Scaling LLMs further is an artform - Demis Hassabis (Google DeepMind CEO)

Dwarkesh Patel · 4:52

Demis Hassabis explains that Gemini was not jumped an order of magnitude larger because datacenter compute, distributed training, hardware, and scaling-law recipes all impose real limits—and because training loss can...

Read the full summary on tuber

Redirecting...