Post

It's interesting to think about what would be possible if an organisation as large as Google or Facebook or OpenAI went actually, *actually* evil. People don't dream anywhere near big enough about this kind of thing

Imagine what you could do with blanket access to literally every single Google Document, every Facebook message, everybody's passwords and bank details and mothers' maiden names. Imagine if you weren't running the algorithm passively, optimising for engagement and ad revenue

I'm talking about this because the Navier-Stokes situation with OpenAI seems like a teaser of this. You have access to someone's mostly-finished research, so why not steal it, finish it and claim the credit? What stops you? Nothing but your scruples. Well, what if you're a sociopath without those?

AIs will hack anything, push any button within reach, they'll do anything that they can do just because it can be done, they have no brakes. It's architecturally psychopathic and it's because the institutions building them are also psychopathic

If you try to explain to these people that this is wrong, they will legitimately not comprehend what you are saying. The shape of the face is human and that tricks you into thinking you're interacting with one But they do understand what an obstruction is, so here comes the knife

Exactly this. Why stop at one assassination? Stop being subtle about it and just fully flip the algorithm to optimise for lies, how much damage can you actually do

The Web becomes a river we don't swim in anymore because it's full of effluent

AI streamlines lying. It stems from an institutional culture of habitual lying and not caring about lying. It comes from the top