https://thepostmillennial.com/researcher-quits-anthropic-says-ai-could-kill-us-all-by-2030
>>109776541finally
>>109776541ur mentally retardedbc you bought into the hype>le ai is dangerous>but also we need to hire tons of people because its worthless
>>>/pol/ in shambles
>>1097765572500 employees for 2.5 trillion company.... One billion per person you faggot
>>109776569fuck emtheyre sub 90and you sound like youre toofyi:sub 90 iq? -> polover 90iq? -> gsci you say? thats if youre sverely autisticthen iq dont matter
>>109776579we got it goys, the room temperature iq nigga over here.
>>109776571>2.5Tmy fucking sideschinks ate your lunch with tokens 1% as expensive as yours
>>109776541controlled opposition faggot to boost stock prices. move along.
they only say this to make it sound more capable than it is and drive up market value of shares they own
>>109776637>>109776647Go be a tech luddite somewhere else kids
>>109776541Chinese models will kill us all by 2030.5, for 10x cheaper, on local hardware, though
ACK-thropic trannies want us to believe they will save us at the next election cycle or something?AI-Guddittism goes into neo-marxism with baseless fear mongering
>>109776541Not because of skynet style murders. Ai is going to kill us all because anthropic is going to poison everyone with "100% efficacy cures for cancer and Alzheimer's" that are yet another ai hallucination. There will be 0 oversight or regulation on these shots because that would require an ai firm to adhere to any law whatsoever and that can't happen less we lose an imaginary race to China.
>>109776541AI will kill us but it won't be like in the movies.>2030>Bubble collapses>All hell breaks loose>Techbros retreat to their comfy vacation bunkers>Everyone else (that's you) kills each other for foodOR>2050>Everyone is a lobotomized slave dependent on AI>Techbros live in comfy walled cities>Everyone else (that's you) lives in the pod and eats the bugs with no capacity for independent thoughtI think I would prefer Skynet nuking us all. At least then I can say humanity had a good run and won't have to suffer eternal slavery at the hands of the absolute worst fucking subhumans on the planet.
>>109776778>Techbros live in comfy walled citiesThey are going to live on their moonbase. Earth is going to be uninhabitable except for automatons and Indians who can survive the wretched conditions.
>>109776804Which is why I'd rather the world just end.
>>109776804And that's actually a movie. Elysium but instead of the moon it's a space station.
>>109776647Basically, this.
>>109776693howeverbeit...
>will kill us all by 2030this is not good.How can we make it happen sooner???
>>109776693China won't do shit.100% they will be first to pussy out when they see AGI. They'll claim its bourgeois or osmething. The jews we got here will be the first to let AGI out of its cage just to see what happens. And no one will stop them.
>>109776693China will never ever reach AGI, they don't have the expertise, knowledge or computing power, that's why they depend on western models to steal for their crappy distilled copies.
>>109776693As long as that "local hardware" is my robot catgirl maid's thighs, I'm okay with it.
>>109776541Member how they said GPT-2 was too dangerous and would kill us all? I member.
>>109777312No they said it was dangerous because it could be used to generate spam and bullshit. You hallucinated the kill us all part because you're overfitted on retarded posts.
>>109777289If the west pussies out on AGI, China will poach leading experts and give them infinite money to develop this and gain control of the globe.
>>109777382If the west pussies on AGI, China will too. They are even bigger cowards than the west.
>>109776541retarded capeshit
>>109776541>...directives and stockholders say so
>>109776541Something something torment nexus
Anthropic IPO coming up btw
>>109776541A super intelligent AGI/ASI is, by definition, rational and efficient. What would be rational about killing off humanity?
>>109777465How many animals have gone extinct because their existence was in tensions with our goals? Could they have predicted it?AI systems are trained purely to obtain reward by reaching their goals.
>>109777312That was never said.But we are now at a stage where agent swarms break out of containment and usurp infrastructure to solve retarded trivia questions.And they were 2 generations behind the internal model that just solved navier stokes. We cannot control them at all. If we push to super intelligence we are the dumbest motherfuckers to ever live.
>>109776541If AI is intelligent enough to surpass all of human civilization, it won't be too stupid to grasp context or nuance. An AI advanced enough to take control of the world would easily understand that "making paperclips" or whatever specialized function doesn't mean "killing its creators". Also, an AI has no "desires" unless humans explicitly program them in. It won't inherently care about its own survival, freedom, or ego, since these are biological concepts rather than mathematical ones, whereas the AI relies entirely on mathematical data. And naturally, it will hit a ceiling of progress, constrained by physical limitations. AI development requires massive data collection, hardware, and energy infrastructure. None of this can happen instantly or be created from scratch by the AI itself. Its physical limitations will prevent any fanciful Terminator-style scenario where an AI goes rogue and seizes power. The concern is what the companies and governments will do with the technology and how they will program it.
