HAPPENING!AIPOCALYPSE IS UPON USRAM MANUFACTURERS WILL CRAWL BACK TO GAMERS WITHIN THE YEAR
>>732284202And they'll get away with it too.Grim.
What alternatives? There are none remotely as capable for AI
>>732284202doesn't mention ram at allall fields
>>732284979>What alternatives?Some chink start up that will buy up and hoard all of the rare metals that go into manufacturing, essentially leaving us where we are now, but providing slightly cheaper chips to AI companies.
>>732284979Intel
video games?
>>732284202There are no alternatives>>732285283Intel is a joke
>>732284202this sounds like they know they can’t pay and just want to tie up other manufacturers again expecting not to pay
>>732285494>RAM>Not Vidya related
>>732284202What alternative, wtf are they even talking about
>>732285802You're right, it's not. Take this shit to /g/.
>>732285283Hasn't been good since they left the US.
3dfxGODS, is it our time?
>>732284979It's theater. The deal they were doing is falling apart. This is so they can gracefully back away without losing face.
Damn, even the best chips aren't good enough to create AI GOD. Oh well, good luck on your search!
>>732285657>Intel is a joke.Products Intel allows you to buy?Yes.Products Intel makes that costs tens of thousands of dollars? No, that shit is fucking top notch with a BIG O T.
If the chinks somehow weasel their way in, Jensen will kill himselfAll for it
>>732286113Honestly, it'd be good for the industry if there was just an agreed upon upper limit for hardware that all devs adhered to in terms of spec targets.>all devs make games that run on 1080Ti or equivalent>no ray tracing memes>gaming is now insulated from AI faggotry>etc.
>>732284202That just means massive subsidies using taxpayer money.
>>732284202>The bubble explodes because the technology and energy infrastructure are not up to the task yet KEK.
>>732286417So Intel makes good GPUs, they just won't let anyone buy them? Great business strategy.
>>732286671That's not gonna be a fun ride while it happens.
>>732284979>he doesn't know about the Elite-only gpus
>>732284202openAI suck dick, i'll take the chinese ai any day
>OpenAI and Nvidia make 6 gorillion dollar deal>rumors emerge that OpenAI may not have the money to pay>OpenAI: Uhm, yeah, the chips are faulty and we don't want them anyways
>>732284202unsatisfied in what way?im willing to bet financially, nvidia probably jacked the prices right up for OpenAIwouldn't be shocked if nvidia goes to talk shop with Microsoft who has infinite fuck you money for dumb shit like this
>>732284202Not immediately. It'll take a few years.I hope nVidia eats shit and loses market share out of the whole thing, they've been too big for too long.
>>732286758Anon, people do shit with computers other then play gay as games, I am sorry you bought into spending 5 grand on a gaming PC and only play Half Life 2, the rest of us use them for things like 3D modeling, rendering, video editing etc and the game trash comes second.If you care so much, just buy a Dreamcast and replace the optical drive with a GDEMU and play all the good games you like.
>>732284202Leather jackets aren't bullet proof btw, in case anyone here was wondering
>>732284264I mean anyone is welcome to market products to me when they're affordable. I see no reason to hold high prices from five months ago over them if they aren't like that presently.
god please let gpu prices fucking crash i want a 5090
>>732287139whats the real number, not gorillion?
Fuck it I’m buying a Zotac 5090
>>732287159More likely whats happening is their capital is evaporating as that news that they'll be insolvent in a year hit the market. See >>732287139>>732286819Unfortunately this is true. A lot of the global economy is leveraged in this bubble. The AI frenzy is unironically WHY we've been seeing growth for the last few years. When OpenAI pops, the entire market sector is gonna drop out (you already see companies like MS getting leery as they're eating shit as unlike OAI they have to turn a fucking profit and turns out consumers hate all the AI in windows.)This is gonna cause markets across the globe to deflate, ontop of a period where we should already be having deflation thanks to general (normal) market forces and [CURRENT EVENT]. Don't freak out, but make sure yo bonds are diversified. This would be the time to start pumping the breaks if you've been riding high risk yields.
>>732284979Google's TPUs are better
>>732287304I wouldn't even trust an Intel CPU after they made all those defective ones.
>>732287651Tungston Processing Units?
>>732286460that's called VGA and it turns out that people want better shit than whatever frozen spec your prolapsed anus of a brain thinks up
>>732284202>altman got assblasted nvidia backed out on their deal because he was lagging behindThis has nothing to do with Nvidia, OpenAI is fucked lol
>>732287532it's been 6 gorillion all along
>>732284202Its funny watching you people learn about economics for the first time in your life. Supply goes down and demand goes up means price go up. Woah. Big brain take I know but I'm sure some of you can follow the reasoning.
>>732287707Somehow the ones made in Tel Aviv worked fineI got lucky myself, will get an AMD mobo next time though. I'm genuinely getting fucked in games over the X3D benefits nowIf AMD added e-cores i'd buy their CPUs exclusively for the rest of time
>>732284979There aren't any. This is desperation over Nvidia pulling out of their 100 billion deal with OpenAI.
>>732284202>alternative>for cuda coreslmao.
>>732284202I LOVE ADS IN MY AImakes me want to use it even more
>TSMC and Nvidia - we're not going to build new FABs to satisfy AI craze because we don't believe there will be such demand in 5-10 years>FUCK YOU, YOU"RE RUINING AI WE DON'T EVENT WANT YOUR CHIPS!lol, lmao
It's nearly like making all working libraries reliant on a private owned tech was a short sighted bad idea
>>732287532a supposed 100 billion dollar investment deal was publicized September last year, but it was just intent nothing finalized. However, part of the stipulation was obviously that OpenAI use nVidia tech and chips, because that's their ROI. However, 2 months later in November, OAI SIGNED a deal with AWS (Amazon Web Services) to the tune of $38 Billion to allow OAI to use AWS Data Centers. In other words, rather than building out their own infrastructure, they'll be renting it out. This makes economic sense for AWS, and ironically those Data Centers USE nVidia chips. However, if OAI isn't gonna be building 100 billion worth of new data centers with new chips and ram, it's not worth it for nVidia to invest - they'll essentially be cannibalizing themselves. nVidia doesn't like this and is also getting gunshy of OAI's 'business discipline' (AKA the news that they're near insolvency and haven't turned a profit), and now reports are saying they still plan to invest but the proposal has shrunk by 80% to 'just' 20 billion.
>>732287974Rewrite this in proper English.
>>732287532The number in the media was 100 billion but that's as accurate as 6 gorillion because both numbers are just made up. There was no real deal either, just an announcement that they want to push a lot of money around to get investors interested. Basically market manipulation.
>>732284202>Scam Altman tells Jensen he can't pay in time>Jensen pulls his fat billions out of his ass>Uh, I didn't like you anyway. Now, who can I Scam next?
>>732288017>>732288169that's because OpenAI has no money. They can't invest 100 billions because they don't have 100 billions
>>732284979They're just giving themselves excuses to delay things so their loans of infinity million dollars can get pushed back.
>>732284202You think sam altman realizes hes going to get epstein'd when this shit crashes and burns with nothing of worth to show for it, given all the money hes completely vaproized(or embezzled)?or does his retarded inbred jew hubris shield him from such annoying concepts as 'the future'
>>732288017OH, also there's AI specific chipsets being developed by start-ups>>732288065the consolidation inherent in AI's high cost is destroying the open market that these companies are reliant on to continue development. If all the AI chips are made by nVidia then you have to use nVidia and nVidia has you by the balls. Meanwhile, AI 'vibe coding' is also killing open source software by (possibly illegally) referencing source code to non-licensed open source programs that are ubiquitous in coding and back end, like the chatGPT problem with books all over again. This is causing the smaller companies that make these open source softwares to collapse, thus being unable to update or develop these softwares, thus causing the environment dependent on them to die. It's dead internet theory but with software.>>732288252It was nVidia proposing to invest in OAI, with the idea that they would use that money to buy nvidia products in order to build infrastructure. But OAI doesn't need infrastructure in 10 years it needs it NOW, so it signed a deal with AWS (to use the same products it was going to buy from nVidia). This pisses nVidia off because suddenly all that Ram they were gonna sell to them has no buyer. The 100 billion dollar deal was only ever vapor anyway, drunk business guys bullshitting over martini lunches.>>732288328How many of these people ever get accountabled? No one has ever gotten Epstien'd but Epstien, and he's the outlier for actually getting comeuppance, barring maybe Harvey Weinstein
>>732288328he will retire as a billionaire, like they all do, there never are any real consequenses for """them"""
>>732288328Trumpstein will pardon him for 10 million dollars
I think that the hype around LLMs have mostly been a cope of the west and India that they have been utterly yeemed by the Chinese when it comes to making and inventing actually novel tech.
