Ted Underwood (@Ted_Underwood): ML is nowhere near passing a Turing test, but it’s getting easy to see that doing so won’t actually persuade ppl of sentience. I suspect we’ll insist on physical reaction to “pain”; also, systems that can play too many diff roles won’t seem to have “a stable personality.” +

Ted Underwood (@Ted_Underwood): In short, our instincts on this topic are probably crappy and speciesist—heavily shaped by our own contingent limitations.

Ted Underwood (@Ted_Underwood): But philosophically this is interesting, because while I believe in meaning-as-use, I think we’re going to instinctively add provisos about “mere theatricality” that make it hard to distinguish that from a theory of “intention.” https://twitter.com/dan_sinykin/status/1433477540367282179