Lex Fridman Podcast
#434 – Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet
Aravind Srinivas
and spent a lot of Google compute. Like I think probably like 400 million parameter model or something, even back in those days. And then this grad student Badano in Bengio's lab identifies attention and beats his numbers with Valus compute. So clearly a great idea. And then people at DeepMind figured that, like this paper called Pixel RNNs, figured that you don't even need RNNs.
0
💬
0
Comments
Log in to comment.
There are no comments yet.