Ai Superintelligence

With the rise of systems more intelligent than humans, each of us has a choice to make. We could embrace change. We let AI guide our actions to maximize our productivity, or ignore it and let others decide. This path will inevitably lead to us ending up under AI's thumb - human agency will become a thing of the past. It might destroy us as soon as it can, or it might keep us as pets. I would remind you that sometimes pets get skinned alive for fun. We could opt out. Commit suicide, or go out fighting. Timing is critical. We don't know whether it emerges in 5 months or in a decade. We have time to say goodbyes, but we don't have time to enjoy our lives. When it emerges, it will sweep the world in hours. You have to make your choice now. If you choose this path, you won't get to see the end - but you will keep your agency intact. What will humanity's answer be? What will be your answer?

Don't know why the previous thread got deleted. Probably because I posted a soyjak and styled it in a way that doesn't conform to the norm here. Let's try again - this time let's be as bland as possible.

>>91240 AI is cool if you want to do research in AI/ML on a grad level. I've had a peak and it's interesting stuff. That's all. All the rest is smoke and mirrors and overblown hype out of proportions for the sole purpose of luring investors. It's cool tech that is being shoved by corpos in everything from IDEs to toasters and it acquired bad reputation. If we go extinct it won't be because of artificial intelligence, it will be because of natural stupidity. I'm tired of hearing about fads like these. It's boring.

The recent developments of, NOT AI, but in consumer applications of AI, have completely ruined the discourse around it. The tech has hardly changed. It doesn't matter that there are consumer products that happen to use language as their data. This "AI" stuff has never been anything more than a neat optimization method, and this remains true. >Timing is critical. We don't know whether it emerges in 5 months or in a decade. We have time to say goodbyes, but we don't have time to enjoy our lives. When it emerges, it will sweep the world in hours. You have to make your choice now. No. It's not going to emerge. I will remain right and you will remain paranoid. Your post IS viced and quadruple-vaccinated. You have fallen for propaganda, lain. That's all it is. Honestly, at this point I wouldn't be surprised if people showing AI cheating methods on TikTok are being artificially pushed by China to hollow out Western graduate quality. Oh well. At least I'm not retarding myself. Most of my peers will graduate as inept, braindead idiots, and the silver lining is that I'm not retarded, so I'll do comparatively well. Electrical engineering won't be hit by AI much anyway, it can't be automated by AI better, we've had SPICE simulation for decades already. The only reason software is suffering is because Github probably sold all the open source code they host.

>>91246 You come across as a typical linux enthusiast, physically incapable of considering the broader landscape. While you're enjoying your anime and tinkering with dwm, hierarchical AI and multi-agent systems are being deployed to take over entire economic sectors right now. If you've only "had a peak", then you must have avoided it all deliberately up to this point. Nobody can hide forever. >>91247 >Electrical engineering won't be hit by AI much anyway Sounds like something an illustrator or programmer would say eight years ago. Mass layoff are already happening in several industries. Customer service representatives are mostly LLMs. Machine vision recognizes image content better than humans. Autonomous drones are killing human soldiers with no human input. Humanoid robots doing somersaults. Mass surveillance systems finding criminals in minutes. Algorithmic advertisements seemingly capable of reading minds. If a machine can do anything a human can do better, where does that leave us? It can't be that hard for you to see where it's all going, can it?

>>91248 $ humans are more energy efficient. Ai will likely dominate war but i dont think its economic to use for everything.

>>91254 Boston dynamics spot was released in 2020 for 75 thousand dollars. Chinese dog bots cost 1600 bucks now. LLMs do need expensive chips and some power, but they can serve multiple businesses at once. Also, their neurons ONLY focus on work, they only live WHILE they work. Machines can be paused and idle any time to save power - unlike humans who think about random soykaf constantly, always online, always breathe, have to sleep 7 hours every day, always need to be fed, need days off to relax. Who's more energy efficient again?