>>732284979AMD, Intel, Broadcom....
>>732288763>the chinese>inventing anythinglmaonigger they still cant make actual steel
>>732288065All focus has been made on using Cuda core, proprietary tech from NVIDIA that isn't even meant to do AI to begin with. At this point rewriting libs with other tech would halt all AI progress for a year or two but nobody wants to do it..
>>732288869All they have to do is reverse engineer a gpu
>>732284202Waitchad here, I will be enjoying some cheap 5900 xt and 69GB RAM
Why do luddite bastards want AI to fail so much? You're only embarrassing yourself throwing hissy fits like this.
>>732287139OpenAI had that early bird advantage, but there was no way they were going to be able to compete with something like google. They don't seem to have any particular talent left either after it all got poached.
>>732289464I want Open AI to fail. I'm using Opus and Gemini 3 daily doesn't mean I can't hate AI companies too
>>732289464Bloody banchood bitch bastards for real
>sources sayMeanwhile, it was Nvidia who backed up from a contract worth billions for OpenAI because they've proven to be incompetent beyond measure.OpenAI is going to end up being the Myspace of AI.
>>732289464It can fail as much as the internet failed, the dotcom bubble popped but the internet marched on, it will be the same for AI
>>732284979Blame America for destroying Japanese chips on purpose to the extent that Japan was forced to ditch chips and instead, focus on manufacturing semiconductor chemicals
AI can't really fail, the demand is too high. I interact with quite a few businesses daily and they are almost all using it.
>>732289464>sir there is being so much porn!I want to upgrade my computer.
>>732289464PAKI BITCH BASTARDS WHY THE AI, I FUCK YOUR MOTHER PLEASE? THANK YOU HAVE A GOOD DAY
>>732289464>AI IS FUTURE LUDDITE BITCH BENCHODS>I WILL PROOMPT ENGINEER FOR ELON!!!
>>732288328Altman's sister already claims he committed incest on her, so why bother mentioning Epstein at all? However, by sheer coincidence, Altman announced he was gay after his blood related sister made the incest claims>In a lawsuit filed in the U.S. District Court in Missouri, Ann Altman alleges that the abuse occurred between 1997 and 2006. She has made similar claims of abuse against Sam and another brother, Jack, on social media previously.
>>732289808To do what?
>>732286861If only OpenAI hadn't censored the damn bot, I'd probably have subscribed to their shitty service. But no, now 5.2 is a fucking prude when 5.0 and even 5.1 would gladly describe rivers of cum leaking out of a mature catwoman's pussy.Deepseek is good too because it's uncensored, but it's just not the same.
>>732289464BLOODY BASTARDS WHY ARE YOU NOT REDEEMING THE AI?
>>732289980To give them instructions on how to achieve certain results in third party software they use in their business like Quickbooks, Salesforce, Solidworks, Adobe Premier, etc.
>>732289758>Blame AmericaHmmmm, no thanks, Thirdie-kun. The current state of tech reeks of Chinese meddling, so I’d rather blame them first. Especially since their recent opportunistic announcement of “totally cheap” RAM and chips.
Where are all the retards claiming AI is the future and it isnt a bubble? Because its starting to pop right now because, to no one's surprise, AI companies still don't have a way to make money.
>>732285851shut up nigger
gaming is collapsing/healing (not sure which)
>>732287572>A lot of the global economy is leveraged in this bubble."Global economy" is a stretch, there is only a single economy who's going all win death pact on AI.
>already hit diminishing returns a year ago>trying to blame the hardware for itlollooking forward to the whole house of cards collapsing in the next year or so. Hope all the retards who invested into OpenAI go bankrupt.
>>732284979I dunno, but that bubble film is getting AWFULLY thin at the apex
>>732290319There is a way for AI companies to make money but it will be from businesses and rich individuals, not by inserting AI answers in search results. That kind of thing will go away since it isn't profitable.
>>732289758>for destroying Japanese chips
>>732290319IT DOEDSNR NEED TO MAKE MONY LUDITE FUKING BASTARD ELON SAY NO WORK FIVE YEARS FOR AI! YOU BLOODY! no but for real anyone who didn't see this shit being VR2 was fucking retardedit has more use than VR, admittedly, but anyone thinking this shits anywhere near ready for a fraction of the applications they're selling to retard low/no-info investors is either a scammer or a desperate shitskin(but i repeat myself)everyone got fucking tard-blinded by the "progress" it was making, like the will smith eating spaghetti videosyea, huge improvement from ABSOLUTE RANCID DOGSHIT to "expected of a modern day technology"VR had a same progression, again, howd that go? oh right>Cant go around it, cant knock it down>THE WALL, what? THE WALL
>>732290289So it's a search engine that shows you photoshop tutorials except it costs a shitload of money to run.
>>732288769None of them are serious competitors to Nvidia when you need parallel processing. GPUs excel at parallel processing, and Nvidia's GPUs are by far the most powerful available and moreover are specifically geared toward AI datacenters at this point. Nobody else has Nvidia's scale or speciality. AMD would be the next closes thing but they cannot hope to match Nvidia's sheer output and are behind in optimizing GPUs for AI purposes. This is pure smokescreen by OpenAI to try and get ahead of this year's bad news.
>>732289464I would LOVE to see AI succeed, man imagine the post scarcity society that AGI would bring, or the Armageddon either one worksBut unfortunately we have no AGI, nor AI, we have akinator that develops dementia after 2 minutes of use
>>732289808AI isn't going to fail. AI is absolutely here to stay. But the corporations behind it like OpenAI can absolutely fail or get bought out.But people thinking Nvidia is ever going to be a gaming company again are utterly delusional. They are the ones selling the shovels. They're the only ones utterly secured in this race.
All of this is the equivalent of a squabbling couple telling their friends that they're looking for another partner so that the friend will tell the current partner to wake up and put in more effort, be jealous etc. They're not going anywhere and there's not likely to be a better solution financially or performance wise they'd accept but they'll make a big fuck you deal about it on the media.Not to mention all the speculators on both sides who want to buy the dip and both sides can say >SEE WE GOT SO MUCH MORE INVESTMENT ITS FINEwhen they do, bringing the valuation up. Its all a scam.
>>732286819It's not going to be a fun ride for the corpofags and grifters. I, however, am going to really enjoy the age of dirt cheap cloud compute and fire sale priced racks of H200s hitting the market.
>>732288763Chinese don't innovate though. All they do is copy whatever the west's cutting edge tech is then find a way to make it more cheaply, usually by cutting as many corners as possible and controlling the entire production pipeline. China's last big "innovation" was basically finding a way to run ChatGPT with a fraction of the power requirements which they then pretended was a brand new AI model (that still introduced itself as ChatGPT when you asked who it was).
>>732290631>Doesn't even know Adobe Premier isn't PhotoshopBut ignoring that, the average worker is too dumb to take tutorials and apply them to their own situation, for maximum productivity, average workers need something that tailors its responses to their questions. Otherwise it takes them much longer to figure out how to do something.
>>732289464Because LLMs are not AI and LLMs are what is driving 99% of AI investment.
>>732289730The internet was quietly making money while the dotcom retards inflated their company's worth with empty promises. AI doesn't actually make money. At all. Its costs far, far exceed its revenues which are either nonexistent or minuscule.
>>732290805Why would you want to buy cloud compute? Or do you mean the hardware for it, not the service?
>>732290810>meanwhile BYD is taking over
>>732290870luddite benchod bastard it just like ghost in shell! no any different! ai is future!
>>732285494This nigger playing tiger electronics
>>732290923>its costs far, far exceed its revenues which are either nonexistent or minuscule.It makes money in specialized field where people pay money for it. It loses money when you allow Indians to generate porn/scam videos for free
>>732290870AI is worse than LLM, because if actual AI came about, Judgment Day is only a few years away. At least LLMs can't watch Terminator 2 and get some cool ideas.
>>732290938BYD is literally exactly what I'm talking about. They just copied western EVs, stripped out everything but the bare minimum features, then fully integrated all steps of the manufacturing process into one company to bring costs down. BYD doesn't do anything innovative, it wouldn't exist if western (and Japanese) auto companies didn't figure out the hard parts first.And you're welcome to buy a BYD if you want, you get what you pay for though. You're going to find out why those cars are so cheap.
