[tux]

I think everyone should now make mistakes so we ca distinguish human vs ai.

[zeristor]

This can be optimised for no doubt, adversarial training is like that