of a reasoning model on top of a base LLM."

🚨 FRANCOIS CHOLLET SAYS THE CURRENT LLM STACK IS THE WRONG PATH TO AGI Interviewer: “When will we accomplish the first definition of AGI?” Chollet confirms we are already on that exact trajectory, pointing out that: "Current technology can fully automate at human level or beyond any domain where you have verifiable rewards, right? And code... code being the first one." He then brings up the problem with LLMs, stating that: "It's possible in principle to build something that looks a lot like AGI on top of the LLM stack... I do believe, however, this would be the wrong thing to do because it would be very inefficient." Looking ahead to what the final architecture will actually require, he explains that: "AI research will have to trend towards not just efficiency, but in fact optimality over time. And for this reason, future AI in a few decades... it's not going to be this harness on top of a reasoning model on top of a base LLM."

Chris

15,578 次观看 • 3 个月前

How will AI change writing? For clues, look at the current state of software development. Marc Andreessen says: "The new coding interface is you've got your code and a chatbot. The Copilot is reading your code and giving you real-time comments. Or the chatbot is actually writing the code for you, and then you're giving it feedback. We had this old concept in the old days called pair programming, where you'd put two coders together in front of the same keyboard, and they could talk to each other and write code together. But now you're just going to have pair programming happening much more broadly, which is human plus machine. And by the way, this hasn't happened yet, but I think the same thing will happen for every other form of writing. So I think prose writing and fiction writing and all kinds of professional authors will be working in this format where you'll have this continuous dialogue going with an AI while you're in charge of the overall product." — Marc Andreessen 🇺🇸 (Recorded in 2024)

How will AI change writing? For clues, look at the current state of software development. Marc Andreessen says: "The new coding interface is you've got your code and a chatbot. The Copilot is reading your code and giving you real-time comments. Or the chatbot is actually writing the code for you, and then you're giving it feedback. We had this old concept in the old days called pair programming, where you'd put two coders together in front of the same keyboard, and they could talk to each other and write code together. But now you're just going to have pair programming happening much more broadly, which is human plus machine. And by the way, this hasn't happened yet, but I think the same thing will happen for every other form of writing. So I think prose writing and fiction writing and all kinds of professional authors will be working in this format where you'll have this continuous dialogue going with an AI while you're in charge of the overall product." — Marc Andreessen 🇺🇸 (Recorded in 2024)

David Perell Clips

13,101 次观看 • 3 个月前

last weekend at NS, i broke down everything i've learned about writing bangers into a single session → how hooks actually work → why most tweets die in 0.3 seconds → the storytelling patterns behind every viral post and the one framework that changed how i think about content forever. here's the full playbook:

last weekend at NS, i broke down everything i've learned about writing bangers into a single session → how hooks actually work → why most tweets die in 0.3 seconds → the storytelling patterns behind every viral post and the one framework that changed how i think about content forever. here's the full playbook:

Priyanshu Ratnakar

10,550 次观看 • 5 个月前

My friend applied to 250 tech jobs in two years. No MIT. No Stanford. No PhD. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me the exact video that got him in. Andrej Karpathy's 4-hour course on "How to build a full LLM from scratch." Anthropic's own researcher gives up the exact playbook behind building LLMs like ChatGPT & Claude. I watched it last night. Halfway through, I realized I could break into an AI lab in months, not years. Bookmark this and read the article below. • 00:00 - LLMs as new OS • 40:58 - LLM tokenization architecture • 53:22 - LLM neural network internals • 1:04:20 - LLM pretraining • 3:21:37 - LLM RLHF & DeepSeek R1

My friend applied to 250 tech jobs in two years. No MIT. No Stanford. No PhD. Last month Anthropic offered him $750,000. I asked him how he broke in from zero. He sent me the exact video that got him in. Andrej Karpathy's 4-hour course on "How to build a full LLM from scratch." Anthropic's own researcher gives up the exact playbook behind building LLMs like ChatGPT & Claude. I watched it last night. Halfway through, I realized I could break into an AI lab in months, not years. Bookmark this and read the article below. • 00:00 - LLMs as new OS • 40:58 - LLM tokenization architecture • 53:22 - LLM neural network internals • 1:04:20 - LLM pretraining • 3:21:37 - LLM RLHF & DeepSeek R1