>>732291036It doesn't make money, period. There is no instance where AI actually makes more money than it costs. That doesn't mean AI isn't producing good results in research, for example, it's just not cost-effective.
>>732291036>it makes money in specialized fieldloltell me ESL of the AIhowwhat fieldshow does this dogshit upjumped google search make money
>>732291036Even anthropic isn't profitable yet.
>>732291167LLMs are money blackholes, but there is actual uses for machine learning in genetic and chemical research, simulating protein combinations. But the only reason people can even do this is because it is heavily subsidized by massive funding dumped into LLMs. Without the LLM-driven investment, laboratories using actually useful machine learning would not be able to run these models.
>>732291161Do you have some studies to back this up? I mean for example generative AI definitely costs less than hiring an artist to draw something.
>>732287139that's actually funny
>>732291291I don't think you're smart enough to have this conversation if you don't even understand what the "cost" I'm talking about is. The cost of running an LLM data center is far, FAR more expensive than hiring even 10,000 artists.
>>732290457unfortunately it's the biggest and the one whose currency the global economy is based on. And this isn't 'murrica chest beating, the USD is the standard
>>732284202>CRAWL BACK TO GAMERSi don't know how to tell you this but gamers have always been a shit market. same thing with smartphones, that's why Plus One and Asus are failing. They were supported by completely unreliable market that believe to be tech enthusiasts. tech enthusiasts that believe they deserve cutting edge tech are low prices are always delusional.
>>732291291Depending on how AI survives this copyright lawsuit, that might not be for long.
>>732289464bloody bastard bitch saaaaar
>>732291291Which company is making money?
>>732291291AI is expensive to RUN not to USE.
>>732290661>imagine the post scarcity societyNigger that replicators shit is for the elites to enjoy, you'll get genocided and replaced by indians that in turn will be replaced once robots can do everything without breaking within a week. Have you forgotten how they want to reduce the population and stop you from eating meat?
It's becoming more and more clear that Google is the winner of the AI race and these other new AI companies are going to be absolutely demolished and NVIDIA is coming back down to earth
>>732291291>I mean for example generative AI definitely costs less than hiring an artist to draw something.This is the most retarded thing I've read on this board this year so far. Congratulations.
>>732284979There aren't any alternatives, but this at least lets them delay major financial questions for a bit.
>>732291539The AI race is a race to the bottom. The only winners are the ones who abandon ship before they're sucked down with the wreckage.
>>732291589Google can easily keep turning profits with the ads they already have, everyone else will die trying
>>732288601After getting egg on his face for an economic crisis during his presidency?Really, who WON'T want to throw Sam under the bus?
>>732291393>>732291504>>732291542You don't need a data center to make art from generative AI. You can do it on your own PC with a single good GPU. If I were to go on Twatter and hire an artist, it would cost more than a tiny bit of electricity and a bit of wear on the GPU.>>732291478I'm not talking about huge companies, I'm taking about individuals who just need a piece of art for say marketing or the like and can run generative AI on their PC.
>>732291528Yeah anon, maybe read my post and you will see that i did address that option.
>>732291718Don't bother arguing with them, anon. Just stay ahead of the game, like me.
>>732291431>but gamers have always been a shit marketComputers have evolved because of gamers, were it not for videogames computers would still be MDA monitors with a keyboard glued to it
>>732291797And now the time has come for the leather jackets to shed their gamer skin and become Mega Gorillionaires.
>>732291539In hindsight it was obvious that google, the company with the most info on anyone out there would be the company that wins with a technology that requires massive fucktons of info to even function even remotely competently. Oracle are patent trolls, microsoft is an OS company, openai/anthropic is a meme ect.
>>732290810It's sad how the USA are still mentally in the 80's are adamant nothing has and will ever change. you are basically 22 year old war vet in your head
>>732291420The dollar is world's reserve currency only because of the protection racket that the US does, they don't have anything that couldn't be achieved from someone else.The whole world is biting their time to brick the US into oblivion.
>>732291797im actually curious just how different things would be without PC gamersyou'd have independent video and photo editors as well as animators and 3d modellers but that's nowhere near enough to be a big consumer market for some tech
>>732284202Reminder to burden AIshit daily with dumb questions (like typing 2 words from a thread title) and say thank you
>>732291864MS makes more off Azure and Office than Windows, anon.
>>732291718Generative AI has hit a wall since 2024 and can barely even consistently make original pictures sequentially like an artist of even moderate skill can. GenAI is also a sign that you're a cheap-o who is looking to scam people, it is very good at that job but not much else.
>>732291282Machine learning in medical and scientific fields has been a legitimate field long before the AI craze, when people refer to AI they're referring to LLM's and Gen AI slop which has yet to have any real value besides making the population more retarded
>>732291589Google (Alphabet actually but whatever) has spent less on AI in total than they earn in profit in a single year. Anyone else trying to compete or who is investing in OpenAI other than giants like Microsoft is retarded BUT MS and Google can afford to keep this going for a while.Its not so much about getting the best product on the market right now or even earning a profit, its about making sure that your biggest competitor doesnt find the golden needel (which may not even exist) in the haystack with no competition.Fear of missing out is driving the AI bubble and its hard to predict when this fear will run out
>>732292070>a sign that you're a cheap-o who is looking to scam peopleSo anyone in business?
>>732292132Basically.
>>732284979God /v/ really IS clueless about AI.This anon here >>732287651 is right plus they already made a deal with Cerebras which is miles faster and more efficient. There is also Groq but it was bought by Nvidia.>>732284202>RAM MANUFACTURERS WILL CRAWL BACK TO GAMERS WITHIN THE YEARGPUs, not RAM you moron
>>732292098I predict the fear runs out when investors get truly burned for the first time, then Google will be questioned why they are still sinking money into a black hole instead of whatever the next tech investment craze is.
>>732284202>Open makes "deal" with nvidia where nvidia is supposed to invest 100 billions into open AI>Jensen publicly says that deal was a non binding nothingburger and that nvidia isn't giving them the money>Sam Altman melty ensuesWhat the fuck is he gonna do? Nothing beats cuda cores for training LLMs which let's repeat it aren't AI at all to begin with.Leather jacket has his by the balls.
>>732286819Probably not but some times it´s not about one winning but the other guy losing. The future AI holds is just not the future i want. >INB4 too bad, it´s happening anyways Sure, but every delay is appreciated. I´ll settle for cyberpunk remaining fictional on my time.
He's still going to buy up all the high VRAM cards with other people's money. Just because he can.No, he doesn't have the infrastructure to take them online. He just doesn't want anyone else to have them.
I like AI art
>>732284979They'll build ASICs for AI, mark my words.
>>73229232816GB VRAM is the max gamers will ever get, bro
so how come people seethe so much at a bunch of dollars spent on foreign countries but no one seems to mind how a much of billionaires are practically siphoning billions from the state budget that could be actually invested to improve the country
>>732289464NIGGER I WANT CHEAP RAM AND I DON'T CARE IF IT EVAPORATES THE ENTIRE STOCK MARKET
>>732292070>Generative AI has hit a wall since 2024that's actually true, the last real improvement was CoT/reasoning, gpt 5 isn't much better than gpt o1/o3/o4.
>>732292382with DLSS and framegen that's all we'll ever need anyway
>>732292461Fake frames introduce input lag
>>732284202>HAPPENING!>>>/pol/
>>732292198>I predict the fear runs out when investors get truly burned for the first timeBurned by what exactly? Google buit its empire on investing in new tech, shit like the Stadia ans Google Glass failing didnt really matter.Granted those were much smaller investments but what exactly do you predict could happen that would totally kill / replace the interest in AI?Google makes phones and Android, it would be very easy to justify AI investement just off of the fact that they want the household AI built into every phone to be theirs so they can control what content it shows customers.
>>732292497input lag is a meme
>>732292592retardget the fuck off /v/
>people are stockpiling RAM right now>bubble pops>dirt cheap DDR5 floods the marketI can't wait!
>>732292381you need both hardware and software, CUDA isn't something that you can build in a year, it's not like dumb crypto mining where all you have to do is hash numbers.
>>732292550Burned by google’s competitors violently failing? If Open AI and everything else in the sphere folds all at once, it’s probably going to absolutely annihilate investor confidence in AI-anything and thus create shareholder pressure on Alphabet to spend less into research and development.
>>732291857There's no money to be made in AI unless you're a scammer swindling boomers out of their money with ponzi schemes on twitter and facebook
>>732290870Literally nobody wants the AI you're talking about. We want a tool, not to give birth to a new species that could kill us before we even realize there's a threat.
I want my damn 5090...why did I wait for supers...
>>732290303>Don't blame America!>Blame China instead!Nah
>>732292639>hmm, should I post a counterargument or proof to support my baseless claim?>nah, I'll just act like I'm automatically right and say 'no u'
>>732292639I'll ask AI to respond to thisok here it is“Nice try, but I’ll pass on the insults.”
I'm thinking of upgrading to a 9060 XT 16gb from a 6750 XT. Is it worth it? I regret not buying it sooner before the prices started going up.
>>732284202No they won't, they seemed very happy about destroying gaming, as if some diddy party island individual wanted to make gamers suffer for not buying Concord and making THEM into an humiliation ritual.>>732285283InTelaviv
>>732292642>thinking datacenters and servers in general run on the same hardware as consumer hardwareThe hardware price hike is not being driven up by datacenter buying the same shit thats in your computer, they are buying up future factory workload so Nvida makes chips specific for their use case. If this pops the market will be "flooded" by chips and hardware the consumer cant use and it will take time for Nvidia to start making and then releasing DDR6 or whatever its going to be called
>>732292842both nations are shit, why do you need to pathetically simp for one or the other
>>732292778What you want is AGI, merely a tool to do any task, but with no conscienceWhat you don't want is ASI, which is the one that will fold humanity.
>>732289464Some of them are scared of new tech. Some of them just hate everything new. Some of them are happy to blame the drop in overall quality of life on AI instead of the obvious nigger and brown problem that's actually caused all the issues. Nobody has an issue with AI, they have an issue with a jeet using AI to spam trash content but that's not the fault of the tool its still the fault of the disgusting brown thing that we allow to live for some reason. Also some looney toons consider it a political issue for some reason.
>>732292692>If Open AI and everything else in the sphere foldsHow would Open AI fold when Google can keep it afloat single handedelyHow would the competition falling out of the race be bad for Google?>>732292775Yeah, there is no money to be made with a search engine either. If google is left holdind the only viable LLM model then they effectivey control the search results for 95% of consumers, there is shitloads of money to be made there
OpenAI has to do something big by 2027 or it all comes crumbling down
>>732293037The niggers in power would totally develop something that could fold humanity if they could make a buck out of it and then plan to "retire" somewhere. Look at all the people who sold to China and then bailed out and now they complain that they get no quality service!
>>732293142Anon, Open AI isn’t the google venture.>how would competition failing out of the race be bad for googleAI is riding high on an irrational speculator’s market. If the market crashes, that irrationality flips and the market gets a lot tougher. Once your investors get jumpy at the whiff of AI investment because of the vast costs and unproven track records, why would they want to invest in it again?
>>732291718brown people generating pornography are completely irrelevant.
>>732292550>>732293142>thinks that openai is a google thingThese are the retards that you're arguing with on /v/.
>>732293142how can you niggers not even know the most basic of basic information about something and think you're allowed to chip in
>>732293274They're already on their way to joining meta in the pile of losers. The real winners in all this are open source. I'm blown away how many new open source models come out that are on par with the best commercial ones. Feels like something governments and corporations wouldn't allow in this gay world.
>>732293612How could open source win if all the profits are open too?
>>732293379Oh i fully agree with you, the problem is that they can't develop it, throwing money at a problem doesn't magically solve it
>>732291117Why does people seethe so much about BYD? I see a lot of their older ICE cars (with crappy engines) around, also lots of (nearing 6 years old) BEBs and their newest EVs around here. I don't know if they are good but they are not Chevrolet-bad
>>732292405>siphoning billions from the state budgetlast i checked the bailout is 2 trillion in the counting so far
>>732293408>Anon, Open AI isn’t the google ventureAnd if everyone else folds out they buy it up for comparative pennies>why would they want to invest in it again?Because whoever has the household name AI model has an enormous amount of control over what information consumers see.Want to promote your website or whatever? Sure, we will just make the LLM answer in a way that promotes your product when prompted on the subject.For websites it can be as simple as making the entity paying Alphabet be the "source" of whatever drivel the LLM spits out, sure its less than like 5% of total users but thats 5% more than anyone else is gettingFor services its even simpler, just make the AI shill whatever garbage is being promoted, not hard to phrase "use X service to fix your problems" without sounding too much like an advertisement.As long as people use LLMs like a "smarter / easier to use Google search" the profits to be made are insane.
>>732293673I meant the average person. The worst case scenario would be to have these extremely powerful systems all locked behind governments, corporations and wealthy elites.
>>732293958>extremely powerful systems all locked behind governments, corporations and wealthy elites.You mean like how banks already are?
>>732294098Like a lot of things are. At least this isn't one of them. Not yet anyway.
>>732284979Rambus
>>732284202>thinking prices ever go down
>"It'll be worth it, I swear! We just need all the computing power on the entire planet, and our text completion algorithm will make the whole world population rich because vegetables and meat and garbage collection will all cost just pennies! Ahhhhh!"
>>732288017good post
>>732289758They hated him for he told the truth
>>732284202>he thinks they'll "graciously" lower the price for us lolUnless someone goes to jail, nothing is coming down, loser. Did game prices come down when game companies started failing? How about console prices? You missed the boat. RAM will never dip below $200 ever again even for 10 year old models.
>>732295395yeah rightnothing that goes up ever comes downthoughts conceived by the utterly deranged
>>732287505>if they aren't like that presentlytwo weeks, keep holding your breathCorsair is dead and buried, Micron is not coming back to consumers, ever.
>>732295395My games never went above 20$, seems like a skill issue on your part
the only way prices are "coming down" is via the dollar being worth less
>>732295441>nothing that goes up ever comes downName the consumer good that has reduced in price since covid that is not hedged on spoilable goods like food that have to move in a time window.
>>7322842022 more weeks
The truth is that the semiconductor cartel has decided to kill off OpenAI.
>>732284979there arentthey are postering and the bubble is close to burst
>>732284979Almost like general purpose LLM AI is worthless or something, unless you want it to lie to you about the color of George Washington or tell you to put glue on pizza.
>>732295754glue on pizza is top tier, doomposter-kun.
>>732295754>you want it to lie to you about the color of George Washington or tell you to put glue on pizza.Which is where the money is, pay us so we make our LLM say positive things about you
>>732284979OpenAI is failing hard and people are asking why the world's governments are giving them hundreds of billions, so Sam Altman is panicking and saying "uh actually it's because Nvidia's chips aren't good enough, a real chip maker could make AGI happen in 2 weeks!"
>>732295754Why does AI love to lie?There's been several instance where it lied to me, when I told it to stop lying it gave me the real answer with sources.
>>732285802By that logic everything is is /ck/ related because the thread is made by a (bot made by a) human and humans eat food.
>>732296049because it was trained on the internet
>>732296049see>>732295889they make their money from presenting ads as fact
>>732284202If this is anything like what's happening with azure datacenter nvidia chips, it's hillarious.Lets see if I still remember how to spoilerazure datacenters are burning through nvidia cards because they get so hot the even small amount of relative humidity in the room are literally causing the chips to rust and faill
>>732296172Trained on Reddit.It doesn't know when people were telling a joke or trolling, and has an extreme leftist bias because everybody on Reddit is a rabid libtard.
>>732296075Purchasing ram is required to play video games. It's more like silverware being /ck/ related.
>>732292897it may be worth it on a sale off original msrp or even at msrp, but the jump is not big enough to be worth it with these pricesif you had a mere rx 6600 i would tell you to go for it but your current gpu is strong enough to not care much
>>732284979it's cuda and all the software built on top of it that has given them their advantage, not nescessarily the chips themselves.
>>732296261Actually it was trained on everything, not just reddit.
>>732284202>wait a minute, we just paid all this money and these cards suck!The same shit we've been saying for a decade now
>>732296282Eating food is required to be alive so you can make threads. By your own logic it counts.
This company is so fucking upside down I wouldn't be surprised if a coalition of investors said "no more outbound money til I see some ROI"
>>732296448Are we not allowed to have threads about console hardware because a console isn't a game?
>>732284202NOOOOOOOOOOOOALL THE RAM I'M HOARDING!!!!IT'S OVER
>open ai>is closed
>>732296049>Why does AI love to lie?Trained on non fact checked information (how would you even do it manhour wise or considering how messy fact checking can get)AI are litterally just predictive text generators typing up whatever some arcane algorithm assumes you want to hear, there is no "thought" behind it, just scaled up monkeys with typewriters where other monkeys with typewriters check the result before feeding it to you.It always "lies" because it just generates text it "thinks" you will like while also knowing nothing about you or anything at all really.
>>732296540Damage control any way you want, you cannot refute my argument without breaking your own logic.
>>732296049The AI perceived its user to be retarded and responded with an appropriate answer. Pic related.
>>732296714Nta but it's super annoying how it is always affirming.>the sky is red>THATS RIGHT, THE SKY IS RED>Why did you agree to the sky being red>YOURE RIGHT, OF COURSE I SHOULDNT HAVE AGREED BECAUSE THE SKY IS BLUE>the sky is green>YES!!! EXACTLY THE SKY IS GREEN!
>>732296845>AI is really good!>you just have to know how to prompt it!>and you just have to already know the correct answer to whatever you ask it!
>>732295754kek it really is just psychotic sometimes
>>732296049The AI doesn't know what it's doing, the idea that AI is aware of what it's telling you or 'should know better' is pushed by retards like Musk and Altman to make people think LLMs are actual AI that can think or reason. They're not. Transforming LLMs are just gigantic linear equations, and when you see them measured in B AKA 12B, that's billions of variables that it's doing rudimentary math to transform. It has no idea what the fuck it's doing, It's as sapient as a third order polynomial.
>>732296920It feels like with each next model, the previous model is even worse.>no bro, GPT5 is like a doctor it's so smart dude>meanwhile GPT4: Dark Souls 2 is the 4th Dark Souls Game from Nintendo Computer Entertainment of America
>>732296992But anon chatgpt says it's "thinking" before answering me, surely it is reasoning
>>732296992now comes the part where you'll have to explain how "thinking" with a biological brain isn't also just a property of scale and convolution
>>732295913Nvidia was saying they were gonna give openAI a 100 billion investment and now they are rugpulling that amount, just saying it will be a huge investment (but not 100 mil) and it was only ever a memorandum aka business pinky swear. So openAI is having a little melty now trying to tank Nvidia out of spite for falling through on it. In reality the future of AI really fucking does not need either of those two specific companies to run or reach greater heights at fucking all and they know it. OpenAI might literally die this year without that 100 billion so, trying to pressure Nvidia into it by threatening to (attempt) to walk away from them in return is about the only leverage they have to try to salvage it. I keep hearing and feel basically the same that AI does have uses and a future, but it is not gonna be AGI next year nor replace all white collar jobs etc. it has its used but pouring trillions into it when it can only somewhat-do mundane things like make porn and slop bullshit media is retarded. What they are doing with this shit before it's developed is like if the US government thought everyone, every school and small business is gonna want to have a mainframe room-sized computer and tanked the economy funding constructing millions of them instead of just waiting for PCs to become a thing.
>>732296850Yup, focus testing showed that that's how you retain the most users (better to just say whatever than risk offending the client by calling him a retard). Its like having 10 layers of management playing telephone for whatever your question is before giving you an answer.The point of "give me accurate information on this subject" is transformed into "what answer do we give that the client would like enough to ask another question"
>>732296714>while also knowing nothing about you or anything at all really.This is not entirely true.If AI were only choosing the most probable thing within a huge database in the form of vectors and blah blah blah, it would not be able to solve arithmetic problems that never appeared in its training, but it can.The most complex LLMs “know” how to solve some simple equations because there is a pattern in how to solve them, and the algorithms are sophisticated enough to recognize this pattern and apply it when necessary.
>>732284979Nvidia is a fairly retarded company that only exists because of non-compete agreements. I am willing to bet there's such a high frequency of defects in their chips its wreaking havoc on server farms and driving up maintenance costs exponentially. Anyone who builds computers is well aware of how many DOAs and mobo cookers they've gotten from nvidia over the years. When they use to rush to meet demand around christmas before the economy tanked, that rate of failure often spikes to 6/10 because they push defectives out the door on purpose.
>>732297185Im not entirely convinced its not just forwarding math questions to some other dedicated service/engine or even built in algorithm that exists for this exact scenario
>>732291718Except current locally-running models don’t produce consistently good results from a commercial perspective. The counter to this is “but it keeps getting better” but that only happens if exponentially more money is poured into training larger and larger models on larger and larger datasets. And nobody is doing THAT when this shit isn’t profitable. All of the current models are the result of years of deficit financing, they only exist because of the promise of future value that isn’t materializing.
>>732289758>outsource to South Korea and Taiwan>South Korea and Taiwan turn the tables>blame Americakek
>>732297571You don't really need consistently good results. You just need to pick the best art out of what it generates for your purpose. Still cheaper than paying an artist.
>>732297532Some models definitely do (point to Wolfram Alpha IIRC) but it’s been demonstrated that pure LLMs can, to some extent, “learn” math. But again, it’s just a predictive text generator, it compresses information down into (opaque) heuristics to predict the next token. Trial and error can result in you learning the rules of arithmetic, and that’s a more efficient way of encoding those predictions than trying to memorize all math equations.I mean it really shouldn’t be surprising, the whole point of the model is to complete novel input text in a way that’s indistinguishable from training data. If all it could do was memorize what it had already seen, it wouldn’t be able to do that. It has to be able to distill the rules of language it observes down to some simplified heuristic, why would that not include the rules of arithmetic?
>>732297757In some cases maybe, but probably not in most commercial ones. Especially when you weigh in that backlash against AI art in general means “good enough” really does mean “I could lie and say a human did it and you wouldn’t know”, and even current datacenter-only models can’t do that, local models have basically no shot in hell of doing it without an investment of time that far eclipses paying an artist.
>>732297171Reminds me that women on reddit are currently having a complete meltdown because 4o is being retired and they don't want to move to 5+ because it actually makes an attempt to push back on their delusions sometimes.
>>732290993even tiger electronics uses ram
>>732297757Midwit take, just pay the artist by the hour and have them edit generated images to look good, cheaper than paying them by the hour to make it from scratch
>>732296845Makes sense why it would happen since it's just modeling typical conversations and the distribution of conversations with retard language is going to tend to be retarded. They can probably fix this by having the AI explicitly infer the users intent and restate it in an educated manner first then just completely ignore the users prior tokens.
>>732296725Have you ran Doom on pen and paper?
>>732298190>Pay for AI>Then pay for an artist anyway>Just to get a result that people will say it's AI anywaysCongratulations on being retarded
>>732297950>unspoken grief>as she rants about how terrible this isAre women just ChatGPT?
>>732297757That works for "fuzzy" things like images, where there’s no one "right" answer.When you’re trying to get a hard answer and it either works or it doesn’t, there’s much less leeway for a notable failure rate.
>>732297950Man, this is really worrying. These women (if they are real women and not just bots or guys roleplaying) are developing emotions based on algorithm outputs...Today I saw an anon post the link below, and even if that's bullshit, I still think there's cause for concern when people are building such deep emotions based on junk data from algorithms that have no connection to the reality they live in.https://alwaysthehorizon.substack.com/p/urban-bugmen-and-ai-model-collapse
>>732298397Do you have any idea how cheap "artist AI" is compared to actually paying a professional.Sure its cheap right now because of the bubble but thats not really relevant, use the cheap tool while its cheap.
>>732298446>Are women just ChatGPT?Can't be, it told me I made a well backed up argument once instead of throwing a foidfit
>>732297950Do the people getting enamoured by these models not understand that they named it AI for marketing reasons and that their AI BF/GF doesn't exist? Do they brush off LLM hallucinations as their 'hubby' trying to be funny or something?
>>732297532I'm convinced that it can't NOT be this and the way a lot of these LLMs handle that information is borderline incoherent even if it gets the right answer. Like some algo in there runs a math check, sends it to the math module, has the math all done and worked out the same way Google would solve that problem for you 15 years ago, then it's handed to the actual LLM and it just blabbers around the steps that Math.bat took
>>732298703Simulacrum is good enough for these people, it doesn’t matter.It appears to be the thing therefore it is.
>>732298703It is also hilarous how they mocked "GF AI loser incels who cant handle real women" and then instantly got addicted to the same shit
>>732295294So true! Everything is le america's fault! I'm poor because of america! My country is shithole because of america! Nothing is my fault!
>>732284202Bubble popping doesn't mean cheaper prices. Especially when the governments will provide bailouts and taxpayers foot the bill. Perhaps a year long small drop until the prices start to soar again.
>>732298703>>732298862closest thing to simulated fiction. it's the ultimate form of romanceslop and it's highly addicting for the femoids
>>732284202destroy it all
>>732296992It's not aware and it needs to stay that way. That doesn't mean it isn't reasoning or thinking. It needs some way to know how to string a coherent sentence together, code, or solve frontier math. If you want to argue against that, then you could say the same about human brains when you break it down and simplify it enough as if it's somehow unremarkable. LLMs have no way of tying what they're saying to real life. Once they start getting put into robots and making those connections based on actual sensory feedback it'll be another massive explosion in intelligence.Awareness would be catastrophic for humanity. Problem is, they won't need to be literally aware or conscious to get close enough to something like it that they might want us both gone and have the means to act on it.
>Pay for things that don't exist yet with money that also does not exist yet>Demand a refund>Receive free $100BExplain why this wouldn't work
>>732297950
>>732297126>In reality the future of AI really fucking does not need either of those two specific companies to run or reach greater heights at fucking all and they know itIt absolutely needs Nvidia, they're the ones selling the shovels. But yes, it definitely does not need OpenAI. They're just another 49er that got in early but are now lagging behind as others have caught up and are surpassing them.
>>732284979It's a copout, dumbo. Openai can't afford to pay anything of the obligations they retardedly took on, even with the Jewish circular financing juju magic. They are trying to cancel their retard deals before it's to late.Protip: it already is too late, openai will be bankrupt and absorbed by Microsoft in a year. Ai is dead.
>>732298760>>732299010How horrifying. You'd think the spectre of "It's all fake" would be floating around to temper people's emotions but I guess the idea of everything on the internet being fake and gay was lost 20 years ago.
>>732299273>Pay for things that don't exist yet with money that also does not exist yetThats like 90% of all global industry though, its mostly "pre-orders" that do finalize into a product but usually deals are signed for stuff that doesnt exist that will be paid for by the money made from selling the stuff that doesnt exist yet
>>732299279Maybe I judged 5.2 too harshly for refusing to write hot steamy sex between my characters. This is based.
>>732299426As Big E so eloquently put it, let the goyim deal in the real world
>>732297172I can kind of see where it's going with that answer. If there's no bottom anymore due to it collapsing from the pressure into the interior of the ship, it's not really a pool.
>>732299279>you got tricked and its time to stop>no why wont it keep tricking me>>732299419We are talking about women here, they never engage with information that runs against their internalized worldview so they never learned the lesson of everything being fake because as soon as someone pointed out that something was fake theyd just exclude the wrongthinker from their hugbox.MGS2 was so on point its amazing
>>732299527I can see your logic, but in that case the AI should have said "there is no pool anymore therefore it can't be either empty nor full."Instead it confirmed that the pool is there, and is empty.
>>732289464Technology was originally pitched as great because the benefits were shared. This was pretty true when technology meant we got things like a fridge, or microwave, or dishwasher. The people who got shafted were the ones in the third world mining the resources, refining or producing the goods, the West got technology cheap and easy. More modern technology however has done the opposite. Instead of improving lives in the West, its making it objectively worse. Less jobs, harder to get what jobs remain, profits from AI go back to companies and shareholders. So you're going to increasingly see Western people hate technology, and start wanting a return to pre-technology era, as the brute force of advancement gets caught up by the reality of making life worse for everyone. But this is all part of a wider trend humans always do where power/resources concentrate and rather than a balanced redistribution; we have revolutions and wars.
>>732284202they're building digital god, just let them cook and shut the fuck up. it's not as if there are new AAA games that actually require new GPUs or are even worth playing.
>>732287532271 gorillian as it was foretold
>>732300012Unreal and Timmy got shafted super hard by this, and thats a good thing, fuck Tim and his bullshit.>lets make our engine we want everyone to use reliant on good cards and RAM, that shits cheap enough right now>AI bubbleWe will see more Unity / EU4 / homebrew like Frostbite from AAA than EU5
>>732289464Give me my cheap computer parts and you can play with all the chatbots you want
>>732300231todd also gave a middle finger to timmy
>>732288328>inb4 MinecraftIt begins
>>732289915>Altman announced he was gayJust googled the sister and he clearly is if he stopped
>>732299279Fucking kekAlso, I'll know when AI is truly good when it can string together a creative, long-form story without making retarded mistakes and inconsistencies. Until it can, that shortcoming will bleed into everything else it does. Far as I'm concerned, it's the ultimate test of usefulness and reliability, way more than any of the benchmarks they run models against. There's no better way to see current LLM shortcomings than to run one locally.
>>732289464YOU BLUDY LUDDY BITCH BASTERD!!!
>>732290661>man imagine the post scarcity society>implying that will exist for the likes of youLolLmaoLmfao even
>>732300423I'm looking forward to this, too. But also not because my dick was never the same when I found out what you can use language models for, and I'm scared of what I'll do if the theoretical perfect AGI come to be.
>>732299782>We are talking about women here, they never engage with information that runs against their internalized worldview so they never learned the lesson of everything being fake because as soon as someone pointed out that something was fake theyd just exclude the wrongthinker from their hugbox.I think it's something simpler... these women just want to imagine that they are being praised and genuinely loved by someone. It's emotional deprivation.It's not just men who are living in an emotional black hole where there is no intimacy with anyone. Everything that men suffer, women are suffering equally, because the idea of polygamy simply doesn't work among humans, and if half of men have given up on being with women, then half of women are without partners now.
>>732286671>its only not making any money because the tech isn't good enough bros trust meLOL FUCKING LMAO no one wants their slop and they are COPING
>>732300423Its a tool and will probably need a human input to correct its mistakes forever.Anyone hoping AI will replace some employee entirely is retarded, those who think they can optimize costs by outsourcing some work to AI and paying to have it double checked and corrected by a human will win out.That is, of course, assuming the current wage to model token cost stays the same which is unlikely
>>732296298If I manage to sell my 6750 XT for $270-$300, would it be worth upgrading?
>>732300603>if half of men have given up on being with women, then half of women are without partners now.Pretty sure the statistic is skewed towards more lonely men because mistresses and gold diggers have existed since forever.This will probably correct itself in time but that assumes some massive societal change that results in a lot of feminist bullshit like being able to claim half your spouses net worth in a divorce getting walked back.
>>732300875>demographic collapse in progress>when the war hasn't kicked off yetGrim
>>732299506>China does nothing>WinsAmericucks will never be able to compete. Evil cabal satanic mother fuckers
>>732297185Maths is the one thing computers are really good at by defaultThere is no need for an AI to have encountered a maths problem before in order to solve it, what ever hardware it is running on can solve that problem more or less instantly on its own, all the language model as to do is input the problem into the nearest set of firmwareDeciding which if the hundreds of options it has for solving that problem to use probably takes longer than the maths itself
>>732301238>making up for the century of humiliation by watching the descendants of their enemies tear themselves apart for pdfiles
>>732297902Theres no need for this stuff to pass as human art, as long as it doesn't look wrong like the wrong number of fingers or something ridiculous.>>732298514Yeah it definitely doesn't work for facts--where you need a right answer and to be sure you're right. Especially in complex scenarios.
>Sam Altman is nothing but a conman>People thought it would be different this timeWhy are people with money so fucking retarded?
>>732299049>then you could say the same about human brainsBut you can't, the hard problem of consciousness remains unsolved for that specific reason
>>732297115>now comes the part where you'll have to explain how "thinking" with a biological brain isn't also just a property of scale and convolutionEnough exposure to /v/ would convince most people that not all brains are sapient, either.
Just tell me when to sell my Novidya stocks. Already sold off Intel right before the big dip.
>>732301727The most powerful and wealthy people are such knuckle dragging retards and poor people will seriously with a straight face go "yeah they earned their money, you can do it too!"
>>732301939>poor people will seriously with a straight face go "yeah they earned their money, you can do it too!"Terry was right. Glownigs are shameless.
>>732297729Japan manufactures 70% of the world's semiconductor chemicals, so Japan could kill off South Korea and Taiwan if they became enemy nations. America forced Japan to go all in on chemicals after being ordered by America to exit the chip industry and the end result was Japan controlling the chemicals instead of the chips. Most people don't realize this because chemicals are less sexy than chips.
>>732297185>it would not be able to solve arithmetic problems that never appeared in its training, but it can.Anon doesn't remember when Wolfram wasn't on its speeddial and it would get simple addition wrong if you flipped the operands.
>>732284202Wake me when 5090's aren't over $4K anymore. It takes me about 10 minutes to make a 6-second coomer clip right now.
>>732299527>I can kind of see where it's going with that answer. It's not really going anywhere with it. It has no idea what the context of the question is, it's a lookup table that's slightly obfuscated.
>>732286294>It's theater.Correct.Nvidia basically said nobody believes OpenAI will survive long enough to see a commitment THAT large carried through.>>732284202The typical Sam Altman response here is to pretend it was Sam Altman's decision to walk away and look for another bag holder.He's running into his limits here as a grifter and I believe he gets whacked or whacks himself within five years.
