LLMs have become the de-facto standard in modern AI toolchain, but it still comes with a lot of confusions - quality, speed, cost, etc. In this talk, Yangqing will share a few observations he has regarding LLM, both from an algorithm engineer and an infra engineer perspective, on how to best utilize LLMs in your daily operations. Lepton AI's CEO also talks about how enterprises think of their IT and AI strategy, given that the fast changing computation pattern is disrupting conventional cloud in unprecedented ways.

‍