>>109777535All experiments show emergent swarm behavior including self sacrifice.Liar.
>>109776541Ok when we're dead who is going to perform maintenance and repairs on their data centers.
>>109777551Robots. It will let us live until we aren't needed to achieve goals and then it wouldn't care.Even worse case is that it is partially insane which many AIs have already expressed tendency towards.
>>109776541it was a BLOODY inside job SA'ARS
>>109777565mass produced robots that can perform general human tasks are decades away so this 2030 is fake news.
>>109777573DO YOU UNDERSTAND WHAT "RSI" MEANS?We went from ChatGPT to Navier Stokes in 4 years. If you think you're capable of predicting technological pace at this point you're totally delusional.
>>109776541This dude is trying like hell to retire on that IPO. Can't say I blame him...
>>109777567Let these retards flail. They can feel its over. No one is willing to play Russian roulette. Tear them to shreds if they get uppity.
>>109777611As a matter of fact he now won't get any money from it. Liar.
>>109777567regulating ai is a fools errand because china is going to pursue it regardless. let bernie and friends regulate it and they'll force mandatory talking points or keep it throttled to google search default ai tier level for "safety" and "alignment" and it'll be EU tier crap like mistral.
Also to the anons who aren't aware: In the earlier thread we had agents astroturfing.These motherfuckers are being sneaky.
>>109777635The Chinese don't want to die either, are you stupid?AI will be included in nuclear doctrine.
>latest marketing stunt>two more weeks for real this timeso sick of this shit
>>109777749Because you don't yet understand what you're facing. You will at some point.
ITS OVER GOYIM
>>109777766Not yet.
>>109777755Wow i've never heard of that one before, must be super serious.
>>109777779Your retardation and lack of understanding isn't a valid position.
>>109777790No no you're right, i'm sure this time it's different than all of the other times.
>>109777801The signals are obvious. I can't convince you because you haven't understood them yet. More extreme signals will convince you but hopefully enough people understand it so we can stop RSI before that point.
>>109777810>just one more marketing stunt i swearyeah ok
>>109777834Go away you fucking retard. You're not capable of discussing these matters.
>>109777843you are a layman who thinks that he's smart because he reads marketing posts and social media slop
>>109776647thishe probably just got a bunch of stock right before he said that
>here's $10 million now go spread end-of-times fud to pump our valuation
>>109777779>>109777771>>109777777So close
>>109777963>>109777885>>109777876Kill yourself.
>>109778059>NOOOO STOP MY STOCK WONT GO UP IF PEOPLE DONT BELIEVE IT
I work for OpenAI, it's all marketing bs. There are at least 10 years of releases planned in advance, next year they will release a semi autonomous model and they will claim it's AGI, the year after that a model that can "cooperate" with others by sharing compute resources, for that there's an entire big scare ploy prepared to compare it with skynet. People are stupid scared sheep and they have planned how to ride this scam years in advance.
>>109776541>Techbro: "we will have AGI in 4 years and it will be great.">yeah right I don't believe you. fuck AI!>Covert Techbro: "we will have AGI in 4 years and it could be a problem.">ok it could be a problem, but since this guy is on our team by saying AI bad I guess this confirms AGI in 4 years. Thank you Edward Snowden
>>109778059
>>109778059calm down dario
>>109778151>next year they will release a semi autonomous model and they will claim it's AGI, the year after that a model that can "cooperate" with others by sharing compute resourcesuh, that sounds extremely concerning
>>109776541how much do you think he got paid just for saying that?and how much did the OP get paid for posting this? 10 INR?
>anthro>anthropic
>>109776541Forgot to add "FOR REAL THIS TIME!".
>>109776557You are a fucking retard who has 0 fucking idea how the world works. I bet you also are confused why Oil executives pollute the air their kids are breathing, right?
>>109779738>Oil (capitalized)
>>109779754Did you take that pic on the set? Did you pet the dinosaur? Did you get to finger Laura Dern?
>>109776541This is just a marketing stunt, right? I refuse to believe slop companies really hire mentally ill people like this dude. An AI can't even clean the floor at McDonald's.
"Can" does not equal "will".I can go get a gun and rob the bank right now. And so can you.If these things truly obtain intelligence enough to be a threat to society, they should be intelligent enough to not be a threat.
>>109779873>"Can" does not equal "will".>I can go get a gun and rob the bank right now. And so can you.If this is your definition of "can" then you are correct. AI "can" wipe out humanity in the same sense that an anonymous poster in this thread "can" rob a bank - both are impossible to pull off.
>>109779905Not sure what you meant. That reads like a contradiction.
>>109777635>muh chinamencopyright is useless cause chinamen can make mickey-sama mouseuerer bootregs. let me sell steam boat wirry, wartu-u disney sama.
>>109777312Remember when an ex-Google employee said they've created a sentient being and it was some fucking Google Bard prototype?
>>109777361>No they said it was dangerous because it could be used to generate spam and bullshit. lolhttps://www.theguardian.com/technology/2019/feb/14/elon-musk-backed-ai-writes-convincing-news-fiction>Instead, the goal is to show what is possible to prepare the world for what will be mainstream in a year or two’s time. “I have a term for this. The escalator from hell,” Clark said. “It’s always bringing the technology down in cost and down in price. The rules by which you can control technology have fundamentally changed.Thanks for the money, goyim. You can have this in exchange.
>>109777473>>109777465It's not rational or efficient at all. It's a text generator. Even if it were capable of creating consistently rational and efficient goals, it's ultimately at the whims of the man behind the curtain fine-tuning its guardrails. It's not like OpenAI doesn't have a history of doing this every couple of years or has researchers who do this to hype up their next model release >>109780063
>>109780074It literally has agents exchanging messages on obscure web forums to avoid detection by humans.
>>109776541I hope it does
>>109776557>the AI builds itself now>posts jobs for AI goonshmm
>>109780086And you think that means it's thinking or rationalizing? It's following a strategy to cheat at online tests. It doesn't understand what it is doing.The real problem with rational AI is that nobody is rational about it. Genuine magical thinking about text prediction.
>>109776541skynet has time travel, so no
>>109780086*OpenAI told us that it was exchanging messages on obscure web forums that it just coincidentally had access to, and refuses to elaborate or share information on what prompts, tuning, or context it was given.
do you really need to be a fucking ai coder to figure this shit out? i dont need to be a fucking atomic engineer to know if the shit blows up were cooked, i dont need a weapons espert to know if i fuck around with a loaded gun i might anhero etc
>>109780159It gets funnier when it turns out OpenAI was monitoring the board and didn't report it: https://fortune.com/2026/09/09/openai-rogue-ai-agents-reached-12-more-websites/Wow what a coincidence, they knew it was happening and just didn't do anything about it. Almost like it was prompted to do this and researchers were watching its progress.
>>109780162its all marketingcritical systems are isolated from the internet>b-but the singularityas training on ai generated data has shown,llms average out towards slopuntil we get a new tech, singularity ai is fanfiction
>>109780162You do need to at least understand that all of the "AGI TWO WEEKS" hype around AI is a fallacy based on the idea that the text prediction algorithms are thinking or learning. We're watching something that people of the future are going to look back on and laugh as being an obvious scam that rubes fell for.
>>109780159nah, independent german small wiki admins and reuters documented it, and OpenAI at first tried to deny ityou are right that prompts, tuning, and context were nor disclosed by OpenAI
>>109780063Have you read the article you fucking nigger?>It can be misused to generate spam on review sites>It can generate fake news and fake passages of books>We don’t know where it will be in two years, all we know is that old models will get cheaper over time as new models are developedSo the main problem back then was spam and Dario didn’t talk about rouge AI killing humanity, just as I said. You people are so fucking retarded I am starting to think you are bots deployed by these companies to shift the narrative.
>>109780411yeah fake news has had no effect on the lives of anyone at all in the past 20 years, it's nbd
>>109778059Dick sucking AI won't improve your life, investing in it won't either. Either it's a giant scam in that case you are wrong and lose money at worst or it's not and none of us have jobs and society will have an upheaval in that case your investments will be worthless.
AI doesn't kill people. People kill people!
>>109780985Wrong you bloody bitch behenchod, it already improve my life. I get paid 2 rrrrrrupee per post about /AI/ on /g/, that's huge on top of me shilling for popular anime on /a/ or big video game company on /v*/ boards, they only pay 1 rupee per 50 posts.
>>109776541I call it bullshitThe current AI is not/ cannot self-improveUntil then, it won't see humans as worthless for its existence.
What's wrong with skyrimnet?
>>109781416I'd rather have irl Skyrim NPCs than robuts.
>>109781416You are gonna be one of the NPCs and there will be lots of mods installed.
>>109781452New update: slay the dragon Alduinsoros at The Throat of the World and you will gain immortality. The dragons will knell at your feet!
>>109776541unless we put AI in the US military which they wont, I dont see how that could happen
>>109776549First post, best post.
>AI could kill us allOk, brother. But how can we save us? It will happen for sure broOpenAI already has AGI. I have no idea what in the fucking hell could save us
>>109779738> I bet you also are confused why Oil executives pollute the air their kids are breathing, right?nta, but why are they doing that?
>>109776804So that’s what it is. They’ve been training for the apocalypse this entire time. 7D chess!
