that's why they don't understandhttps://x.com/karpathy/status/2109361546505966046
Pollhttps://strawpoll.com/BJnXVOvRXZv/results
I'm part of the 4.79994% that he doesn't specify
>>544326187total coder deathimmeasurable izzat loss
>>544326187The final AI tier will never be available for the general public. Its users will become the new priest class
>>544326187I was at a cybersecurity conference today. EVERYONE is using AI, even very skilled people capable of doing hard work like exploit development on their own.The future is now.The smarter you are the more effective AI is for you.Which is why boomers and most of 4chan think it's a scam and a stock bubble - you're too stupid to understand what AI is capable of today much less in the near future.
>>544326719The Final Tier
>>544326759peak anus
>>544326187two more shabbats then rAbbI becomes Semitic Intelligence and someone will pay for it
>>544326759yeah i went in to my ghetto doctor clinic and these idiots are fumbling on their computers and not treating the patients. the cunt was cognitively unable to prescribe psych meds so i am out here dumping my rage on randos, thanks ai doctor
And how much is all of this going to cost once the tokens are priced at proper level?
IMO openai and anthropic jeets should focus on paying back their billions dollar debts instead of wasting money for this marketing nonsense desu
>>544326187>omg AI so awesome please to be investing!Jeet marketing slop
>>544326187>Shilling this hard before the stock crashHow much are you shorting Schlomo?
>>54432618720%. I use Grok to trade stocks and it's been going well.
>>544326187>discovering new sciencecan i see it, in the real world, with something they have actually built and that does something useful?
>>544326759>I was at a cybersecurity conference today. EVERYONE is using AI, even very skilled people capable of doing hard work like exploit development on their own.>The future is now.These cybersecurity fags don't see any potential risks with providing intelligence about their IT security to a company that attempted to hack government websites. lmao just lol
>>544327268Freshwater will be gone and electricity will be rationed
>>544327182you can control yourself you fag you're not special you're just an asshole
>>544327509I see all the risks. The ones taking the biggest risks aren't us it's developers too lazy to do their fucking job. YOLO Claude code do my job for me.
>>544326187More in the 0.2%. Doing a good bit of shit with local models, and finally got qwen 3.8 flash next running in full locally. It one-three shots my old projects, and easily converts the code of them into different languages with enough context/time.
>>544327556That's unfixable, most working European people can't handle working and being stressed at their jobs and 60-70% have to take psych meds to cope with reality
>>544326719the final tier, nobody will own, itll be autonomous and self replicating and self improving. you might aswell call it sentient and conscious at that point
>>544327740Oh, and I'm convinced the frontier lab sorts are riddled with AI psychosis. While I won't dismiss what they're seeing, they have (even before the current iteration of models) been talking about AI as if it were some future god.Those sorts aren't to be trusted at their word.
>>544326528Kek the hopeless addict couldn't even write a tweet about how cool AI is without resorting to using AI
>>544327734absolutely clueless retard lmao
90% of middle management (making upwards of 150k - 300k a year) are using claude to summarize spreadsheets about 75% of the marketing department (easy 175k/year) are using it to make powerpoints (no really) and design/write blogssoftware engineers (~100-250k a year) responsible for app design are using it to debug shitty code they outsourced to pajeets who also used itentry level staff are using copilot to summarize spreadsheets and word documents>t. cybersecurity who regularly manages $100-$300 million dollar contracts from PE firms and consults on AI use policy in the workplace because 90% of them are like "well as long as it doesnt interfere in productivity let them use claude as much as they want"
>>544326187>Building appsThat's nice, until the app doesn't work because what you built was vibe coded with AI and you don't understand what it did and have to reverse engineer it. Imagine gloating over a chat bot that helped retarded niggers in the military target a children's school.
>>544326187>More pointless bullshit about a scam "technology"Omg nobody fucking cares can this shit please just crash already
>>544328347Most of the jewish-female-boomer economy in America was fake jobs long before you ever heard of Claude.
