AntiDyatlov

AI safety is a completely separate question from the hard problem. Also a very tricky one, given these things are still black boxes.

Yeah, we're totally fucked, there is no scientific theory that can tell you what is and isn't conscious. For all we know, my laptop, not running any LLM is conscious and always has been. Or my chair. Or a proton. This consciousness thing is a nasty problem for the scientific worldview.

I can recommend Void Stranger. Real mystery box of a game. The first ending seriously hooked me into figuring out all its secrets.

So the power grid is having experiences? Computers too?

Well, actually, probably not. If you say we're made out of meat, you end up with the hard problem of consciousness.

I'm imagining a purple cube in this moment. Is the purple cube made out of meat?

A very good outcome for AI safety would be if when improved models get released, malicious actors use them to break society in very visible ways. Looks like we're getting close to that world.

How can human races not be a meaningful genetic category? Aside from the phentoypical differences, the prevalence of some diseases varies by race.

How can the service know that the users have really entered into a relationship?

If you have time to post on Hacker News, you definitely have time for proper grammar.

I think the contrived scenarios you come up with need to not have a trivial solution. Everything about my brain is optimized for KISS, it breaks everything to turn down simple solutions to reach for something more complex.

Well, dethroning the 0.01% seems more feasible than starting and winning a civil war against the right. I think the right is not so pro 0.01% as you think, the killing of the United Healthcare CEO had strong bipartisan support, at the level of ordinary people.