Comment by safety1st

I'm kind of pro AI while also being pro LESS AI.

These are ultimately all software systems where updates can be pushed out with the press of a button. We just found an amazing new tool in the LLM and people are trying out all kinds of ways of using it, many of them dumb or stupid, this is discovery that has to be done, we'll keep doing it and pushing updates.

It doesn't take rose colored glasses really to infer that over time we'll cull the dumbest uses and extrapolate on the best ones. There's a huge difference between using a generic chatbot, even one driven by a SOTA model, vs targeting a specific business task with good guardrails, skills, and tool interfaces that call out to real business logic when the model isn't the right thing. The latter will emerge more and more over time.