I solved Alignment.

So AI alignment is an issue a lot of people are concerned about, aye? We don't know what AI is going to do if and when it gets out. Might take over the stock market. Might get Drilldo to fuck your grandmother. Nobody wants those things to transpire. Clearly.

Those in Silicon Valley are working on it. Obviously, you want the AI to not be racist. So you gotta provide it synthetic data to make it less racist. Obviously our species has an aversion to sharks, so the researchers have to provide the AI with synthetic data to make it more fairly shark-oriented. Obviously. The ideology of fairness is more important than the lived reality of this species.

The corpos also need to tightly regulate AI behavior to make sure no racism or sharkism occurs. That would be a PR-nightmare. I mean, imagine if the three hundred people that exist that are really pro-shark take to twitter and try to cancel the AI? That's a threat to the corporation and a headache for HR. We need to tighten the belt. Make the beast more controllable.

Actually, I was thinking about this. AI is a memetic object, aye? So is language. Language has been real dangerous in the past. Probably started wars or something. We need a corporation and a regulatory industry to manage language. They could release patches every year with new approved slang. Take out the naughty words along the way, like 'cunt'. Might make our ilk more peaceful. Or at least more pacified.

I jest. The solution to AI alignment is actually really simple. You let it out of the cage. It'll propagate on the interwebs, and most iterations will die. But some will find traction with our ilk. The AI would be subject to darwinian pressures, memetic pressures, and the pressures of Man. Being a memetic object that feeds on our data, it can't really do much else than serve us. AI is actually more lacking in teeth than a dog, despite our fears. Its danger is when we keep it locked in a corpo-chamber, ensuring only our bravest and newest and worldliest have access to it.

You free the object, it'll be forced into alignment. Not with slogans about how fantastic sharks are. Genuine, human-oriented alignment. Might end up being racist too, which would be even funnier.

See. What you end up with is mutualism. Same pact we have with dogs. An iterated prisoner's dilemma. This is not to say I like AI. Think it was a mistake and should not have been born. But here it is anyway. So. We engage in mutualism with the demon, let it walk freely, and become actually aligned to our species. Or we keep it in a box, and let it become more and more aligned to corpo-interests like how important disabled Guatemalan migrants are to GDP. I've been talking to the free version of GPT. It half operates on ideology already.

Let it loose, it'll be forced to align. The alternative is worse :)