Topic: The Necro Wars (Read 356168 times)

silverspawn · « **Reply #2275 on:** January 05, 2022, 11:28:18 am »

I'd be surprised if I get anyone to study it, but that should work. But the problem is more the network stuff. Data security and cryptography are difficult to compress but at least topics I know something about and like

silverspawn · « **Reply #2276 on:** January 05, 2022, 11:34:08 am »

I am considering just asking for a good textbook on internet protocols or something like that. I should read more books in general. Learn more things other than AI and math

silverspawn · « **Reply #2277 on:** January 05, 2022, 01:29:50 pm »

So supposedly Don't Look Up is a movie about existential risk. Gonna see what it's all about.

Ignored it before because it looked dumb on the titlecard, but this is super unreliable.

silverspawn · « **Reply #2278 on:** January 05, 2022, 01:53:04 pm »

Yeah this is really good.

silverspawn · « **Reply #2279 on:** January 05, 2022, 02:10:16 pm »

It's extremely frustrating, which I know is the point. It captures well how much people care about credible xrisk

silverspawn · « **Reply #2280 on:** January 05, 2022, 02:13:28 pm »

insanely frustrating jeez it's too real

silverspawn · « **Reply #2281 on:** January 05, 2022, 02:28:14 pm »

I can't believe I'm saying this but I think there is at least a 1/3 chance that Trump would have done a better job than the president in this movie

silverspawn · « **Reply #2282 on:** January 05, 2022, 02:35:02 pm »

All of this is super well done but it's not psychologically healthy for people like me

silverspawn · « **Reply #2283 on:** January 05, 2022, 02:36:26 pm »

I just hope it ends with humanity wiped out because fuck humanity

silverspawn · « **Reply #2284 on:** January 05, 2022, 02:38:18 pm »

god I hate all these people

silverspawn · « **Reply #2285 on:** January 05, 2022, 02:46:07 pm »

I will say I do think this is probably too cynical but maybe that way it sticks better

silverspawn · « **Reply #2286 on:** January 05, 2022, 02:50:25 pm »

Seriously this has to end with humanity being wiped out. If you show us fucking up to this degree and then it works out, that sends all the wrong message

silverspawn · « **Reply #2287 on:** January 05, 2022, 03:11:19 pm »

Please let this have a bad ending please

silverspawn · « **Reply #2288 on:** January 05, 2022, 03:17:20 pm »

This is so difficult to watch jesus

silverspawn · « **Reply #2289 on:** January 05, 2022, 03:18:36 pm »

Maybe Trump isn't so bad after all

silverspawn · « **Reply #2290 on:** January 05, 2022, 03:26:40 pm »

This seems like a relevant quote:

If people try this repeatedly with other corrigibility training tricks on the level where plots are easily observable, they will eventually find a try that seems to generalize to the more complicated and intelligent validation set, but which kills you on the test set.

A way to die with less dignity than that is to train directly on what should've been the validation set, the more complicated domain where plots to kill the operators still seem definitely detectable so long as the AI has not developed superhuman hiding abilities.

A way to die with even less dignity is to get bad behavior on the validation set, and proceed anyways.

A way to die with still less dignity is to not have scaling training domains and validation domains for training corrigibility. Because, like, you have not thought of this at all.

silverspawn · « **Reply #2291 on:** January 05, 2022, 03:32:47 pm »

God I can't do this

silverspawn · « **Reply #2292 on:** January 05, 2022, 03:44:10 pm »

(real spoiler) The film made me hate humanity but the fact that it was made is obviously a good thing. And yes, bad ending.

silverspawn · « **Reply #2293 on:** January 05, 2022, 04:50:06 pm »

Ok everyone is saying the president is modeled after Trump. I guess that is true. Probably I just forgot how awful Trump is and this is more visceral

silverspawn · « **Reply #2294 on:** January 05, 2022, 04:54:16 pm »

Anyway I rate this unironically a 10/10

silverspawn · « **Reply #2295 on:** January 06, 2022, 06:38:21 am »

I the amount of chess games that you try to win by putting your knight on c2 (as black) or f2 (as white) as a measure for your intrinsic worth as a human being. Less is better.

silverspawn · « **Reply #2296 on:** January 06, 2022, 06:41:15 am »

Artosis always says that the way someone plays starcraft says a lot about them as a human.

This sounds like a meme or character, but I suspect it's quite true, and not just for starcraft. I'm much less sure what exactly the correlation is, but it does take a different kind of person to play terran/protoss or (admittedly to a lesser extent) to try/not try winning by putting your knight on c2.

silverspawn · « **Reply #2297 on:** January 06, 2022, 06:43:42 am »

Unrelated, and it doesn't affect me anymore, but I think dominion tournaments with unlimited thinking time are a bad idea.

Why? Because you're rewarding the extent to which someone has no social scruples. And complaining about this immediately makes you an evil person, so naturally everyone says that Of Course It's Perfectly Fine To Take Forever On Every Move.

You can do 30 minutes per game plus 5 seconds per decision or something. It can be a generous limit. Just don't do no limit. Bad incentives.

silverspawn · « **Reply #2298 on:** January 06, 2022, 12:29:48 pm »

The way in which I feel like society is significantly more competent than depicted in Don't Look Up is that plenty of billionaires are not, in fact, idiots.

Some are, for sure, like Zuckerberg. Probably more than half. But not all of them. Vitalik Buterin, for example, the inventor of Etherium, and he's just a phenomenal guy. Gates is not an idiot. Musk is... arguably responsible for making things a lot worse, but obviously not a bad guy.

The most comprehensive single book that makes the case for AI risk is Superintelligence by Nick Bostrom, and I'm pretty sure all of the three people I just named read it. Bill Gates' recommendation was printed on the cover. Yet for some reason, he's still doing global poverty stuff instead of the thing that actually matters. Musk read it and wanted to do something, and well he fucked it up about as bad as was possible. But it still gives me hope because it feels like things could have been different. Maybe this Everet Branch is doomed, but there should be a bunch of other worlds that are doing better. One where Musk founded CloseAI and Gates poured billions into AI safety instead.

silverspawn · « **Reply #2299 on:** January 06, 2022, 12:32:07 pm »

And maybe one where Many Worlds was considered first and no-one ever took Copenhagen seriously, and maybe society goes from giving zero fucks about many worlds to giving just slightly more than zero fucks about many worlds

Dominion Strategy Forum

News:

Author Topic: The Necro Wars (Read 356168 times)