Interesting question, it's gonna be a challenge, but I think the relays are key here, you can make a million bots pretending to be humans but you can't make them pass through paid relays forever ;) you destroy the incentive to create new bots, something that twitter fails miserable because there is no real, and physical punishment when you get banned, called out etc..
Discussion
Yeah true. Its within twitters interests to have bots actually as it 'fills the room'.
An open protocol that creates a hash of a human fingerprint or biometric that allows verification could be cool. But needs careful thought. I'm not sure captcha would cut it as proof of human.
Metrics around posts that give insight to likelihood of bots might be important also.
All just thinking out loud!!