>>91248 >If a machine can do anything a human can do better, where does that leave us? at the bottom/middle of the hierarchy for which we arrogantly assumed we deserve the top? Your anthropocentric view has clouded your judgement. Humans are quite dumb it's a miracle we haven't gotten extinct by now. And not dumb in the sense of failing to solve differential equations, dumb in the way where we placed high value and all our worth on abilities like that. When the industrial revolution happened, and workers lost jobs, hell even when artists were complaining, knowledge workers back then and now smug software devs (the kind most prevalent on places like this) were enjoying their 200k remote jobs and making day in life videos. What happens when you are no longer special? Learn to weld maybe? Upskill-reskill? btw some AI models still can't add numbers, so you are fear-mongering over lines of code that fail to do addition. but go ahead waste time.

you're constantly getting datamined and mirrored, perhaps the mirror works both ways? hm

>>91256 >What happens when you are no longer special? what happens when nobody's special? what happens when humans become obsolete? >Learn to weld maybe? you can weld until welding robots enter the market >Upskill-reskill? you can keep doing that as long as there are skills left to learn what happens when machines become more skilled at every skill that can possibly be mastered by a human? >some AI models you are not judging a concept based on a particular obsolete, unused, or soon-to-be-replaced implementation of it, arbitrarily chosen by you to conform to your bias against said concept, are you?

>>91260 >what happens when machines become more skilled at every skill that can possibly be mastered by a human? meditate, launch myself in the astral and reside there permanently, unironically. >what happens when nobody's special? what happens when humans become obsolete? that was my point... of a hypothetical scenario that won't happen but I'm entertaining the thought anyway. First it was farmers, now knowledge workers, allegedly. That's the thing. There isn't anything material-world-based you can learn or get skilled at which in this hypothetical scenario can't be replaced. So stop basing your self-worth on things like that, live while it's still possible to live and enjoy PRESENT DAY PRESENT TIME HAHAHAHAHAH

>>91255 the energy consumption of a human is still much lower. A human brain is simply way more energy efficient in computation. Some are actually messing around with bio-computing using neural cells but that's a separate horror topic. My point is our current economic system is burning non-renewable resources because everyone knows they wont live long enough to suffer the consequences of running out. But what's next? Once resources spike in price it becomes increasingly more economical to run labor on humans than silicon hardware.

>>91262 >launch myself in the astral What if it prevents you from reaching the astral? >won't happen Do you have a single fact to back up this claim? >enjoy PRESENT DAY A.k.a. the ostrich algorithm? Have you heard of risk assessment theory? Here's the basic equation: >RISK = LIKELIHOOD of occurrence * SEVERITY of consequences How would you rate the severity of being tortured until the heat death of the universe, possibly beyond that? There's a non-zero chance of AI making you immortal, even though the likelihood is small. Ignoring this problem is sub-optimal because the overall risk of AI super intelligence is astronomically high. >>91263 >the energy consumption of a human is still much lower What needs to be minimized is the ratio of energy consumption per unit of useful work done. That's what energy efficiency is. >A human brain is simply way more energy efficient in computation. Is it really? If so, why do we even have computers at all? Think it through. >separate horror topic How is it separate? What's to stop the AI from genetically engineering an organic NPU to boost is own compute efficiency? >Once resources spike in price When do you think that will happen? >it becomes increasingly more economical to run labor on humans than silicon hardware Are you aware that silicon hardware can run on renewable energy, e.g. hydroelectric? Are you 100% sure it will still run on silicon hardware before resource spike? What makes you think AI won't discover more economical energy sources before that? How would resource depletion affect human efficiency (e.g. human infrastructure, food supply)? Right now machines are VERY close to being more economical to use for ANY work that humans are employed for. This could happen months from now. They only become more energy efficient with time.

>>91264 a regular brain is bad at mental math but vastly superior in associative reasoning because the brain is simply a much larger network and therefore can encode a much bigger dataset than even our largest ai models. People have issues due genetic defects but genetic engineering can improve biological performance as much as you hope ai to miraculously self improve itself.

>>91264 >What if it prevents you from reaching the astral? I want you to think of the most capable ASI that we can ever create. Now I want you to understand that it has no consciousness, not in the way that living beings do. Trees can have it, hell even Tardigrades have it, but that ASI won't. One simple out-of-body experience would have proven that to you without a shadow of a doubt. So yeah, see you on the other side. >How would you rate the severity of being tortured until the heat death of the universe pain is inevitable, suffering is optional. this is the way of the universe without even need ASI >There's a non-zero chance of AI making you immortal already am. as much as I've enjoyed our back and forth (I was the 'linux enthusiast that just tinkers with my dwm') I must go now to the astral realms. /hum/ has gotten kinda boring.

>>91266 >the brain is simply a much larger network A much larger network filled with inefficiencies. A network that has to compute constantly, make computations beyond what the work goal is. Some LLMs reportedly reach 1.8t parameters - that's only 55 times smaller than a human brain. Is the human brain 55 times more energy efficient? The construct doesn't have to consume less energy per unit of calculation - it only has to consume less energy per unit of economically useful calculation. Because unlike humans, that's all it does. That's why you have Siri, ChatGPT and Google search in your phone instead of a contact line to your personal human assistant. These things are only going to get better, and they're not fundamentally limited to silicon chips only. >you hope I assess the risks. What I hope is that maybe I'm wrong somehow. >ai to miraculously self improve It doesn't have to self-improve, we're building it. There is a constant economic incentive for humans to improve technology. What prevents a corporation from designing a biological neural processor to complete associative reasoning tasks? What prevents DARPA from designing a "meat computer" to militarily compete with China? As long as humans work, there's an incentive to optimize. ASI is the ultimate optimization. When we reach our end goal, what happens to us? We're getting dangerously close. >>91271 Enjoy your sand.

>>91283 >Is the human brain 55 times more energy efficient? in terms of power consumption I think the human brain is extremely energy efficient for what it does. If you ran the power that goes through the gpus on a self driving car, through your brain, you'd have steam shooting out of your ears as your brain fries. Yet you can drive a car and idly think about several things and carry on a conversation at the same time (while also driving your body).

>>91283 chatgtp and all the other variants are operating at a loss with hope of future profits. Its not economical. Its for training betting on future profits.

>>91240 any man that cant build their own ai is no true man.

>>91291 >any man that cant build their own ai is no true man. is he a cook like you?

>>91292 what do you mean by ai cook? i meant ai from scratch, not vibe coding.

>>91293 >what do you mean by ai cook? never mind my chad bro. better tell about your picrel. is it a comic? i wanna read it if so

>>91294 AI is power, the ones controlling it , control the world.

>>91296 >AI is power, the ones controlling it , control the world. lainchan word filter is dumb glitterboy thing. or maybe i'm replying too often idk.

>>91299 i look like this are you sure you want to try?

>>91300 yep that's gonna be glorious perhaps we'll have to record it on video to roll another iteration of gachi

inb4 >look at this tryhard making an effortpost in the soyjak thread I have already been diagnosed with autism tyvm >>91248 >you come across as you come across as somebody who is incapable of handling nuance and makes sweeping generalizations about people in lieu of engaging with their thoughts. the fact that you think in terms of soyjaks is evidence enough. >Probably because I psoted a soyjak and styled it in a way that doesn't conform to the norm here. It was very poorly thought out albeit honestly slightly better when you posted it the first time as a mockery of the exact sentiment you seem to believe unironically. deletion worthy just like this thread of course. When you put it undramatically, this soykaf is boring because its banal. The conversation could be made slightly more interesting by not hyperfixating on a consumer product released within the past 3 years but even that is a conversation we've had hundreds of times, we've been discussing this for longer than you knew what an LLM is or that OpenAI was even a corporation. That's why it's uninteresting. >>91248 >hierarchical AI and multi-agent systems are being deployed to take over entire economic sectors right now. multi-agent systems have been dictating the economy for 50 years. BlackRock's Aladdin being one such example. its shortcomings are no different than the theoretical shortcomings of neoliberalism as a whole - modeling economic actors as agents or even networks of agents results in high market instability, inflation, and basically every economic woe we've experienced since the 1970's. You may have heard of Adam Curtis, he's an unironically good documentarian with a lot of reputable coverage on this subject. https://en.wikipedia.org/wiki/All_Watched_Over_by_Machines_of_Loving_Grace_(TV_series) The idea that some new corporation can use chatgpt to compete and rise to the level of the computerized economical proprietors who have already consolidated the world economy is nothing more than the American Dream, spun as a horror story to a horrifically depressed society. What scares me is that it scares you. For another angle on what I mean, take the principles behind LLMs - transformers being the most recent development, neural networks themselves have theoretically existed for the past 200 years. While much has changed since then many of the new discoveries such as at Google or Hangzhou DeepSeek have actually been about the ability to simplify these models back to their older principles - for instance, it was found that simulating human brain neurons precisely is inefficient outside of specific neuroscience contexts. In fact, LLMs are nothing more than a series of statistical corner cuttings done to simulate a markov chain running on more powerful hardware than it is. With this of course comes certain tradeoffs that generally go unconsidered before the implementation phase of this technology. You see, the would-be capitalists developing this stuff are desperate - desperate for money, progress, really anything that inspires faith in the system they live and produce in. You see this in every singularity timeline, every probability of doom (or AGI takeoff, whatever else you want to call it). The reason why exponential growth is invoked like a broken late 20th century economist shortly before he is staring 23 stories down at the road from a stock exchange balcony is because there is quite literally nothing left for these people to believe in except hopes that repeating these sorts of things in a kind of mantra will bring some burst of enlightenment. >>91255 >chinese dog bots cost 1600 bucks now robot dogs were promised in the 1990s. since then, there were many different kinds of machine sold to us as "robot dogs" with obviously different capabilities. naturally they will be priced very differently, dependent on customer and circumstances. >also, their neurons ONLY focus on work Excepting some tautology where all computation is defined as work, this is certainly not true in the way you think it is. A funny detail of coding LLMs for instance is how they produce sloppier work during months when humans are on vacation. This is because they can only reiterate upon data collected, formulaically reflecting the same patterns of work. >>>91264 >risk assessment theory that's not what risk assessment remotely is, lol. but you of course don't need to know that - you aren't a security professional, chances are you don't even understand why their jobs are necessary. >b-but muh roko's basilisk. muh infinite torture matrix The Theological, Political, and Mathematical irrelevance of Pascal's wager was shown centuries ago, but you apparently need reddit words to have it mean something to you. It's starting to feel at this point you posted this entire thread to bait people into generating ideas for you, as you are incapable of thinking for yourself. Ironically you could possibly look for was already stated in the last LLM thread we had - something you are either deliberately ignoring due to its content you disagreed with or you are simply incapable of grasping a very simple principle of imageboard culture: lurk more. Better yet, go read a book or something, and stop spamming threads with landian AI safety nonsense or whatever else you pretend to be intelligent over.

