[thr0wawayf00]

> Online platforms will go towards requiring real human IDs for participants.

I just don't see how AI won't be used to manipulate these kinds of barriers. Once AI reaches a point where it can truly equal or surpass humans in terms of operating online, how are we going to prevent it from circumventing those barriers? If AI can generate code from a chat prompt, who's to say it can't solve a captcha in the future? And once that happens (because we all know it will at some point), how are we going to be able to differentiate?

[Nathanba]

real human ids = passports, bank logins

[madars]

There is a market for KYC'd accounts. Someone operating an AI bot farm posing as Real Humans would just buy these.

[thr0wawayf00]

Ok, but the systems that are responsible for creating those IDs are automated, which means that they can be learned and reverse-engineered. There are lots of passports and bank logins floating around that can be used to train AI.

Do you see the problem now?

[int_19h]

The other problem is that AI can be put in charge of phishing operations. Once you devise the correct prompt to get past the filters, it understands what phishing is, and will quite happily write phishing emails for any audience you care to describe. Combine that with an automatic mailer, and you could just fish for IDs (along all the other profitable stuff), and then use those IDs to spam more etc.

[mike_hearn]

Passports contain digitally signed X.509 certificates in their chips. No AI can learn to forge a digital signature (we hope).

[baq]

So you buy them signed at the source.

[mike_hearn]

The only passport you can buy is your own, right? The signing keys are held by national governments. I don't really understand the threat model here.

[baq]

A government sells you passports without validation of your identity.

[Nathanba]

they are not automated and require real world steps