All of that wannabe AGI horseshit is preventing these things from specialization and becoming more efficient and useful IMO.

In my recent usage of LLMs I’ve come to think of them as a type of eye-poppingly expensive, inefficient interpreter.

Losing a lot of the bullshit could make them efficient and inexpensive. To an extent, that’s what Anthropic has already been doing by making models that are mostly code focused.