Reposting this here. https://twitter.com/gordonbrianr/status/1599835588102066177?s=61&t=dHcCANGfY1-3T6JobMqKMA
@cognition @machinelearning @cogsci
@cognition @machinelearning @cogsci
Brian Gordon on Twitter
“There is (or seems to be, at least) a Quinean type of conceptual emergence here, where the deeper-layers of the LLM seem to have captured the logic used by people to construct the corpus on which it was trained. I take this as something of a vindication of the Two Dogmas model.”
