Yes they do. Don't they? It's a very unfaithful simulation of a life.

Right, but The Sims didn’t use a LLM to power their agents.

The point here is they are strapping a supposedly non-agentic LLM into a new test rig and are able to observe agentic behaviors.

It’s very obviously not claiming that this is impressive from a gaming SOTA perspective. It’s just surprising that ChatGPT can do this sort of thing.

I don't think this is surprising at all?