Okay, that makes sense. I had a cached prior that {autoregressive LLMs can’t continual-learnLLMs won’t appear agentic enough to convince the public to pause before they crack CL} which seems much less likely given Anthropic’s recent models. Thanks for helping me clear some cobwebs. :)
Okay, that makes sense. I had a cached prior that {autoregressive LLMs can’t continual-learn