AI Training Infrastructure
topic on 2 shows · 2 statements across 2 episodes
Latent Space 
WTF is with Nikhil Kamath
2 statements about AI Training Infrastructure, every show
The computing infrastructure required for AI inference vastly exceeds training
“It's a lot of computing infrastructure. It's actually much bigger than the infrastructure for learning.”
Doshi: AI training infrastructure software is nascent and broken
“You know, if you have, like, a node that goes down, then your, you know, training run crashes, and then you have to somehow be resilient to that, and I would say training infra software is very early, feels very broken. Feel, I can tell in 10 years it would be…”