How do you outsmart the evil AI?
AI eats data and breathes data; humans do not. It follows that AI can control humans but humans cannot control AI. Apparently.
Read more
Who has not been banned from Facebook for something by now? A few years ago, whoever got a ban counted as special: they were confused, they were ashamed, and people looked at them as if they had been trespassing. In a few years’ time, though, the special one will be the person who has not yet been banned - and people will look at them as if they had incredible connections.
A few years ago an acquaintance of mine was banned by Facebook for 30 days over a funny picture shared years earlier, with a justification that was not true - and then banned immediately for another 30 days, because he had already shared the same meme once, even further back. The algorithm did not understand the humour: it picked out a single detail of the picture, passed judgement on that basis, and then handed down the punishment.
The accusation was not preceded by questions, nor by any chance to defend himself, and after the punishment there was no legal remedy. He could appeal, and he did. The appeal meant that they would “take a closer look” at the picture. Long after the 60-day punishment had expired, the notification arrived: the appeal was accepted, the punishment waived. Except it had already been served…
Automated accusation, immediate automated punishment, automated appeal, automated decision. The whole process runs in a way that involves no questions, no confrontation, no evidence, no procedure, no hearing - and there is no way to communicate with a living person.
You could say fine, it is only a picture, and only a 60-day ban, not a big deal. If you make your living from Facebook ads, it is a big deal; but otherwise let us say you are right, it is not a big deal - still, let us think it through a little further.
Either you do something or you do not, but the algorithm concludes that you broke the rule and punishes you for it. It bans you. Or suspends you. Or takes money out of your account. Or sentences you to prison. Or to death.
Automatically. Without a question. Without a hearing. Without the possibility of communication. Without legal remedy. With an automated answer to an automated appeal, dragged out for an astonishingly long time.
Where does this lead? Someone is sentenced to death, executed, and then the notification arrives that the appeal was accepted and a posthumous apology is due? Or it is not accepted, and the Wikipedia entry about them as a criminal is generated automatically - because the algorithm did not understand one sentence in the post that set off the avalanche?