Business plan: We set up a server farm running predictors. By predictors I mean something that throws dice using a pseudo-random number generator with different seeds. Each predictor is identified by its seed.

After a time, some of them will have correctly predicted something (be it elections, sports, the weather, …) We then pick one of the correct predictors at random, start marketing that predictor as an AI. Call it Deep Britney.

Maybe it works for future cases, more likely it won’t. But at the same time, we keep running the remaining predictors and after people realize that the first AI is bullshit, we come up with “a new version” which has correctly predicted everything the first one got right AND everything it got wrong.

I believe we should be able to keep the scam going for several years before anybody catches on.