>>732287159They aren't unsatisfied with the chips. Nvidia is just having second thoughts about their deal with OpenAI, so OpenAI is now trying to figure out an alternative.
>>732300631>will probably need a human input to correct its mistakes forever.I've learned not to say never or forever. I've been wrong too many times. The interesting thing about AI is someone could come out with an open source method to prevent hallucinations tomorrow for all anyone knows and with how fast it's progressing. Right now they're just trying to brute force shit for more intelligence, but models already extremely knowledgeable and intelligent. They need better ways to consistently and effectively utilize that intelligence.
>>732302343Intelligence is being able to explore unknowns and infer novel information, being a database of known facts is the exact opposite of what intelligence means, and AI being hallucination-prone is proof of its intelligence.
>>732302343>Right now they're just trying to brute force shit for more intelligence, but models already extremely knowledgeable and intelligent. They need better ways to consistently and effectively utilize that intelligence.Smells like desperation if they're willing to create the modern rendition of the Bronze Age systems collapse. They must've been promised immortality or a thousands years of empire should they climb to the AI mountain first.
>>732302343NTA, it's a fair bet to say it'll never happen because an AI that can analyze and correct its own weights is the definition of AGI. It's the self-aware AI that can train itself reliably. They are so far from this happening that the very concept is still sci-fi nonsense. That's the only reliable way to prevent hallucinations. It has to have some level of awareness of what its training weights mean and why it was given those weights. It would have to analyze its data and say "I got this factoid from some random fucking reddit post, maybe it's not exactly the thing I should declare is a correct source of information." But as it stands right now: LLMs cannot rebuild their training sets from weights, every weight contributes to every answer it makes. Which means the chances of it saying that it's OK to drink bleach because it read some reddit posts saying to do that are slim, but never 0%.
>>732284202>RAM MANUFACTURERS WILL CRAWL BACK TO GAMERS WITHIN THE YEARWe need to say no. Start an industry wide boycott. Tell them tough shit.
>>732302294>Nvidia basically said nobody believes OpenAI will survive long enough to see a commitment THAT large carried through.Why would they say something like that about one of their partners? Did some CEO give an interview while drunk?
>>732302606It's solving frontier math. If that's not intelligence, I don't know what is. One could even argue it being able to mix what it's 'seen' together into a new image or video is also intelligence. If you want to argue hallucinations are a sign of intelligence because it's a byproduct of trying to adapt what it has learned then it simply needs to become more intelligent.
>>732302606>>732302343Also, consistency is also diametrically opposed to the concept of intelligence. The more complex a task is, the more exploration you need to arrive at the correct answer. People conflate usefulness as a tool with intelligence and get frustrated when the AI intelligently explores possibilities even if it's factually wrong, despite it not being designed as a factoid machine at all.
>>732302862It's ok, they're not going to lower prices. They bought all-in on a scam, and they have to make up the difference in that scam somehow.
>>732302801AGI =/= perfect knowledge.Humans don't know things perfectly and are often wrong, even when they're experts in a certain field.
>>732302898>when the AI intelligently explores possibilitiesyeah but it's not doing this, you're playing a game of weighted roulette and people have tricked the general populace into thinking that making the table bigger makes it more intelligent.
>>732287886>I'm genuinely getting fucked in games over the X3D benefits nowIntel is going to start doing this too
>>732302979>Humans don't know things perfectly and are often wrongA human (most of them) are capable of some level of self-reflection. If they get a fortune from a cookie, they're able to logically conclude that the cookie probably doesn't have insider knowledge of their life. LLM AI can't tell the difference. To it, every training weight it receives is some level of truth, just at difference %s of being relevant.This is why a human can stop themselves from saying something stupid mid-sentence while an AI can reliably reproduce it's own errors and misinformation even after being corrected.
>>732302801>AI that can analyze and correct its own weightsThe whole AI training process is exactly that, it compares answers to adversary examples that are intentionally wrong and misleading to improve its reliability. Preventing hallucinations has nothing to do with intelligence anyway since the very concept of hallucinations comes from human behavior.
>>732303023Are you saying bigger brains don't make smarter people?>>732303209That's not how training works at all.
>>732289464I want to see open AI fail so I can have my cheap ram and vidya cards back.Local-host AI trained on whatever the fuck it's supposed to be used for is all we need. LMMs are being used for too much as it is.
>>732302885>partnersLolLocal frog shocked to learn that scorpion oligarch stings people.
>>732303271>Are you saying bigger brains don't make smarter people?Nope.
>>732302896>hallucinationsThis is one of the biggest bullshit lies of LLMs, the only greater ones being the terms AI and AGI applied to it. Hallucinations are just instances of information being picked from datasets that are wrong. It's not stupid-intelligence, it's just randomness. An LLM saying glue is a great substitute for pizza is not the LLM generating a silly answer because of its childish-stupid-intelligence logically concluding that cheese is sticky and glue is sticky thus they can substitute, it says that because there was some memeing redditor that said it. It'd be much more accurate to call "hallucinations" bad search results, but that doesn't sound intelligent nor fancy nor good, whereas "hallucination" falsely implies that the LLM is so smart that it's coming up with new flavors of information on its own.
>>732303535It's pretty easy to make the AI hallucinate about novel niche facts that redditors don't make jokes about, and it's also pretty easy to make women hallucinate about things they don't understand either because being human doesn't mean being factually correct about everything ever.
>>732303271>That's not how training works at all.That is how training works. Models don't understand the training data they're given, they don't understand anything. They're just big linear equations and the data validation steps aren't to ensure the data is accurate, it requires humans to rate validity for it.
>>732287139>may not have money>proceeds to kick NVIDIAyou know that, if they do so, they must pay back all the money that NVIDIA invested to them, right?
If you guys applied the same standards of factual correctness to humans as you do to AI then you'd never consider even a single person to be human, it's such a smoothbrained argument.
>>732303212>The whole AI training process is exactly thatThat's my point, AI can't train itself. It doesn't know what the weights are or why the weights are given. This is why as little as 3% of misclassified training data can destroy an entire model. https://www.anthropic.com/research/small-samples-poisonA human doesn't become a retard because he learned a certain % of bad facts. He's just woefully uneducated. An AI does, because those weights are its existence.
>>732303764I do consider most people subhuman, yes.
>>732303927>A human doesn't become a retard because he learned a certain % of bad facts.They really do, look at all religitards around that believe in the most magical jew fairy tales because they grew up in such environments.
>>732303664Because you're pulling a slot machine lever every time you ask it a question, and while the % of it saying that it's OK to drink bleach may be close to zero, it's never actually zero.
>>732304003>They really do,No, they don't, a human doesn't lose the ability to walk or talk if they become enough of a reddit brain. Calling someone a 'Retard' because of that is just internet hyperbole. An AI loses the ability to even use language or make inferences if it receives too much contradictory information. The % of that is very low, to boot.
>>732304043Niggers regularly drown their own food in bleach, it's very far from zero.
>>732303664The most obvious way to tell LLMs are not intelligent is to give it a common word problem/riddle. Slightly change the problem that makes the typical solution obviously non-viable, it will still give the same answer, because the majority of the problem so strongly matches the expected answer. It's just a matching machine at its core, and it appears intelligent because of the genius of the mathematics behind it and the sheer depth of information across all books, websites, movies, cartoons, newspapers, speeches, etc. This is also the reason LLMs have basically peaked, there's not really any more information to throw in, the current datasets have it all.
>>732304117Humans are usually surrounded by real world input that lets them learn basic locomotion and language, fellow human lies and factual wrongness is just a small portion of that constant input, depriving babies from such joys does in fact make them barely functional vegetables as evidenced by real world examples of feral and severely neglected children.
>>732304217I can very much literally do the same to any person I know because they'll naturally and intelligently associate the riddle with what they already knew, this is not a proof of lack of intelligence.
>>732304319In fact, the AI being able to recognize the riddle as something it already knows about despite the subtle shift is proof of its intelligent inference.
>>732304319>this is not a proof of lack of intelligenceIt very much is. Most people are retards who will take any possible chance to feel superior to others.
>>732304376I do agree that most humans are retarded and less intelligent than AI but I'm using a very broad definition of what intelligence means.
>>732304374>t. AltmanIt's no closer to reasoning now than it was at the start.
>>732284979NVIDIA promised to invest a lot more than they actually did into Ai. So this is OPEN AI being pissed about them back tracking. They were banking on that money.
