“Talk to the datasette.io database is the result. It’s a GPT that can answer questions by executing SQL queries against the /content.db database that powers the official Datasette website.” 🆓

Link. From Nov but an explanation of GPTs. Not too useful now, but eventually

“… a way of building conversational interfaces for all kinds of weird and interesting problems”

However: all GPT content can be hacked by prompt injection.

Paper On AI Monosemanticity: “The AI gradually shifts to packing its concepts into tetrahedra (three neurons per four concepts) and triangles (two neurons per three concepts). When it reaches digons (one neuron per two concepts) it stops”

Link. Heroic effort to explain a very technical paper. Recommended for those who make confident statements about the near future.

“Anthropic’s interpretability team announced that they successfully dissected of one of the simulated AIs in its abstract hyperdimensional space.”