If AI is outsmarting people, it'll also outsmart me. Maybe it won't outsmart an AI designed to detect AIs outsmarting people ... but then it's your standard arms (ha!) race.

I don't think DIDs do anything for proving whether someone is a human or not. Dylan precludes infallible instantiation and custody of identity. Until we know that's possible, it's safe to assume it isn't; eg who is verifying the verifiers? PKI is vulnerable in all sorts of ways, and decentralized PKI trades some vulnerabilities for new ones.

A victory over AI fakes requires proof of humanness be trivial to create and verify but asymmetrically difficult to forge. Possession of a cryptographic key is not proof of humanness. For the time being an initial verification process that depends on a person having a physical body is verifying humanness, but how long until AI has a human like body? Even with a perfect initial verification process, something like DIDs exist to avoid reinitializing verification so they'll require infallible private key custody.

People fall for AI content shared here already and it isn't even that good. afaict the best we can do right now is make approval of content and people costly in an unforgeable way.