Sergey Brin On Model Convergence, Prompt Engineering & Transformer Evolution

There’s a very special kind of house in California called AGI House.

Apparently, it is the de-facto location for hosting largely informal events that bring the local leading edge AI community together for talks, discussions etc. Sometime early last month, Google DeepMind gathered there for an event and afterwards, Sergey Brin held an informal sort of fireside Q&A session. There’s a video recording of the session doing the rounds, and you can find it on X here.

It’s probably one of the most objective assessments of the state of the AI that I’ve seen in a long time. In particular, his thoughts on transformers, model convergence and AGI are quite illuminating. Let’s dig in…