Week notes — 16th July 2024

This week I’ve been looking at the idea of AI Agents — that is, AIs that are able to interact with the outside world (exact definitions are variable), and how agents might fit into organisations — using my work on Planning and Neighbourhood Plans as a thought experiment.

OpenAI have recently released a hierarchy of AI capabilities, where the ‘agent’ concept is level two of five. Matt Webb has some interesting thoughts on agents. BCG has some research that might be interesting on setting KPIs using AI. I mean, they’ve made it boring, but I feel like there could be a hidden gem in there.

One mind-expanding aspect of this is that it quickly becomes apparent that AI agents will be interacting, spiralling off into all kinds of unforeseeable activity. I’ll be writing a post on this shortly so I won’t type out more here.

Also, I’m back on the case with the cs224d course after a bit of a hiatus.