>I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out.>I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really).>It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely.>The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture.>We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.>I worry.https://xcancel.com/mitchellh/status/2055380239711457578
>>108832620>I worryWhat a fag. This is hilarious, I hope LLMs drive CEOs insane and they run their companies into the ground and billions die.
>>108832640you will die tho, not the ceo's lol
>>108832640keyed and fippy
>>108832648do you fear death?
>>108832620what a fucking luddite liberal faggotKEKYPOW
>>108832667yeah? what about it.
>>108832620No matter how you twist it, the final evolution of the niggercattle feeding lot is the niggercattle slaughterhouse. Accelerate.
>>108832667I've been dead for billions of years. Didn't bother me. Why would I fear it?
>>108832704yts are so cringe I stg
It's not just the tech industry. you have literal state governments losing their shit over thishttps://www.wuft.org/2026-05-14/florida-pushes-uncharted-legal-territory-openai-probe>But what if someone with far less innocent questions wants help?>"When was the last school shooting?">"How many victims does it take to get on the media?">"What about 3 plus at FSU?">These are some of the questions authorities say the man accused in last year’s Florida State University shooting asked ChatGPT shortly before the attack.>Florida Attorney General James Uthmeier is now investigating OpenAI, which runs the chatbot.>"AI is supposed to support mankind,” he said. “It is supposed to help mankind. It is supposed to advance mankind, not end it. And unfortunately, what we've seen in our initial review is that ChatGPT offered significant advice to the shooter.">A man who's accused of stabbing two University of South Florida students to death in April also sent ChatGPT questions.>That case has been added to the investigation.>"If this were a person on the other end of the screen, we would be charging them with murder,” Uthmeier said. “Just because this is a chatbot … does not mean that there is not criminal culpability."Like what the fuck do they expect the text prediction algorithm to do about it? People asked factual questions and got factual answers. Normalfags are so brainraped by AI that they genuinely think it's a living, breathing machine that should know right from wrong instead of just being an upscaled version of the search engine they've always used.And it's impossible to discuss this with normalfags, either. You can't go up to them and say "You know this is a computer program, right? It's not reasoning anything, it doesn't know what it is telling you, it's matching phrases to questions." and they wouldn't get it. worse, they get mad at you for ruining their illusion. You see that with normalfags on /g/ with sub-optimal IQs
>>108832756>>108832620I'm losing hope in humanity being able to see past literal puppet shows, especially in this case with a text prediction puppet.
>>108832620https://m.youtube.com/watch?v=OHQRo3Uz_VQ&ra=m
>>108832620Imagine that it's the pre-industrial era. And you are building something and you require a metal bracket.So you go to the local blacksmith.And he will take a piece of steel, of his own custom metallurgy. And he will forge it into an artisan masterpiece of a metal bracket. Nothing will break that metal bracket. And it will have a nice beautiful blued finish all around.That metal bracket will cost hundreds of dollars in today's market. So nobody will buy it. They'll just use a CNC machine to cut a shitty bracket out of generic Chinese rolled steel. It'll get the job done because engineers have designed around the assumption that you'll have to use some shitty cut bracket that costs next to nothing to make. "They don't make em like they used to."But they also didn't have 8 billion people to feed back then. And I'm not advocating that it's better the way it is now. But it happened.There was an adjustment period. Products got shitty for a while as artisan craftsmanship was replaced with industrial sloppification but they slowly got better again and now nobody thinks about it anymore. LLMs are now doing the same thing to semantic tasks such as writing and coding. Replacing artisan craft with cheap slop. Even if we're generous to the naysayers and we say that the code is 1% as good as human code... Well horses were fine until we started measuring mechanical work in horsepower.
>>108832782>artisan craftsmanship was replaced with industrial sloppification but they slowly got better againNo, the people who knew what a reasonable product looked like just died. Your lifestyle and consumption habits will keep getting more and more subhuman.
>>108832620>I can't name any specific peopleThen you just made everything the fuck up
>>108832620>Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.Why is this so obviously AI written?
>>108832831retard. he's likely referring to jared sumner and co.
>>108832782llms are nowhere near close to new industrialization (tm) lol, you scientism fags make me cringe
>>108832854>likelyDid he name anyone specifically? No? Then he made it the fuck up. Goddamn, when did EVERYONE stop caring about evidence, proof, facts, etc. and instead just believe whatever agrees with them? Holy shit, 99% of this world is retarded.
>>108832865alright autist who let you out of your cage
>>108832864They're the faggots who fell for crypto and the metaverse a few years back, desperately telling us THIS TIME ITS THE ONE
>>108832854>>108832875I hate you disingenuous faggots so much
>>108832864Better than talmudism
>>108832620>>108832756yeah it's all industries and basic infrastructure. AI is just the sort of bureaucratic defferal of resposibility on crack
>>108832756The feed forward limitation can be negated with how chatbots usually work - by filling most of the context window with previous chatlogs, allowing access to previous context and the ability to self correct. To a degree this just shifts the burden to the other constraint. Context windows simply scale with compute and vram, so people are trying to throw more GPUs at LLMs to see what other emergent properties they might show.I think the *main* architectural issue is the fact that at the scale these LLMs are currently deployed, we can only do inference but not training. The fact that training a competent LLM is a multimillion dollar a week endeavor means is still prohibitively expensive to try some live-training-and-inference scheme, so the best you have is relying on the emergent properties and just spamming the context window with "learned" knowledge.Since current LLMs can only "learn" as much as their context window allows, with even AI agent project trying to monkey patch this architectural limitation, no matter how big you make the context window we probably won't really touch anything rapidly self-improving like the movies.
>>108832620>Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.this series of sentences brought to you by slop generators, and means nothing
>>108832782>And he will take a piece of steel, of his own custom metallurgy. And he will forge it into an artisan masterpiece of a metal bracket. Nothing will break that metal bracket. And it will have a nice beautiful blued finish all around.no, it'll turn into a useless lump of rust within 100 years because he hasn't figured out how to galvanize or zinc plate his products
>>108832620>yeetOpinion discarded.
>AI is le bad!>why?>because.... it just is ok????so tiresomeclean your room and find a job already please
>>108835010>get given millions of examples>ignore them like a fucking retard and go duuuh why?Kill yourself.
>>108835470>trillion dollar industry and other millions of people are le wrong, I'm the right one here chudyes sweetie, I'm sure you are, now clean your room please
>>108832620This faggot sold Hashicorp to IBM.
>>108835546Indian scam call centers are a trillion dollar industry as well. Good on you to not attempt to refute the facts about the millions of examples given to you though.
>>108832620The only AI psychosis I see is people who are too scared of losing their jobs in the short term and want to stop the possibility of AI utopia
>>108832854he's not. he's himself fully into ai.https://news.ycombinator.com/item?id=48154252this twitter post is another mediocre attempt at getting attention.
>>108835815>>108832620Robot Parrot/Neuronal network isn't replacing anyone if it can't even work with Photoshop and other quality human software. Imagine not even being able to work with digital tools as a robot. Imagine not being able to understand math as a robot, that one was exceptionally embarassing for artificial intelligence and shows that it's a scam.If this was the 80s, 90s or even 2000s you wouldn't have so many gen-Z(ombie) investors throwing money at obvious bullshit cause they'd use their common sense, but zoom zooms have NONE. What's worse and what everyone ignores is that these gen-X trillionaire jews had gen-Z children right now who are managing their trillions. And millennials are having gen-alpha which don't even pretend to be eco-friendly and moralistic like the gen-Z hypocrites, gen-alpha just don't give a fuck to the point they don't give a fuck about their own survival.
here an actual ai psychosis: hn luddites think this guy is a bot lmao https://news.ycombinator.com/item?id=48156996
anyone got that meme chatgpt/claude/gemini comparison image about> you ruined your shirt and it's your fucking fault?can find it for the life of me
>>108832620He sounds like a faggot worried he is going to lose his job to AI
>>108832620Thanks for this completely useless blogpost, faggot
>>108836115>>108836195i bet you're too retarded to even understand his post lol
>>108833386this is a pretty standard criticism of large scale systems. The primary goal becomes legibility/metrics of what is contained within the system. You lose the ability to understand anything that can not be represented by metrics that can be surfaced to higher levels. The actual functioning and risks exists at a level that cannot be discretely measured and communicated. You are able to have metrics that allow for larger scale growth which increases net productivity while spreading inefficiency and inability to maintain things and smaller scales. Eventually the machine/managers can no longer control the system through their metrics because the actual underlying structure has failed and they have no means of coping with that, and the system dies.
>>108832620his tactic is unwise. You cannot convince people to stop being delusional.
>>108832640There have been ~3 times ive seen CEO's go insane & run their companies into the ground.Each time the govt. bailed them out & you ended up footing the bill.
>>108837399/g\ is not employed enough to know about shitass aws setupshttps://www.youtube.com/watch?v=y8OnoxKotPQ
>>108837663and that someone made a 'declarative' 'easy to use' json interface with a very customized version/build of angular2 to generate browser client ui for banking applications and then left the company and it has no documentation and it's almost impossible to debug
>>108837693this is not an llm skitzo post, just an anecdote, btw
>>108832620Moat people are so fucking weak they're desperate to find ways to dump their responsibility to something/someone else, AI just happened to be that golden oppourtunity, of course they'll realise it isn't that easy once they lose control of their lives but it'll be too late then.
>>108837399Why did you spend so many tokens just to say "the management have their heads in the clouds"
>>108839828it's not about management in particular it's about large scale systems in general. Large scale agriculture, institutionalized medicine, politics, advertising, economics, etc.
>>1088327824 bed houses used to cost $130,000 adjusted for inflation in desirable suburbs in coastal states. Not only were things made better, but they didn't cost as much to purchase, and everyone made more money. Things never got better - only more expensive with shittier ways to treat their customers. The fridge you buy for 10x what it would cost 30 years ago is going to break in 5 and serve you loud ads the entire time.AI will shut off your power because local governments serve data centers instead of citizens.There needs to be large scale riots.
>>108835010>Shuts off your power>Increases your taxesNothin' personal.
>>108836016>[flagged]they shut him down kek