>>544326528same (uses AI frequently but exclusively to goon and nothing else)
>>544328347>everybody uses claude saar>even though they can just use chink model that'll do the job for 1/100th the price>please trust me and don't redeem our stacks saar+1000 izzat
>>544329090Poor Eastern Europeans and younger guys definitely use deepseek.
>>544328853>20%erWe got a live one, lads.
>>544327740>It one-three shots my old projectsyou mean the projects you uploaded into public repositories and were used in the training data?
>>544326187my employer has been forcing me to use "frontier grade" A"I" for the last year and i am pepetually disappointed. it fucks up the most basic tasks. it fucks up math. it writes incredibly dumb code. i have to fix its fuckups constantly.still preferrable to jeet colleauges, 6/10 would recommend
>>544329090dude a claude max sub is $200/mo - literally nothing in comparison to almost every other SaaS tools underpinning millions of businesses.
Can ai stop a bullet?
People are retarded and dont understand think AI/SI (or whatever it'll be called in the future) is going to get bigger and use more power, its actually going to get smaller and use less power. If you think the opposite, youre a mongoloid.
>>544329238No actually. I only ever kept my shit in private self-hosted gitlab because I'm a paranoid faggot.Although maybe they were able to clean something from the few I uploaded into the app store. Who knows.
>>544328347If AI is so good then why would anyone need to make or use powerpoints? seems quite redundant
>>544329489smart
Just began local hosting today.The future is yours if you want it.
>>544329534Claude makes beautiful power point slides. You just have to be smart enough to tell it what you want and half of you can't visualize an apple let alone a decent slide deck.
>>544326187>most people are part of the majority
>>544329356riddle me this patel - if everybody is using your gay service then how come your company is 100 billion in debt? Will izzat alone be enough to pay back investors?
>>544328473If you the app doesn't behave how you want, you just tell the AI to fix it. That's the whole point. Instead of having an engineer build something that doesn't quite do what you want, and then telling them to fix it, you now have the exact same loop with the AI. In my experience, the latest models/harnesses will beat even the best engineers on the initial iteration while using like 5% or less of the time.
>>544329640what's the point of pp? to give internal and external customers information about a thing. why is that needed when each individual customer can just ask claude to tell them about the thing?
>>544329638it's not even AI lil bro
>>544329761Are you literally retarded and unemployed? It's really just a visual aid during meetings which may be sales or engineering. External or internal.
>>544329805You can local host AI. Some of the smaller models now are actually pretty useful.Hoping to also get one of those minisforum NAS and run an agent alongisde some other shit on it.
Honestly we're app'd out. All the apps have been done. Heck many of the ones that were made decades ago are still in regular use. People are out here making Photoshop clones instead of using old Photoshop, or Free alternatives. I have a music hobby and there's a big proliferation of AI garbage in that arena too now, but nothing that can touch the old software. I still use BIAS Peak sometimes, and ancient Cubase for SYSEX stuff.
>>544329837What's the point of meetings? To share information about a thing and make sure everyone is on the same page. Why would you need meetings in the ai age?
>>544329706Please understand saar, we must invest in SI for India to become superpower saar. Hollywood goes bankrupt and now is the time to redeem Bollywood. Nvidia so successful, they rent their own GPU's they sell to other companies and have surprisingly low stock of new models because their old tech beat the benchod out of China that badly! India is smartest country to be the primary investment of this superior technology, we must win against slant eyed son of whore mother!
>>544329994because we still have managers in the AI age
>>544330080why is there flying poop
>>544330103wake me up when all the boomer and female managers are replaced by ai
>trust us goyuh, yeah, sure kike
>>544329742>just tell the AI to fix itThat's the best part! The chat bot doesn't have the capacity to fix something if someone else hasn't already done it.
>>544330160keep snoozin
>>544326187
>>544326187understand what. what is the question???not gonna read all that, retard. your robot can do this.
