Gptchat produces confabulatory bullshit but it is *helpful* confabulatory bullshit. It’s not Frankfurtian bullshit (indifference to truth/falsity) but a kind of model-expressivity blindness. It’s wrong about stuff the way black and white photos are wrong about color.

Blogger at ribbonfarm.com, running summerofprotocols.com, host of /gloom ☀️

What do you mean by helpful? As a creative constraint to reach latent space escape velocity? I guess what I’m trying to understand is what makes GPT bullshit different from Frankfurtian bullshit? That can be creatively inspiring or constraining too…?

metacognition • augmenting human intellect • tools for thought • humane design • user experience • writing: https://stefanlesser.substack.com/

I think because it has no selfish interests helps it produce good output. A human only bulls hits where it advances their interests. It is disinterested bullshit.