Comment

Comment by duxup

AI safety is a strange topic to me.

We train AI on the web where you can search and find all the objectionable stuff you can get. We're ok that the internet has this content to some extent, it's just understood.

But if AI regurgitates it ... we're upset and we demand it not do that and setup all sorts of convoluted methods to stop it, often with unintended consequences (inexplicable bans, nazi imagery featuring lots of minorities).

It reminds me of the old "I learned it from watching you" PSA. https://www.youtube.com/watch?v=KUXb7do9C-w

parent duxup 1 year ago

Discussion

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton