Stephen Wolfram: Unraveling the Mysteries of Large Language Models, Intelligence and Autonomous Crypto Agents
The Delphi Podcast
00:00
How a Neural Net Can Invent Something New
I think what is, I think you were also kind of thinking about this point that sort of can a neural net sort of invent something new? Well, it can certainly write essays that have never been written before. It's the question of whether it can really kind of figure out something that goes beyond what we humans already kind of know on the web. But when it comes to figuring out a new fact about science or mathematics or something like this, no, it's not going to be able to do that. That's not, that's not its nature. The power tool way of getting new knowledge is to use this idea of computation which interestingly sort of emerged between the time when neural nets
Transcript
Play full episode
