Hacker News new | past | comments | ask | show | jobs | submit login

It'd be interesting if we could train AI to spot fake accounts. A good sized sample would be lovely...



It would also be interesting if we could convert lead into gold. Technically true, but not a novel observation, and an incredibly difficult problem.

Edit: specifically, none of the interesting parts of this problem/idea are in the phrasing - the only thing to do is just go out and implement it, and that's very difficult because (1) you're essentially trying to solve the Turing Test ("is this a computer or a human?") (2) most of the techniques that you might use to heuristically make this determination can either (a) be defeated very easily or (b) be defeated by another AI made using similar resources+techniques to those used to make the first.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: