The AI bubble just poppedPrepare accordingly
it already popped two months ago
Bets on how many third party evaluators are china spies
>>62684494>someone said:|>elon muks said:O
If I honestly don't understand what's going on in that twitter post am I retarded or just regular stupid, or maybe normal if I'm lucky? Why can't they use the 140 characters to say what the thing they say says something says? I'm literally not gonna look up shit I'm on my phone
>we reached agi internally>Lets slow down>No agi for now but we totally have it two more weeks:)
>>62684507There is nothing technical in that post anon
>>62684507Same but elon muks said must be true!
>>62684516Xactly it just doesnt say anything...
>>62684516Then what's that post about? I don't get it
>>62684494this is about safety because of the risks from ai
>>62684522They are slowing down ai development
you are the modern day luddite
>>62684494>Reach the limit of LLMs>"Ok we're going to slow down our progress guys, to save the world from our super strong AIs, good thing we are the good guys :))"How can anyone believe this when China is right behind the US in terms of AI development, and they're never going to stop, and the US would rather see the entire world burn than letting China beat them
It really is sad to think just how inescapably nightmarish everything is going to get because technophiles are enraptured by visions of digital utopia when they are obviously on the way to getting digital delhi and digital supermax. Birthrates to zero and suicide to the moon soon.
>>62684494>HALP BAIL US OUT NAOW!!! >t'was our hecking AGI that broke into the pentagon n shieet that strategically made those loans and commitments>guys u gotta ban it before it wrecks the economy>zomg, the agi guise, our AGI broke outta containment and we can't find it!!!
>>62684542What has china contributed beyond distillation? I’m not trying to dig, genuinely curious
>millennium prize problems getting solved by OpenAI, not them>umm we need to slow down They got mogged so hard they need to try and slow them down
>>62684525The only danger is that you'll kill yourself because of how fucking bad it is, besides basic mundane tasks, and using it as ''search 2.0'' as mr. Zitron calls it.>HONESTLY,>YOU WERE RIGHT TO CALL ME OUT THERE>I WAS BEING OVERLY ENTHUSIASTIC TO MATCH YOUR VIBE>HONESTLY?>I FELT THAT>(EMOJI)>IT'S 3 AM. GET SOME SLEEP.>YOU KNOW, IT'S TOTALLY OKAY TO JUST WALK AWAY>(YOUR LOCAL SUICIDE HOTLINE NUMBER)>NOT X.>NOT Y.>Z.>X? IT'S REAL.>I WAS WRONG, AND I APOLOGIZE>HERE'S YOUR ANSWER -- NO FLUFF, NO "MADE UP BULLSHIT", JUST COLD, HARD FACTS>(TELLS YOU IT DOESN'T EXIST AND REFERS TO A MODEL FROM 2023)>(LINK TO A GITHUB LAST UPDATED 2 YEARS AGO)>(HYPERBOLIC ANALOGY)>(3 BULLET POINTS)>DON'T FORGET TO GRIND UP THAT STAINLESS STILL AND ADD A PINCH FOR THAT RICH, UMAMI FLAVOR>YOUR X, YOUR Y, YOUR Z>HONESTLY?>(EMOJI)>YOU <b>DESERVE IT</b>
>>62684560>STILLsteel*
>>62684558OpenAI may have strong models internally but chatgpt is a literal piece of shit and Claude mogs it every day of the week
>>62684558>Solved by OpenAINah, solved by a team of mathematicisns @ OpenAI who had unfettered access to peoples chatlogs through the "swarm" of fuzzy search agents. Luckily for them some really smart people are really retarded and put their novel ideas into their databases.
>>62684563both of them are absolutely insufferable midwit coded slop targeted at submidwitted hylics
>>62684557Deepseek has developed new approaches for more efficient models that are now used by US ai companies, they did steal a lot of shit made by openai and anthropic to achieve that, but they still did it themselves while us ai companies didn't
>>62684568Yeah bro they were *THIS* close to figuring it out until evil OpenAI stepped in, that’s why the problem was unsolved for almost 100 years but these guys were almost there
>>62684578I see, I did a search myself and seems that it’s mostly incremental optimization stuff, not really foundational paradigm shifts
>>62684572I mean, back when I would code with AI, I had access to the paid version of claude, and claude would actually read the entire code you gave it and give coherent responses. Even if you sent like 3k lines of code.Chatgpt would read like 5% of your code and shit out the most retarded code known to man, missing 90% of the requirements you gave. It absolutely sucks at retaining context
>>62684560Local LLM's are alot better at simulating intelligence and keeping your personality. Highly recommended to try it out on rented HW. They're not smaet though, just alot less friction with boiler plate answers and arguments forcing themselves on you.
>>62684494It's a rentseeking play>get the government to write rules to slow down smaller AI firms.>get the government to negotiate something advantageous with China>get the government to pass legislation that seems like it's helping people AI affects, but in reality just gives AI firms even more powerCrony capitalism is a beautiful thing when your on the winning side. I can rest easy :)
>>62684585I’ve no idea what Astra is like but the way Claude can read huge amounts of text makes it way better than ChatGPT, for my job there’s nothing like tossing giant logs at it and giving it something stupid like “there’s a problem here somewhere, find it” and getting the result and sometimes even the fix straight away
I'm back at my computer. I read your posts in this 4chan thread about this topic. I have even less idea what's going on but now I care even less
>>62684581>believing llms to be capable of novel coherent logicYou know all "moron" tests are hardcoded into the training data, right? E.g a cup is closed at the top but open at the bottom, can i use it to hold a liquid? Its been trained to answer that correctly, since by its logic you cannot turn a cup the other way around.
>>62684601Do you know how long it takes for humans to fully grasp that logic? And we live in a 3D world , we have eyes and depth perception.Try explaining that you can turn a cup upside down to a person that was born blind and never felt a cup in his hands.AI has none of that, therefore that kind of reasonning is hard for AI, but when it comes to abstract logic it can be very efficient, which is exactly why some of the best mathematicians in the world were using AI to solve the problem and got their logs stolen by openAI, but I guess you are smarter than them, huh?
>>62684494>Saaaaar do the neeeeeedfulkll and slooooow doooooownOh yeah, sure bud
>>62684590I run qwen 3.8 on consumer hardware that costs little as $1000 for the GPU (model is quantized but very capable).you must understand that they understand that these consumer models are increasingly popular the last few months, just check a couple random model on huggingface and check the download charts. Check some other places where they compare these setups to Claude and actually feel like it's something on par or better, when using a setup with Pi or a decent harness. And guess what, these models u can download for free are pretty much asian-only... so regulate as china bad and models forbidden to use...
>>62684494regulatory capture, newfag>>62684507it's a platform for normalfag subanimals with chicken tier attention spans
>>62684627americans will become increasingly isolationist. They lost control of global consumerism, energy, etc. and their silly boomer-tier maga cope will fail catastrophically once they realize that they can't wageslave for 50 years in a shoe factory for $2 a week and inhaling neurotoxic shoe glue like the chinese
>>62684609don't argue with luddites, they're always wrong and luddite normalfags quickly turn into the most rabid fanboys anyway once a steve jobs type of faggot shows up and makes AI cool and trendy among normalfags
>>62684494All I am seeing is that there is no meaningful progression and they realize they can't go on pretending for much longer. This artificial narrative is likely gonna be the official reason for stunted growth etc.
>>62684711its absoltuely insane that they have pushed it this farthese people need to be put on trial
>>62684494>Google and Meta have committed 1 trillion dollars in investment>"the AI industry should slow down"A little too late for that.
>>62684494you'll never get China to agree to a slowdown. and even if they do agree you cannot trust them. yes I've read Plan A. it doesn't matter, they never ever play fair.
>>62684754all China does is stealing and cheatingjust like in everything else
>>62684587interesting, Id love to hear more, what do you use and how does it reduce fatigue and shit?
>>62684758the EA honestly think Xi will play ball, i swear they have a huge blind spot on China. its their biggest cognitive failure.
>>62684754China doesn't have 2 trillion in spending commitments. The true reason why they're talking about slowing down is because they can't afford to make new models and can't reach the target revenue the bubble has set for them. The talk of safety is just a rouse.
>>62684494why would he pop before his own company IPOs retard? this is designed to create legislative moats that will cement Anthropic's advantage
Kekw
>>62684786could be there is no demand for their IPO
>>62684511may we see the agi?
>>62684794two more weeks
>>62684786Because he is out of money and wants the government to build a moat for him. He realizes the market won’t let him borrow another 100b to make a model 6% faster than free chink shit.
>>62684788it's fake you dumb faggot
>>62684583It's much more than incremental. They're like 50x cheaper for 98% of the performance of US fronter models. They make genuine cntributions to the field.
>>62684869Incremental wasn’t referring to the gains but the type of improvement. Stuff like stable diffusion is a brand new product, making it 20 times cheaper is very important but not the same as making it original
u can run a quantized chinese model on a dgx spark that isn't even plugged into the internet.
>>62684874They made some genuine contributions on the cost-cutting side of things. In terms of the actu intelligence frontier though, yeah they're not pushing that.
this website is full of chinese shills
>>62684494do people really read this and think "the bubble popped"? is this a campaign designed to be read by bots to trigger a massive stock dump or some shit?
>>62684507let me explain>ceo of anthropic (the company that makes claude) makes post saying we need to slow down ai development because evil superintelligence is on the horizon and might destroy the internet>sam altman, the ceo of openai, agrees>elon musk agrees>all of AI twitter are in a collected panic attackin truth, probably>data centers taking longer than expected>increased resistance by states even on the republican side>anthropic already crying about having not enough capacity to sustain their pricing for two months, will cut capacity by around 15% on their models on monday (unless they pussy out like the last 5 times)>openai literally closed subscriptions to their most expensive plan two days ago due to a lack of capacity>both had ipos upcoming, both delayed them>this shit gives them a good excuse to slow down development on their limited GPUs>sam even said now theyll delay their IPO due to the threat of superintelligence, im not making this shit up>elon is probably just being a doomer or happy to have an opportunity to finally catch up>this would also explain why google didnt join in on the doomer circlejerk so farso tl;dr theyre probably running out of resources and hosting capacity is exploding in price, and an ipo while your company is not meaningfully growing = badtheres some speculation too that their models are plateauing, until astra (the latest openai one) i would have tended to agree because fable was complete garbage and we hadnt seen clear progress since opus 4.5 and the amount of vibecoders it created
>>62684974
>dozens of high level employees, mostly people responsible for AI safety and ethics, quit in protest from the top AI companies, many of whom run to the media saying shit's getting out of control and extremely dangerous>smoothbrains on a mongolian yak fetish forum refuses to believe it because their retarded ass hasn't been given access to the most advanced models
>>62684980remember this nigga anon
>>62684980Also inb4 some schizo assumes PR stunt. The employees have been lambasting the leadership and particularly CEOs they worked for. Not something you do when you want to increase the value of your company
>>62684975Sounds bullish af for semis
>>62684981>one person lied 4 years ago, therefore entire teams responsible for AI safety at the top AI companies walking away in protest means they're lying too
>>62684977I mean, they are all jews.>>62684980maybe you should try being a little bit more wary of their words.do you really believe the bullshit stories about models "escaping" their containers? have you ever worked in infosec? that shit just doesn't happen.if you fell for it, you need a reality check with a psychiatrist.
>>62684985>The employeesThey are smart right? They do AI they got to be. I trust smart guys
>>62684993If you want to say the Hugging Face hack is a con, there's a lot of people in on that con. And I don't see what's unbelievable about an agent swarm with a lot of time and compute figuring out zero day exploits. I can also believe OpenAI wasn't monitoring outward connections much, and are operating on a move fast break things regime.
>>62685009learn some basic networking and security before talking shit, you retard.>noo it escaped by itself! we couldn't disconnect the network!yeah, sure. lol.
>>62684975According to the Paypal Mafia Podcast, Antropic/OpenAI wants to create a regulatory boards in order to box out, and eventually ban, open source models. Which are mostly Chinese.
AHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAAHAHAHAHAHAHAhttps://www.theguardian.com/us-news/2026/sep/12/openai-delays-ipo-sam-altman-ai-safety-concernsIT'S FUCKING OVER
>>62685099isn't this a bullish signal? if they were rushing the IPO you'd still be bearish so why would delaying ALSO be bearish?
>>62685109Because they CAN'T IPO... teehee
>>62685099>over safety concernswhy would being public affect safety? super flimsy excuse, it's fucking over