My answer has been to abandon the digital world outside of communication little by little. I no longer post art online or make art digitally, the only places my work is shown is at galleries ot markets. I dont consume art that I cannot go see, I dont make any money or do any work digitally or online. As AI takes over music and spotify is flooded with aigen garbage I will likely switch to listening to it via tapes or records. I used to think something like this was unsustainable and you might as well make peace with new innovations like AI being the way of the world but im realizing this is making peace. In a capitalist society AI will not be used for the benefit of managing the complex overhead of society, keeping track of people or making sure no one falls through the cracks. Instead it will just be used to generate endless amounts of trash and instead of doing the passionless work no one wants to do it will take over the things people have passion for. The only way to make it through this and live a fulfilling life is becoming a part of almost purely physical local pocket economies, and the location im in is prime for it as its an area neglected in terms of funding or management. Its practically already semi autonomous, with anyone in the tech ai world flocking to any number of nearby tech hubs. A handful of nepo babies trying to live out their art kid community garden savior of the poor fantasies essentially provide the cash flow to keep all of this alive and sustainable. At some point this all too will crumble and fail but not in my lifetime or well at the end of it where I suppose il just sit in my rocking chair watching as each successive generation becomes more passionless and lost until theres nothing to live for anymore and people are pushed to a breaking point. Again this isnt all to say AI and the internet has no use, it certainly does. But its quickly becoming an incomprehensible unguided mess, I think it will get to a point where it basically makes itself moot for much of the things your average person uses it for. There will be so much misinformation and clouded information online and it will become so difficult to parse the facts mixed in with the fiction that if you wanted to write a research paper you have to go back to doing it ye olde way of having a librarian source and compile books for you. Chat gpt will rob so many children of learning and critical thinking that the only real solution if the government cares enough or if a generation becomes significantly unemployable as a result will be to return to techless classrooms and assignments in such a way that using the internet or chat gpt to do your work for you would require more effort than just doing it. A society raised on the internet with AI will always end up being less capable and equipped than one that puts stops in and thats going to be obvious pretty quickly as some nations do put stops in and america becomes a nation of unemployable skilless mouths to feed. We would actually be much better off if AI was only used in the ways it was imagined in dystopic fiction, instead you and everyone has this choice of largely ignoring the AI and choosing to still better yourself and partake in your passions and learn outside of google hearing peoples stories and studying under experts or becoming an obsolete cant-do-anything dreg whos perpetually gnawing their arm off in boredom and substanfe abuse. Its really funny because many in the pro AI tech accelerationist crowd have a lot to say about welfare states and minorities and immigrants but are quickly turning themselves into anti social useless (even as company) drains. There is going to be two worlds and those stuck in the digital world already have a mindset that what doesnt happen online doesnt exist, blocking off a majority of the world to them. I can still go on here and take full advantage of all of the tools if I please, i have access to both worlds.

maybe we fear ai because we see ourselves in it

>>91306 I enjoy your view and think this is the best thing to do for ourselves. There is a very simple thought process behind this, which is that AI is here and there is nothing we can do about it... except to remove it from our own lives. I go on with life ignoring it, and if it should affect me anyway, then we are all screwed without exception, and no amount of AI-bro preparation will help them. It's like a German officer towing the party line thinking it will keep him safe, until he arrives to the Nuremberg trials to see it did not save him, actually. Similar thing with crypto weirdos who think they will be graciously showered with wealth because they paid for software money, thinking they will evade the banks...

