skybrian

[Epistemic status: a basic explanation, high confidence.] Chatting with an AI seems like chatting with a person, so it’s natural to ask an AI chatbot to explain its answers. But it doesn’t work how you’d expect. An AI chatbot doesn't know how it decided what to write, so its explanation...

I will speculate a bit about what language models might be doing. Hopefully it will help us come up with better research questions (for those of you who can do research) and things to try (for those of us informally testing the chatbots), to see how these language models work...

I read an essay that would be good for Less Wrong, but I'm not sure how to post it. The "link post" feature seems to be about posting your own work? How do you share someone else's work?

In Bayesian jargon, people talk about having a "flat prior" when they mean being very uncertain about a question. But more often you just give a probability that you think a question is true. We don't normally draw probability distributions to share with each other. I'm wondering if there's a...

A micromort is a 1-in-a-million chance of death. Wikipedia's article has a few examples of risks of death measured in micromorts, but it seems like there must be a lot more of them. What are some interesting examples you use as a baseline for comparison?

I've seen a few people debating what GPT-3 understands, and how this compares to human understanding. I think there's an easier and more fruitful question to consider: what does it believe? It seems like it doesn't believe anything, or alternately, it believes everything. It's a category error, like asking what...

I'm looking for examples of GPT-3 prompts to achieve certain effects. Is there a forum where this is discussed? In particular, I'd like to see how people generate tweets.