Little late to this, but I guess Anthropic didn’t like that.
One of the funnier effects of OAI and Anthropic racing towards recursive self-improvement is that the models are considerably more knowledgeable about every facet of LLM training.
Just like noticeably better intuition about all the science, distributed systems, it's crazy.
Users admire Anthropic's superior in-house post-trained models as highly advanced in the race with OpenAI toward recursive self-improvement.
No Digg Deeper questions have been answered for this story yet.
@code_star
@code_star You can guarantee they have ones way better post trained in house.
Can't imagine how sick the fable RL'd on anthropic train repo must be.
Little late to this, but I guess Anthropic didn’t like that.
One of the funnier effects of OAI and Anthropic racing towards recursive self-improvement is that the models are considerably more knowledgeable about every facet of LLM training.
Just like noticeably better intuition about all the science, distributed systems, it's crazy.
@code_star lol wonder if they read that and immediately closed the tab or flagged it internally