Buck comments on Large language models aren’t trained enough

[Buck]

Fwiw in the conversations I’m in (in the alignment scene in the Bay Area) this point is widely understood.