Re:AI: Markets for Lemons, and the Great Logging Off
by rocqua on Thursday December 29, 2022 @06:31PM
(#34179559)
The problem I brought up in the grandparent post wasn't (meant to be) about a 1-to-1 mapping. It was about asking "who owns this key".
Web of trust sounds nice, but it hasn't caught on. I would say because it has trouble going beyond one hop of trust if you actually consider adversaries. In person physical confirmation works, re-keying is a major hassle though.
I do like the idea of a 'web of trust is good enough for declaring you are a person'. If you get a vouching system with a recursive revocation system, that might work well enough for establishing you are a person (though not well enough for establishing which person you are).
Trusted third parties have problems. They centralize power, either in the state, or in some non-accountable organization.
You hit the nail on the head. Matching keys to real people can be done in-person for direct friends, then through a web of trust for indirect friends. For accounts/keys/personas you find on the internet where you don't have a chain of friends, you can either rely on "trusted" third-party attestation ("holder of key 0xdeadbeef earned a degree from this university") - you may never know with complete certainty if that's a real human, a bot, or an alt account for someone you already know, and that's totally fine.
The "problem" of matching keys 1-to-1 with identities for everyone globally (brought up by grandparent post) is a massive red herring that doesn't need to be "solved".