🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent). I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed

New 3h31m video on YouTube: "Deep Dive into LLMs like ChatGPT" This is a general audience deep dive into the Large Language Model (LLM) AI technology that powers ChatGPT and related products. It is covers the full training stack of how the models are developed, along with mental

The coolest autonomous coding agent I've seen recently: use AI to write better CUDA kernels to accelerate AI. AutoML is so back! The highest leverage thing you can do with your compute resources is to increase the future productivity of the same compute. It aligns all the stars

What Ethereum needs is a lot of young blood who shared the cypherpunk vision. All OGs are jaded. It’s on the next generation now.

ZKP is a key to crypto privacy solutions Open-source models with local deployment and user-friendly fine-tuning tools should be the key to AI privacy protections. To do that, we need knowledge distillation and more powerful small models Though I only shared a few ideas on the

Some interesting ideas 1. Small but smart model Beyond knowledge distillation, how can we achieve emergent intelligence (like DeepSeek-R1-Zero's "Aha moments") in smaller models? Could techniques like lower-precision training or specialized reward functions during the RL phase

Beautiful words incompleteideas.net/IncIdeas/Bitte…

It's so fun to see RL finally work on complex real-world tasks with LLM policies, but it's increasingly clear that we lack an understanding of how RL fine-tuning leads to generalization. In the same week, we got two (awesome) papers: Absolute Zero Reasoner: Improvements on code

amazing work!

In the 2010s, video games were a fantastic playground for testing RL algorithms. Projects like VizDoom (vizdoom.farama.org) and Super Mario (en.wikipedia.org/wiki/Super_Mar….) were used in these research papers (arxiv.org/abs/1705.05363), alongside milestones like Atari games with DQN

What is vibe coding like these days? A100/H100 as your debug machine 👇

I agree with Jensen. If you want AI development to be done safely and responsibly, you do it in the open. Don’t do it in a dark room and tell me it’s “safe”. Article archive: archive.md/CC5VZ