Lex Fridman Podcast
#434 – Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet
Aravind Srinivas
big enough in quantity and high quality and do the right evals on like a lot of reasoning benchmarks. So I think that ended up being the breakthrough, right? Like this, it's not like attention alone was important. Attention, parallel computation, transformer, scaling it up to do unsupervised pre-training, write data, and then... constant improvements.
0
💬
0
Comments
Log in to comment.
There are no comments yet.