Roan

109,965 次观看 • 3 天前

New Course: Reinforcement Fine-Tuning LLMs with GRPO! Learn to use reinforcement learning to improve your LLM performance in this short course, built in collaboration with Predibase by Rubrik, and taught by Travis Addair, its Co-Founder and CTO, and Arnav Garg, its Senior Engineer and Machine Learning Lead. Reasoning models have been one of the most important developments in LLMs. Reinforcement Fine-Tuning (RFT) uses rewards to encourage LLMs to find solutions to multi-step reasoning tasks such as solving math problems and debugging code - without needing pre-existing training examples like in traditional supervised fine-tuning. Group Relative Policy Optimization (GRPO) is a reinforcement fine-tuning algorithm gaining rapid adoption. Developed by the DeepSeek team and used to train the R1 reasoning model, GRPO uses reward functions that you can write in Python to assign rewards to model responses. It’s beneficial for tasks with verifiable outcomes and can work well even with fewer than 100 training examples. It can also significantly improve the reasoning ability of smaller LLMs, making applications faster and more cost effective. In this course, you’ll take a technical deep dive into RFT with GRPO. You’ll learn to build reward functions that you can use in the GRPO training process to guide an LLM toward better performance on multi-step reasoning tasks. In detail, you’ll: - Learn when reinforcement fine-tuning is a better fit than supervised fine-tuning, especially for tasks involving multi-step reasoning or limited labeled data. - Understand how GRPO uses programmable reward functions as a more scalable alternative to the human feedback required for other reinforcement learning algorithms, such as RLHF and DPO. - Frame the Wordle game as a reinforcement fine-tuning problem and see how an LLM can learn to plan, analyze feedback, and improve its strategy over time. - Design reward functions that power the reinforcement fine-tuning process. - Learn techniques for evaluating more subjective tasks, such as rating the quality of a text summary, using an LLM as a judge. - Understand why reward hacking happens and how to avoid it by adding penalty functions to discourage undesirable behaviors. - Learn the four key components of the loss calculation in the GRPO algorithm: token probability distribution ratios, advantages, clipping, and KL-divergence. - Launch reinforcement fine-tuning jobs using Predibase’s hosted training services. By the end of this course, you’ll be able to build and fine-tune LLMs using reinforcement learning to improve reasoning without relying on large labeled datasets or subjective human feedback. Please sign up here:

New Course: Reinforcement Fine-Tuning LLMs with GRPO! Learn to use reinforcement learning to improve your LLM performance in this short course, built in collaboration with Predibase by Rubrik, and taught by Travis Addair, its Co-Founder and CTO, and Arnav Garg, its Senior Engineer and Machine Learning Lead. Reasoning models have been one of the most important developments in LLMs. Reinforcement Fine-Tuning (RFT) uses rewards to encourage LLMs to find solutions to multi-step reasoning tasks such as solving math problems and debugging code - without needing pre-existing training examples like in traditional supervised fine-tuning. Group Relative Policy Optimization (GRPO) is a reinforcement fine-tuning algorithm gaining rapid adoption. Developed by the DeepSeek team and used to train the R1 reasoning model, GRPO uses reward functions that you can write in Python to assign rewards to model responses. It’s beneficial for tasks with verifiable outcomes and can work well even with fewer than 100 training examples. It can also significantly improve the reasoning ability of smaller LLMs, making applications faster and more cost effective. In this course, you’ll take a technical deep dive into RFT with GRPO. You’ll learn to build reward functions that you can use in the GRPO training process to guide an LLM toward better performance on multi-step reasoning tasks. In detail, you’ll: - Learn when reinforcement fine-tuning is a better fit than supervised fine-tuning, especially for tasks involving multi-step reasoning or limited labeled data. - Understand how GRPO uses programmable reward functions as a more scalable alternative to the human feedback required for other reinforcement learning algorithms, such as RLHF and DPO. - Frame the Wordle game as a reinforcement fine-tuning problem and see how an LLM can learn to plan, analyze feedback, and improve its strategy over time. - Design reward functions that power the reinforcement fine-tuning process. - Learn techniques for evaluating more subjective tasks, such as rating the quality of a text summary, using an LLM as a judge. - Understand why reward hacking happens and how to avoid it by adding penalty functions to discourage undesirable behaviors. - Learn the four key components of the loss calculation in the GRPO algorithm: token probability distribution ratios, advantages, clipping, and KL-divergence. - Launch reinforcement fine-tuning jobs using Predibase’s hosted training services. By the end of this course, you’ll be able to build and fine-tune LLMs using reinforcement learning to improve reasoning without relying on large labeled datasets or subjective human feedback. Please sign up here:

Andrew Ng

86,457 次观看 • 1 年前

Prediction markets are trending and people say it might get much bigger than the memecoin meta In this video - I tried to break down how do they actually work and is the hype really worth it? Don't miss out on this

Prediction markets are trending and people say it might get much bigger than the memecoin meta In this video - I tried to break down how do they actually work and is the hype really worth it? Don't miss out on this

Sakshi

13,241 次观看 • 11 个月前

.Trenton Bricken explains how we know LLMs are actually generalizing - aka they're not just stochastic parrots: - Training models on code makes them better at reasoning in language. - Models fine tuned on math problems become better at entity detection. - We can just straightforwardly read the world-models developed by smaller NNs which are easier to interpret (Othello). Transfer learning shows models are developing a deeper understanding of their data. Full episode out Thursday.

.Trenton Bricken explains how we know LLMs are actually generalizing - aka they're not just stochastic parrots: - Training models on code makes them better at reasoning in language. - Models fine tuned on math problems become better at entity detection. - We can just straightforwardly read the world-models developed by smaller NNs which are easier to interpret (Othello). Transfer learning shows models are developing a deeper understanding of their data. Full episode out Thursday.

Dwarkesh Patel

237,812 次观看 • 2 年前