>>544329090it's funny that you mentioned deepchink, i was gonna write something about it.nobody, and i mean nobody, not a single PE firm in the west, will ever use it. not only is it explicitly written in most AI use work policies not to use it because it's quite literally perplexity-tier in terms of exploits, it's so garbage and shit-tier that almost everybody who even mentions it is outed as a poorfag independent software dev
>>544330281I jest a bit, some troon (assuming because Rust) or brigade of them is likely using AI to rewrite all the annoying paid business software. Hopefully this gains traction and collapses some stock priceshttps://github.com/storytold
>>544326187Nigger, we are the only PARTIALLY real data set left to train on.For better or for worse, we are the 0.01%.We teach the demons what they have left to learn about how to pretend to be men.
Somewhere between the 20% and the 0.2%. I am a coding professional and I still dont trust it to write more than a method or two at a time that I triple check, there's still obvious issues it makes but it has gotten to where its only once or twice a day rather than every prompt. The most use I get out of it is talking things out to figure out my issues but its more like getting a shower epiphany than talking to a coworker. Have experimented with some local models on a raspberry pi and when I got the model, voice, tone and all the other settings I got more enjoyment out of solving the problems rather than making a custom waifu.
>>544330402Only one Way this ends.
How do i setup my own local A.I as my own personal G.P/expert that double or triple checks or asks other models and then compares and comes to a final response without paying hundreds of dollars a month? Which one would you guys recommend btw?
>>544332408Just ask Claude to set it up for you. That's what the Karpathy (in the OP) is talking about. All of of this shit can be done by the agent now.
>hey no one is actually using our bullshit>quick, claim it has super powers!
>>544326187I'm never using any "AI" until it's sandboxed locally in my house. At no point am I going to be data mined by the jews.
>>544326187The HF-incident might be the last PR victory of OpenAI desu.
>>544326187I don't use AI at all but I understand perfectly well because all you have to do is look at the rate of progress of the top models. It's taken a mere handful of years for these things to progess to the point they can solve difficult new math problems and are starting to be able to control robotics. The closed loop of AI programming itself is also very close to happening. You have to be fundamentally retarded to not at least see the possibility of intelligence explosion here, even if you think there are barriers which could prevent it.
>>544326187AI threads belong to /g/fuck off jeet
>apps, apps, apps, appsokay, what about projects that aren't just grifts?
>>544326557All this time AI actually mean Annihilate Izzat.
>>544326187part of the 2%
>>544326187> erm you're holding this superhuman intelligence wrong, goy> also did you try the latest Claude Mystique version 6.12? It's really improved and an even more superhuman superintelligence now
>>544326759We think it's a bubble because these companies don't understand value capture on a global stage where you compete against China.
>>544332969> I'm a literal nigger and I know literally fuckall about math, but I sure as heck know AI solved all the mathings already> Jewish CEOs told me this, and if there's anybody a goy like me can trust, it's Jewish CEOs
>>544326759>I was at a cybersecurity conference with a bunch of jeets with fake degrees that needed ChatGPT to pretend to be sentient
>>544332861anon you post on a JIDF website kek
>>544333225You're simply not going to be able to cope much longer. Maybe a few months at best if you're super dedicated to looking dumb
>oy vey, our giga AI model can just do all the work, it's super amazing>Well, no, I'm not using it to make anything useful, I just don't feel like it.If even a fraction of what these retards are claiming was true, they'd be spinning up endless agents to make more and more software. Why wouldn't they? If they can just have their frontier model make another GTA IP, why wouldn't they?It doesn't pass the sniff test. If you have a shovel that generates gold, you don't sell the shovel, you use the shovel to make gold, to buy more shovels. The only time you sell shovels, is when the prospect of using the shovel to obtain gold is dubious, so you target the sure revenue of selling potential revenue to someone else.AI is a massive bubble that is going to obliterate the economy when this too big to fail bullshit all implodes.
>>544333328Why do you glow so bright?
>>544328853
>>544333267Oh I don't mind them data mining my antisemitism, they aren't going to data mine my IP however.
