Lex Fridman Podcast
#434 – Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet
Aravind Srinivas
Okay, maybe you can use historic data. I only give you data until Q1 and see if you predicted well for Q2 and you train on that signal. Maybe that's useful. And then you still have to collect a bunch of tasks like that and create a RL suite for that, or give agents tasks like a browser and ask them to do things and sandbox it.
0
💬
0
Comments
Log in to comment.
There are no comments yet.