[WheelsAtLarge]

I’ve been thinking that we’ll end up with specialized language models for different tasks, all connected by a router that picks the best one for each job. Put together, that could feel like a “super LLM”, not true AGI, but close enough that it might seem like it to most people. Because of that, I don’t think Mythos is a bluff.

I don’t think LLMs will ever be truly sentient, but they’ll likely appear that way. It won’t be real AGI, just something more powerful than any single human. Think of it as the oracle at Delphi. I'm not shocked that there's a better model coming. I'm sure there will be more powerful models to come after it too.