Using learnings from small scale derisks to train larger model
@_jasonwei
-
Handling the Waiting Period During o3 Model’s Extended Thinking
By
–
Biggest unsolved problem in artificial intelligence is what to do in the 3-5 minutes when o3 is thinking
-
OpenAI o3-mini Model Achieves Impressive Performance Benchmarks
By
–
Throughout my time at OpenAI I have found Hongyu Ren to be absolutely ruthless. No mercy for evaluation benchmarks whatsoever: o3-mini is 83% on AIME, 2000+ codeforces elo. Every o*-mini model is so performant and fast. Congrats @ren_hongyu @shengjia_zhao and crew!
-
O3 Performance Shows Rapid Progress in RL Chain-of-Thought Scaling
By
–
o3 is very performant. More importantly, progress from o1 to o3 was only three months, which shows how fast progress will be in the new paradigm of RL on chain of thought to scale inference compute. Way faster than pretraining paradigm of new model every 1-2 years
-
Measurable Metrics Enable Optimization in Reinforcement Learning
By
–
Law of reinforcement learning: if you can measure it, you can optimize it
-
Rising Demand for AI Safety Work as Model Capabilities Increase
By
–
I am not an AI safety researcher so I have no skin in the game but several forces point to the value of and demand for safety work shooting up like crazy in the next year or two: 1. With each new generation of models, capabilities grow, leading to increased surface area for
-
Chat and Canvas: Future of Human-AI Collaboration
By
–
Chat is a super crude interface for interacting with AI and canvas is a first step in a future world of collaborative environments where human and AI will work together
-
Jason Wei Excited to Meet Friends at NeurIPS Conference
By
–
Excited to be at neurips, always so much fun to meet old friends and make new ones
-
Sora’s Video Generation Could Transform Computing Interfaces
By
–
Sora is amazing. If you extrapolate the scaling trend in video generation, in a decade using a computer will just be interacting with a sora model, which is crazy to even think about (not sure if it will come true) https://t.co/UNx7HX6muQ
— Jason Wei (@_jasonwei) 9 décembre 2024Sora is amazing. If you extrapolate the scaling trend in video generation, in a decade using a computer will just be interacting with a sora model, which is crazy to even think about (not sure if it will come true)
-
From Skeptic to RL Enthusiast: A Two-Year Transformation
By
–
2022: I never wrote a RL paper or worked with a RL researcher. I didn’t think RL was crucial for AGI Now: I think about RL every day. My code is optimized for RL. The data I create is designed just for RL. I even view life through the lens of RL Crazy how quickly life changes