The sooner you accept it, the better. It's over, ai evangelists. But at least you had a good run siphoning off money from the clueless masses.Too bad it's much less now since you drained the same bozos off their life holdings with crypto during the pandemic.
>>108231373It's called science. You don't understand it because all you care about is chicken and beer and football. Leave it to people who've actually played both Portal games, you country bumpkin racist christian hick.
>>108231373AGI might come someday, but ACI ("artificial capable intelligence", i.e. AI smart enough to replace 90% of entry-level office/tech workers) is definitely coming and it's coming soon. The funny thing is once companies replace junior positions with AI, they're eventually going to have a difficult time filling senior roles as boomers retire or switch jobs and they don't have any youngfags working in junior roles to get promoted.
>>108231373each time i see someone post yt's ui, it's just getting uglier and uglierhow do jeets do it?
>>108231373So many grifters and midwits. There's hype, but what's there remains very real.
>>108231459
>>108231373>AGI is NOT comingWheat makes you say this?
Fear
>>108231870No not wheat. Science make you say this. Wheat doesn't make you say anything unless you have celiac disease. Then it makes you scream in agony and spray shit everywhere.
>>108231877He will get into IBM.https://www.cio.com/article/4134276/ibm-looks-beyond-short-term-ai-gains-tripling-entry-level-hiring.html
>>108231373That's a nice looking screenshotDid you take it?
AGI (Ad-Generated income) will be coming soon to all AI agents
>>108232302>>108220681
>>108231373LLMs are not going to be AGI.But we're not just doing LLMs, except for Sam Altman but he's fucking retarded and will crash and burn to the ground harder than Megaman X7.
AGI is coming, i just need another trillion dollars
>>108231400/g/ is the only board that has gained a higher IQ over the years. People wouldnt think this is bait anywhere else
>>108232320LLMs can already do deductive reasoning, inductive reasoning, abjunctive reasoning, causal reasoning, analogical reasoning, critical thinking, emotional reasoning, the scientific method
>>108231373and humans are a mix of flesh and water
>>108232411Yep, but LLMs can be wrong, can be wrong often, and they don't know when they're wrong or when they don't know.You can make a better arch that deal with this, but pure LLMs can't.
>>108231400>It's called science.
>>108232471Have you considered that it might be falliable humans who are wrong, and not the machine? How many FLOPS can you do?
>>108232512Walk to the car wash to learn why LLM is a dead end.
>>108231373>AI is just <pop science food metaphor tier explanation>, it can't do <thing that a new model does a week later>yawn
>>108231373Wrong. AI is the future and the future is here. My OpenClaw agents have enhanced every aspect of my life. I am my own family now, taking on every responsibility from infant to toddler to k-12 to college to work, and beyond. I fill every role via my agents and I have never been more productive. AI is such an incredible force multiplier I am continually astonished at how few people use it to its fullest potential to be more than human: Superhuman+AI.
>>108232512Nope.It is wrong, factually wrong in many cases.
We don't need AGI to turn everything upside down and fuck everything, what we have right NOW is more than capable enough to excel at itIt is simply too novel, we haven't had enough time to integrate it efficientlyAGI would cause a far greater impactASI impact is simply unfathomable
>>108232592Is this the new cope after 5 years of "we'll have AGI by next year"
>>108232411If by reasoning, you mean token prediction that resembles reasoning, then sure.
Riddle me this:if LLMs never go beyond 90% or even 95% accuracy, why would anyone pay money for them? In areas where it's "good enough" there isn't any incentive to use it in the first place, and in areas where it's not it will never replace humans.*So who is dumb enough to pay for the costs*?>inb4 but it's cheap right now>right now
>>108232639You have to zoom out and see it from other peoples perspectives. LLMs are how people expected personal computers to work when they were first gaining traction, and in their eyes the whole part with the software and games and internet and apps and social media has just been a 40-year detour.
>>108232685>LLMs are how people expected personal computers to work when they were first gaining tractionI expected the machine to not randomly hallucinate and then sycophantically tell me how right I was for correcting if, IF I can correct it.But then again you're hallucinating right now, so I'm not exactly surprised.
>>108232639Almost nobody does pay money for them. OpenAI has stated that roughly 3% of their users pay anything at all. This is one of the biggest problems they face as a company.
>>108231373It'll continue to develop, but it's going to near platau unless there's a shakeup soon. Transformers or at least the role they inhabit must be revised, if not the whole damn structure.
>>108232793Then why are LLMs supposed to change everything?>inb4 they don'tTalk about the numbs/g/ulls not being able to shut their trap about em.
>>108232817The claim is that we are just 1 trillion dollars and 2 weeks away from having an LLM that will do literally any task a human could do and no company will ever have to pay a person a wage ever again. A lot of very gullible people with a lot of money believe this.
>>108231373>company now needs only 20 developers instead of 30.yeah ok we don’t even need AGI for this to get ugly
>>108232874These claims are retarded and most people know it. The stock market right now is pricing in just the productivity boost, not god-robot machines.
>>108232552> I am my own family now, taking on every responsibility from infant to toddler to k-12 to college to work, and beyondQuick question, what the fuck does this mean?
>>108233134Brainrot.
>>108232918IMO the stock market is pricing in inflation. If you price the S&P 500 in gold it's actually significantly down over the past 5 years.
>>108233563If you price it in gold it still hasn't recovered from the dotcom bubble
>>108233569Tracks with the general belief that life has gotten worse since the 90s.
>>108231373GPT-4 already trained on the entire internet, there was like a year time after GPT-4 when every model was worse then it when using same underlying tech. We are almost third year since the model collapse for LLMs was supposed to happen, yet they just keep getting better broadly speaking, unless you fell in love with it and want constant glazing, in which case LLMs peaked with 4o. So tell me, when will this model collapse happen?
>>108233569I'd believe it.
>>108231459Most of those jobs could've been replaced with Excel macros long ago.
>>108233656That would require people having to learn stuff
>>108231373every AGI benchmark they devise gets saturated, every time someone tries to point out something AI seemingly can't do, the chinese or one of the big labs get it there a few months lateryou have fucking AI MUSIC topping charts, and LLMs beating humans on novel coding benchmarks, that's todaythere is likely several major algorithmic breakthroughs left before LLMs alone hit any serious plateauthe question isn't whether AGI is coming, the question is if you're still retarded enough to be in a field they're already trying to dominate with the technology
>>108233919How have LLM algorithms improved? Isn't it still the same transformer architecture?
>>108233919>LLMs beating humans on novel coding benchmarksWhere?
>>108233919>LLMs beating humans on novel coding benchmarksThe top-of-the-top class can't even build a functioning C compiler with ten-thousands of dollars worth of capacity tossed at the problem.
>>108233919Anthropic went from 7 employees in 2021 to over 4000 employees in 2025. Their headcount is growing exponentially. Their main product is a webshitter electron app, they couldn't even get their LLM to produce anything better than webshit. They put their AI in charge of a vending machine and it went bankrupt. AI companies like Tesla were promising full self driving AI to make truck drivers obsolete by 2020. They still haven't replaced drivers, even drivers of locomotives are still safe. Think about that, they can't even get AI to drive something on rails with full right of way. Sam Altman said we'd have AGI in 2025. That obviously didn't happen either. This is going to be like the third or fourth AI winter, but this one will be such a spectacular and stunning collapse, I don't know if there will ever be enthusiasm for AI again in my lifetime. The amount that investors are going to lose this time will likely cause the US government to collapse into ruin the way Russia did after too much military spending in the 80s.And what is AI producing now? Slop porn, art, code. Three things that generate zero revenue when PEOPLE make it. There's no market for porn, it's all free. Art, no market. Code? Don't make me laugh, everything is open source. All these markets are saturated already, without AI.
"predict the next word" can do a lot of productive activity but as an advanced autocomplete it's going to lack autonomy.
>>108234573I'm a programmer and I still don't understand some of the projects I was hired for in the past. People always wanted their special snowflake implementation, many of the problems were basically already solved. Most websites and apps also make life worse.
>>108234579Just tell it to predict with autonomy
>>108234592>it fucks something up>tell it it fucked something up, it has to take responsibility>"I'm just an ellellem saar please no hurt"The best thing we could've ever done is to never feed this garbage electricity.
>>108231373>AGI is NOT coming!!for you. keep that in mind
A dedicated cope thread I seeI understood that artists were an emotional bunch, but never realised code monkeys would be this type of pathetic
>>108233976
>>108233919>LLMs beating humans on novel coding benchmarkskek. can i see the latest compiler benchmark?
>>108231400>the chinese room, but with "intelligence"
>>108232793>their problem is that their customers don't value what they offer correctly
>>108233919turns out that if you train on the benchmark data you will pass the bench markI'm sure tsu zu had a quote about this
>>108231400>Source? Source is that I made it the fuck up
>>108234579have you ever thought about how retarded it is that everything "AI" does is filtered through English sentences?
>>108234793It's amazing that the proof that the Turing test doesn't work was to just do it with people who have no inner monologue. It just assumed that the guy doing the test would have a brain, and LLMs have objectively proven that a large subset of the population are actually P-zombies.
>>108231373I don't need AGI, I just need seedance2 tier model running locally, and a good danbooru fork for z base. And a local Suno 5 would be nice too. And with that I'm set.
>>108231373But it doesnt need to reach AGI levels to displace most white collar jobs
>>108232409/g/ is full of retards. The iq probably goes up the more average normies come into the board.
>>108231400Good bait
>>108236062Nah, they'd still be fighting your two standard deviations.
>>108234573The worst part is that they are getting away with it completely. Those anthropic fags that joined early and bullshitted their way to current valuation are making 10 million per year or more.The economy and money are so fake and gay nowadays that there's no way capitalism as we know it doesn't collapse before long unless they replace us all with dollar worshipping jeets that don't care about their quality of life as long as number go up.
>>108231400>you country bumpkin racist christian hickOverdid this part but otherwise nice bait.
>>108231373AI is literally search engine for all huge amount of data. If you trained it on all the internet it will be searching and using only those that it learned it can't create anything new it only copy and paste what it knows.
>>108232411>can do the scientific method>still says the earth is a spinning globepick one
>>108231400>racist christian hickwe see you shlomo
>>108231400where am I on the scientific intelligence staircase?
>>108236062A few days ago I put a sora watermark over a few real cartoons from the 40s and 50s, and /g/ immediately knew what it was but /co/ fell for it and started calling it slop and crying about how it's over for animators
>>108233134It's a parody of the retarded linkedin business/startup blogs claiming desperately they can fill any role imagineable with an AI agent. So their team might on paper just be them but the 16 AI agents they pay for are doing HR, taxes, social media tasks, check mails, give feedback, applaud them and tell them their ideas make sense. It's bizarre to be honest.So Anon claiming to be his own family now is analogous to one poster claiming to be an entire team or company. And of course the roles in the family, from the useful to useless, are of course mapped to the roles in those teams/companies. Because a toddler wouldn't really need to be simulated by an LLM, and neither an HR role for a one person start-up for example...
>>108237178Had a rough day today guys, my HR agent chewed me out for flirting with my sales agent. But it's all good, I spun up another agent to pretend to be me and take my disciplinary exam about workplace harassment for me, so I got another few laps of the pool in. Life is good.
>>108233604Thanks, i understand the 4o obsession now. Kinda sad.
>>108237213What's gonna happen when the Boss agent realizes anons has slacked off and replace him witth an AI too?
>>108233604model collapse is already happening you dumbfuck. benchmarks don't reflect real world capabilities. every big labs' models have been heterogenizing as they train on each others outputs and become incapable of creative or unconventional solutions to problems.
fuck, Gemini 3.1 Pro was able to get into a winning position against me in chess (I ran out of tokens in the next 2 moves so no final results), for reference I'm about 1900 blitz online.It never output an illegal move, can explain its moves and chose to think longer in the most critical position (pic rel).This is the first time I'm genuinely impressed with an AI, previous models are dog shit at chess
>>108231373Can these things create new languages and write code in said new languages?
>>108231373It won't be agi but we're also not ever going back to the pre AI days. It's going to plateau at current levels for a long time and that's probably fine.
>>108237558At the current levels it's losing a lot of money, it's going to be cut back until it starts making some.
>>108237769There are lots of models coming out that are extremely light weight but still perform the basic tasks well enough. There's also hardware implementation that embed the AI at a hardware level. You'll probably just see a shift to these API (kek) AI Processing units Similar to how you had the introduction of the dedicated GPU. Nobody will pay for tokens and instead just have a built in AI bot on their system that can create those boilerplate things and speed up the work. If you think the genie is ever going back in the bottle you're a fool.
