Astral Codex Ten Podcast cover image

ELK And The Problem Of Truthful AI

Astral Codex Ten Podcast

00:00

How to Train a Neural Net at Chess

A r c notes that neural nets can have multiple heads on the same body, different output devices with different goal functions. For example, when you train a chess a i by making it play games against itself, one head would be ered for making black win, and the other for making white win. Pursuing these opposed goals, in qotes, is what makes the neural net good at chess, expertise that both heads have full access to at the same time. Is an image that doll e generated. It's a two headed elk. The two heads are connected at the neck. Are you feeling lucky two? Two heads are better than one. Extended far enough, this line of

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app