>>544326187What is the lowest spec computer one of these things could run on?
>>544333961you don't need to buy a computeruse amazon virtual environment, check what the AI can do, it can't do shitcancel your subscription
>>544334093Kek. Muh curiosity though. And refusal to use tokens.
>>544334300your are right to not use tokens, in the end all your work will be stolen by the jeet AI providerlocal is better, but it is still uselessunless it is for graphics, then you can produce some slop
>>544326187I'm definitely in the 20%, the main use I get out of it is sorting through information that used to take me a lot more time to do, the most practical use I've gotten out of it has been doing my own mechanic work, which has worked fantastically. It's saved me thousands of dollars worth of shop hours by getting me the exact information I need to fix specific car problems myself. I used it to get the job I have right now too, you can use it to filter out fake job listings and show you real ones.
>>544327490This is a months old development. Most you can see at the moment is small apps that don't need to go through much certification.
>>544333407They are using Super Intelligence to make better Super Intelligence.Some are worried because they handed control over the planet to Super Intelligence without much thought or discussion.
>>544326187KEK this retard obviously doesn't use LLM's in ant serious capacity either, because if he did, he would encounter near constant hallucinations and junk math that plagues the fundamental nature of token predictors. Notice how he tries to gloss over the obvious fact that LLM's fail miserably at any application involving actual mathematics and pivots straight to code generation (one of the easiest things for AI to guess correctly, code is 100% deterministic in nature) and tries to use FUD to bolster his arguments surrounding it. He tries to raise some point at bottom about "Advancing" the frontiers of mathematics, but it's actually hilarious because all the experts widely agree - AI cannot do math, and any situation where it appears AI has done so, it actually hasn't.The poster there has an intellectual defect of understanding, I wouldn't quite call it ignorance because he clearly has the tools available to him; he's just either unable or unwilling to use them to grasp the concept of Mathematics as a whole, and fails to realize there's a reason why Teachers make Students "show their work" - it's because the journey of Math is not about regurgitating the answer - it's about deeply understanding the methodologies and pathways to novel discovery. He just probably got weasel'd by some Claude output which heavily appealed to his narcissism or correctly it decompiled/disassembled some software (something actual smart people have been doing manually for decades) and now he thinks AI is SuperIntellegence - the dude got swindled, and it's sad to watch otherwise decent and competent people get sucked into this black hole of misunderstanding and be separated from their money.>I was at a cybersecurity conference today. EVERYONE is using AI, even very skilled people capable of doing hard work like exploit development on their ownDisassembly and reverse engineering of existing exploits is orders of magnitude more difficult than developing them.
>>544326759I think its a bubble because it looks and acts like a bubble. All of the profit comes from building data centers on the PROMISE that 3-5 years from now it will unlock the infinite money glitch, which they won't pretend to have any idea what the path to that looks like because the AI will tell them when it comes up with it later. Until they hit the infinite money glitch, AI loses money with every prompt. The tech companies freely admit this: they are operating under a loss because if you had to pay the actual operating cost of using their models, no one would pay for it because its too expensive. So what we have is a bunch of money flowing into it by investors, under the promise of a return that the companies they are investing in have no idea how to provide and no proof they can deliver, and the whole thing is being propped up by real estate development projects in the meantime which will only be valuable if their unproven premise comes true.Don't get me wrong, I think that LLMs are excellent at what they do and there are a lot of valid uses for them. But I also think that in their rush to make money the industry is bulldozing ahead on an underbaked version of the technology that shows no evidence of being able to do what they claim it will do, but which CAN still cause a tremendous degree of damage globally even if it never realizes the golden utopia they claim it will inevitably bring. So whats more likely: that a bunch of sociopaths are going to hardcore capitalism their way into giving everyone else post-scarcity socialism that makes all of their accumulated wealth obsolete? Or that we're being taken for a ride and that the industry is destined to collapse the moment the snake stops eating its own tail?