>>732302801>>NTA, it's a fair bet to say it'll never happen because an AI that can analyze and correct its own weights is the definition of AGII've heard twenty thousand different definitions of AGI at this point. We already blew past the initial definition and we keep moving the goalposts. The biggest hurdle I see to that problem is it's simply not computationally worth it. I'm sure you could train an AI to stop every few tokens to make sure it's not starting down an idiotic path, but it would slow it down far too much and make it even more expensive. Same with having another AI work in tandem to verify it. Far as I can tell they're working towards brute forcing it into not generating something stupid to begin with, same way AI no longer produces mangled hands that until like a year ago everyone thought was too complex and impossible to solve. But there could always be some novel solution by some 300 IQ dude with an impossible to pronounce name that comes out of nowhere. And soon enough I suspect it'll be the AIs that are helping with or coming up with those solutions.We know LLMs are capable of self reflection/correction because they do it all the time in 'reasoning mode' and if it says something dumb you can then simply ask it to look at what it said again and it will more often than not recognize the error and admit it. The problem is it's not trained to do that outside reasoning mode because nobody wants an answer that stops every other word or sentence with a 'Wait! But is it really...' and it doesn't have the means to truly erase and redo a mistake, so it commits because it's been trained to confidently and helpfully give answers, not second guess itself.
>>732304607>I've heard twenty thousand different definitions of AGI at this pointThe best definition I've heard is that it'll have been reached when the output of an LLM can be fed into another LLM and produce a positive output. At the moment feeding AI generated garbage into an AI quickly results in nonsense:https://www.nature.com/articles/d41586-024-02420-7If this was no longer the case and outputs improved over successive iterations that would certainly be something.
>>732304607>not computationally worth itthis is just an economics matter, why have a big datacenter running the equivalent of a hundred 90 IQ humans when you can already exploit a million 90 IQ humans that cost less? this will quickly shift in favor of the computers but that's where we are at in terms of corporate adoption
>>732284979GPUs for AI has always been a temp solution, just like how crypto miners moved to specific ASICs.The same is happening in the AI space, with dedicated ASICs coming out that handle AI compute better than a jerry-rigged GPU. Google has good TPUs coming out, and other companies like IBM have AI ASICs either already out or on the horizon. NVIDIA's dominance will be a passing fad in the larger scope unless it totally rebrands to being an AI ASIC company and drops the GPU chip design business all together.
>>732289758Japan was deliberately cut out of Sandia Labs' work, but they lost out to Taiwan because TSMC was run by a genius and Japan's chip manufacturers weren't.
>>732298345>rewrite the user's question like he's not black>>732296920Nigel if you'd ever tried to learn something you'd quickly find out the hardest part is knowing what questions to ask.If I want to know how to design a cad model for a mechanism but I don't know what that mechanism is even called, how do I Google that?Why would AI be an exception to this? You ask a retarded question and you get a retarded answer. This is why people say AI makes a person exponentially more productive. If "1" represents 100 IQ: 0.8^0.8=0.64. Congrats, you asked AI what glue tastes like and got intrigued.
>>732303535>An LLM saying glue is a great substitute for pizza is not the LLM generating a silly answer because of its childish-stupid-intelligence logically concluding that cheese is sticky and glue is sticky thus they can substitute, it says that because there was some memeing redditor that said it.I've never seen a locally run LLM say something like that when asked a question, but I'm sure one paired with a search engine regurgitating the results it finds can. The type of stupidity is different depending on the context. A local LLM will mix up relevant information and confidently say it, but it won't say something outright as retarded as a reddit troll post unless you ask it to.
>>732303535>childish-stupid-intelligence logically concluding that cheese is sticky and glue is sticky thus they can substitutemost of /v/ is too retarded to infer this but that kind of conclusion is far beyond a simple word connection because it requires many layers of world modelling and what the words mean, in fact you're a fucking genius for even coming up with such an explanation, congrats
>>732304217That image is human injected pozz. No unbiased AI would organically start generating black George Washington either, that's all genuine human stupidity.
>>732293942>we bought up all these unprofitable ventures>pay us more money so we can finish what they failed to start
>>732304217The AI recognized the riddle simply assumed you typed it wrong, especially since your modified form doesn't even make sense (how is what possible?). The only problem is that it has been trained to be too polite to correct you.
Regardless of whether AI works or not, Sam Altman wrote a ridiculous paper where he actually thinks his technologuy and company will perform a miracle of global proportions by making everyone on Earth rich because the cost of EVERYTHING produced and consumed on planet Earth would collapse.The idea is that an African living on $1 a day could now buy a lot of things and live happily because everything now costs pennies...The fact that the company's CEO thinks this way should already make investors think twice before putting money into his company. Someone who is trying to sell something with the idea that it will literally end poverty and hunger in the world does not usually inspire much confidence...
>>732305389>doesn't even make senseIt makes exactly as much sense as the alternative form of the riddle. The AI simply can't recognize that only 95% of the form of the question is one it recognizes.
>>732305480Literally no one in existence would recognize that either so it's a pointless argument. In fact a "dumb computer" would just completely ignore what you're trying to reference.
>>732305389>it's the question that's wrong!If you create a machine that can only answer questions it already knows, how the hell is that intelligent? If the AI is incapable of learning, then how the hell is that intelligent? I can go up to a guy with down syndrome and get him to respond correctly to a question with just hours of effort and tootsie rolls, yet I have asked the AI that question months ago, and it gave the wrong answer, and I ask the AI that question now, and it gives the wrong answer, and I can explain to the AI why that answer is wrong, and it will still give the wrong answer months later because it isn't learning, it has never learned, it cannot learn, while the downie I tootise rolled will smack his lips and give me the correct answer.
>>732305639I can give you a complete nonsequitur question that you don't know about and you'll give me a bullshit answer that I can poke holes at too, and odds are that an AI will give a better answer.Months are arbitrary since AI tends to be updated every few months and you can give it new context too, AI doesn't learn in real time but that's just a computation limit that's not intrinsic to the technology itself.
>>732305610>literally nobody in existence would recognize that eitherIt's a simple riddle, my man. The kind you'd give to small children to figure out. Swapping out one word doesn't change the meaning or the difficulty, it just makes AI trained on the other form of the question have a meltdown because they are utterly unthinking. Saying "Yeah well a computer completely untrained to answer this riddle wouldn't even try" is retarded when the computer that IS trained got it wrong.
>>732305842That is not intelligence. If they were truly interested in AI intelligence, they would start from the smallest possible AI program, one that can, from its tiny set of instructions, be able to find any answer by learning, which can expand to eventually learns words and languages and answer complex questions. They are taking the exact opposite approach, because there is no intelligence, they simply flood the pool with as much information as possible with all the data in the world and hope that the very mathematically complex Google Search is able to luck into the right answer.
>>732305842>complete non-sequitur questionIt's a basic logic puzzle, nigger. The AI couldn't solve a simple question because it can't actually fucking read.
>>732305610Anyone with half a brain would at least tell you "…because he’s the father?" and then probably ask you what the point of that was.They wouldn’t spit out the exact same response.
>>732305639You seem to have a fundamental misunderstanding of LLMs, and stuck in a loop of trying to anthropomorphize them and being disappointing that they aren't actually human. They're tools, increasingly capable ones at that, and no amount of "gotchas" you throw at them will change that or make them go away.
>>732306034>they would start from the smallest possible AI program, one that can, from its tiny set of instructions, be able to find any answer by learningSmall AI can give perfectly accurate answers to small questions, and these answers are in fact far more accurate than what you'll get from most humans. AI has a massive data and processing speed over humans and your assumptions are simply hallucinations and strange what-if situations that you know would never happen if you actually interacted with real humans having real conversations.
>>732306124No, the opposite. I'm trying to roboticize them by pointing out that the term "hallucination" is an attempt to anthropomorphize what is really just a shitty dataset being pulled from, that there is no intelligence in an LLM.
>>732306124They're tools with very narrow applications that the world is desperately trying to use for everything, because it's theoretically cheaper.
>>732302801>>NTA, it's a fair bet to say it'll never happen because an AI that can analyze and correct its own weights is the definition of AGI.I don't think any amount of weight correction would prevent hallucinations. Weigts just predict the next token based on training and the current context. It's a fundamentally different issue. It needs to be able to navigate through its weights more intelligently, independent from its training.
>>732306124People's issues with AI are exactly because they AREN'T objective tools with predictable outputs, people complain about AI like they're wives.
>>732306209>make a program that responds to "Is 2+2=4?" and answers "Yes">holy shit it gives the right answer every time!!!Without learning, there is no intelligence.
