Oh I’m so sorry Zoe.

Mostly for natural language semantics but partially for higher-order logic.

Sorry meant informally that talking up the stuff I do.

That’s close to what Sarah Moss covers on our main upper-level/grad formal course.

websites.umich.edu/~ssmoss/Phil...

I'm talking my book a little here, but I think a really careful modal logic course (or unit) is a pretty good mix of translation between natural and formal language, proofs (in various guises) and modeling.

Where I am the dispute is whether you should have a C20 style metatheory/Gödel course as the compulsory course, or something that did modal logic(s), maybe some second-order logic, plus a bunch of probability and lambda calculus stuff.

We call the latter formal methods and I think it's more helpful

Yep. Put another way, the sin, or at least externality, it’s trying to reduce is road usage.

You see a bunch of people saying that papers should include acknowledgments of all uses of "AI". I never know whether the people who say this intend to include the uses of LLMs as search engines.

Completely separately, there is a very interesting Wojtowicz and DeDeo paper - arxiv.org/pdf/2407.14452 - arguing that LLMs make it harder to signal sincere willingness to cooperate across a variety of social situations, by making it cheaper to send previously costly signals.

In U.S. philosophy an annoying norm has developed that reference letters aren't taken seriously unless they are padded out with pages of description of the candidate's file. This signals some familiarity with the file, but it's mostly tedious to read/write. Maybe that will go away with LLMs.

It's hard to get a single volume that attempts anything like what Russell did. The volumes that @histphilosophy.bsky.social is doing based on his podcast, e.g, this one, are great, and accessible without doing a philosophy major. But you'd need to read a lot more to cover what Russell covers.

My favorite heuristic about LLMS that they're usually unhelpful when they give you 1 option, and usually helpful when they give you more. Exceptions in both directions of course, but it's a good heuristic.

I don’t think they are uploading their own papers with the prompt “Tell me some objections to this”; that would be interesting to see how they found it.

If they are they haven’t told me. The most I’ve heard is that some use it as super-Google, asking for explanations of unfamiliar terminology. I assume it’s not 100% reliable, but probably better than guessing or random web searching.