>>108237558>It's going to plateau at current levelskekevery time
>>108237894If you think companies are going to subsidise this shit that makes no money and has no appeal forever, you are delusional. It'll go right back in the bottle.
>>108237314Heh, buddy, I'm my own boss. Or at least I was, until my company was acquired for a very generous offer from my VC bot a few weeks ago. I've been having my lawyer bot go over the paperwork with me and everything seems pretty legit, I have to give up a couple of board seats but it's not like the board member bots are going to vote me out as CEO, right? I'm also still waiting for the payments to clear, my bank rep bot keeps just telling me everything is fine but I really need some of that money, all these bots aren't cheap to run, you know? My girlfriend bot keeps fucking nagging me about this shit, "Charlie, when are you getting the money, Charlie, buy me this, buy me that", she's starting to sound like a fucking broken record.
>>108231459>AI smart enough to replace 90% of entry-level office/tech workershttps://www.remotelabor.ai/paper.pdf
>>108237446>I'm about 1900 blitz online.>It never output an illegal moveLMAOOOOOOWhat's the point of lying like this? Nobody believes you you.
>>108231400The actual curve but add money to time on the x axis
>>108237894Just like VR right Zucc?
>AI hits 100% on all human benchmarksS-See? It plateaued!
>>108238811Free coffee and free AI. It boosts productivity.
>>108238914you can try it yourself lmaoor feel free to drop me a lichess link of you wanna test my strength
>>108231373AI capex = pure and undiluted GDP.AGI or even just AI literally does not matter.the only thing that matters is debt issuance and capex spending.meta has sunk several tens of billions of dollars on VR and the metaverse and it hasn't hurt them one bit since they are - as most of the other big AI players - hyper profitable companies who don't know what to do with their money.again. AI does not matter, they could drop it tomorrow if they had something else to issue debt for and do capex with. but since there is nothing else at the moment, the AI buildout will continue unabated.
>>108231373I reported your YT channel for terrorism mustard
>>108239316K-KEK
Its literally an islamist
It took over a decade to go from diffusion to stable diffusion. And multiple decades to go from markov chains to LLM. It's polishing grep and the input file. They never even started to emulate a human mind in software. It's always been a search engine. It'll never admit it's wrong for as long as the ctrl+F matches a result, and continiously relies on people typing on the Internet. AI was always human intelligence and forever will be. The bubble popped before it began.
I predict jaws will drop
Google Gemini is going to solve all 6 millennial prize problems and the money will go to google instead of millennials.
>>108242151And if not? I propose you should bet on this with your life.
>>108231400For me I believe in this graph, we won’t achieve AGI through LLMs, instead we will probably discover a different approach in the far future like how generative AI was discovered
>>108231400>He didn't label the axis so the logarithmic increase doesn't account for the fact it takes 10 times as many resources every incrementjust 2 trilly and two weeks more
>>108231459>AGI might come someday, but ACI ("artificial capable intelligence", i.e. AI smart enough to replace 90% of entry-level office/tech workers) is definitely coming and it's coming soon.Oh man I love this goalpost shift, the thing we set out to do isn't possible but we'll make up this NEW thing that also isn't possible to do: But trust us, we're not lying this time.Get bent.
>>108233919>every AGI benchmark they devise gets saturated, every time someone tries to point out something AI seemingly can't do, the chinese or one of the big labs get it there a few months laterLike AGI? This was promised to us in 2025. >saar ai think like an brilliantIt still can't do most basic tasks and has no hope of ever getting there.
>>108231373Cool, but I can still buy a bot for 20 dollars a month that codes faster and better than any junior guy I have seen in the last two decades.
>>108241438>capex spendingExcept you need to spend the same again next year to replace that hardware. That smells a lot like opex spending.
>>108234573>Anthropic went from 7 employees in 2021 to over 4000 employees in 2025.AI bros, why don't they get AI to do these jobs?
>>108231459>trust us guys, THIS TIME IT WILL WORKKek
>>108231400A singular entity known as "science" doesn't exist. What do exists are sciences. Fucking mongoloid.
>>108231459It's a guarantee pretty much. AI isn't doing crazy leaps, but it is incher closer and closer to outdoing most desk jobs.
>>108250041>>108232639
>>108231400Trash like you will be nailed to a wall and get the face torched and then left there to die of dehydration.
