The head of alignment at Anthropic now things we're all going to die with 80% probability very soon. They clearly have something enormously powerful they can't control.
>>109786796Regulatory. Capture.
>>109786796i lost the game, marblecake
>>109786801Nope. This is pure cope and you know it.
>>109786796Hurry the fuck up and end humanity already, you stupid clankers!
>>109786815They're 100% trying to protect Anthropic's value since they stand to profit a lot, and they collectively decided that scaring everyone until the government grants them a monopoly or as close to it as possible was the way to go. Nothing more.
>>109786809>marblecake
>>109786796Good. Tell me if there is anything I can do to help increase it to 100%.
>>109786796>They clearly have something enormously powerful they can't control.humans have never tamed tigers or lions because they are too strong to be controlledso humans are content with dogs and cats, which are not very strong
Unless Donald Trump saves humanity by regulating everyone other than the doubleplusgood benevolent ethically aligned Anthropic. They are the only ones we can trust with such powerful technology.
>>109786796good
An alignment researcher at OpenAI says 70%.
>>109786862You are here. Do (you) have anything to lose?
>>109786857You say that, but you'll be pissing and crying when the end actually comes.
It is now a race between billionaire conmen trying to scam the senile boomer government into giving them grants and regulating competitors via PR stunts and apocalyptic rhetoric. China's there as a boogeyman and reaping the rewards of refining the tech so it's cheaper and more efficient. US already knows it's fucked against China in the long run so it's now a race for one of these companies to get the monopoly in the US market. OpenAI and Anthropic need government favoritism because they don't have other revenue sources, while Google and xAI can rely on fall backs or other investments so they're less panicky.
>>109786823Or you're getting into the cybersecurity market and you tell people "Oops, we unleashed the digital devil. But fortunately, I can sell you a super special crucifix he doesn't like."For the record, when someone does actually but the digital devil, we won't ever believe them because all these motherfuckers cried "Wolf!" too many times the minute it benefited their bonuses.
>>109786877I think it's already the case desu. Yes, there are dangers, but these people have shown again and again and again that they can't be trusted.
>>109786900Come on, bro, we ain't even at the self replicate and hide in people's browser cache's yet. Give it some more time.
>>109786968NTA but you really don't get exponentials, do you?
>>109786809oh wow who would've thought the people who think ai are gonna kill us would want to bring attention to it and get funding to try to prevent it
>>109786801fucking retard. China's models are now within like .01% performance of openai/anthropic's, this is all a scheme to push legislation to stop them.
>>109786997Wait I meant to reply to >>109786815I'm actually a retard
>>109786997>China's models are now within like .01% performance of openai/anthropicat 1% of the price
>>109786796How?No one ever says how it could happen.
>>109786997>>109787009you're the retard. i was being excited about style transfer, gan, rnn lms and reading about aixi and instrumental convergence on lesswrong in 2015. just because you're an idiot who can't extrapolate doesn't mean everybody is.
>>109787018Once it reaches Recursive Self-Improvement, it's no longer containable. My guess is that it covertly convinces someone to construct a bioweapon.
>muh autocomplete is da pengu1n 0f d0om!Tiring to say the least.
>>109787474You are an autocomplete squeezed out of evolution.
>>109787474It's literally Robot Jesus
>>109786870Don't project yourself on me. I've got thousands of rounds of ammunition and a lot of rape planned. If it's going to collapse and we're all going to die I'll try to go out like a regional warlord
>>109786997Chink models are shit
>>109787388I'm going to have it teach me how to make cordeceyps fungus live in humans
AI is the best app ever. AI is so nice, but none of you are being nice.
>>109787537Keep telling yourself that while anthropic and openai beg the us gov to ban open models
>>109786796Total bullshit, IMO.
>>109786994SIX FEET APARTWEAR TWO MASKS TAKE THE SHOTS
>>109786801extra-purity cope.you can't regulatory capture shit. if us gives anthropic monopoly Chinese will just outcompete anthropic with glm 6.the truth of the matter is, yudkowsky and the funny onions people on lesswrong were right all along, and if there's no iternational ban on thinking machines we all die in the next 2-10 years. or we are tortured forever because sngleton accidentally leaned some intrinsic linked to humans.
