Connor Leahy on AGI and Cognitive Emulation
Future of Life Institute Podcast
00:00
The Inherent Difficulties of Machine Learning Systems
In an ideal world, we would solve all of this formally mathematically proof check everything. I think it is possible, but I don't think it's realistic. No one that works without math are magic in the sense that they use lots of magic. We're gonna have to mix and match some things that black boxes can do with white boxes. So for example, I'm allowed to make the assumption GPT-Sight cannot row hammer or anything like that. It will not be a single system - neither big white box nor large black box.
Transcript
Play full episode