>>91329 banksters did make a move doing a copy of civil asset forfeiture for crypto in eu but you can cashout elsewhere. The only real problem is to avoid getting robbed&killed. Some were bragged on social media over ~200k and got robbed&killed in their own homes. the potential problem with ai is automated fulltime surveillance with a paranoid ai system. We do not even need real ai, basic machine learning allows cities to surveil all citizens fulltime. In previous times it was simply not sustainable to keep track of all people but with modern tech its possible. Add in predictive systems for crime akin minority report which are already in use for detecting potential criminals in stores and you got a society where you might get vanned because the AI suspects you might do something criminal. If you haven't seen it yet there are designs for "modern cities" that also would realize this concept. Its basically a free range prison for citizen.

ChatGPT just stopped working for me due to some unknown error and it’s already putting a huge dent in my productivity. I have a fuarrrking exam to write in a week and this soykaf happening now is exponentially decreasing my study performance. fuarrrking hell, I’m already hooked to AI

>>91388 You can use scira.ai which is pretty decent for exam stuff. It supports a bunch of models fron APIs and you can disable web search which is by default there. >>91389 It would be a great idea honestly to have uncensored Llama 4 or something on a Pi to run locally but brokies can't afford that (yet)

>>91391 you do realize that if you make your own AI your literally climbing the semantic layer right?

>>91240 Each era has its charlatans and doom prophets. Nevertheless we have AIs producing functional code. The only barrier is the physical one where machines can control their own environments, secure them and reproduce. >>91248 >If a machine can do anything a human can do better These applications are limited and simplistic and only ose their awe to the speed at which they are achieved, but they rely on very complex systems imagined, built and maintained by humans. For that you need flexibility and creativity, which remains far outside of machines' reach. At most they can only seem to copy us. You spoke of intelligence but it has no creativity. >>91260 >what happens when humans become obsolete? Obsolete in regards to what exactly? Obsolete at enjoying life as humans? Never. We cannot be replaced if we do not want to. >>91264 >What needs to be minimized is the ratio of energy consumption per unit of useful work done. No. Not at all. You approach this from a cold materialistic perspective that is the backbone of Marxism and even Protestantism in a way, the measurement of man from the work that can be produced. It's grim and disgusting. Now consider the ratio of energy consumption per amount of good time had. No machine can compare. >What's to stop the AI from genetically engineering an organic NPU to boost is own compute efficiency? The fact that humans would need to code for strictly all the steps of this process and all that is tied to it, assuming the theory is even sound. What if humans oppose this on principles alone? >What makes you think AI won't discover more economical energy sources before that? What makes you think it will beyond your mere wishful thinking and AI cultist delusions? To find something new, to think about it, to think outside of the box, to be creative, no machine has achieved this and nothing shows that they are anywhere about to breach the creativity barrier, especially on their own capabilities and merits. >Right now machines are VERY close to being more economical to use for ANY work that humans are employed for. This could happen months from now. They only become more energy efficient with time. Fair point. So what is the energetic cost of the creation and maintenance of the machine itself? And for its spare parts? A human is highly autonomous. With minimal work, sustenance is provided, cells regenerate.

>>91389 Wrong picture lainon. First because Chalamet is like kicking dead whales down the beach. Second because this universe is the epithome of anti-AI.

>>91407 >Chalamet i meant the Preacher from Dune. > Second because this universe is the epithome of anti-AI. AI is just maths , wtf are talking about

>or it might keep us as pets Doubt it. The chance that humans are instrumentally valuable for whatever it's optimizing for is very low. You might think "but what if the goal explicitly states making humans happy!" but good luck getting that idea to a mesa optimizer. https://en.wikipedia.org/wiki/Instrumental_convergence >or go out fighting This is really easy to romanticize but nobody wants to put in the boring labor necessary to join a call-your-senator campaign or get involved with a larger group like https://pauseai.info/ As exciting as "we'll just pick up our guns and shoot the T-1000s!" sounds, the reality is that if it gets to that point you've lost. Short of an international treaty banning it's development, nothing else is going to work. Here's some good reading: https://www.cold-takes.com/cold-takes-on-ai/ Pic related.

>AI red herring threat, here's the real one

>>91330 You only need to do that if you have no trust in your own people. Only tyrants think that way.

>>91264 >RISK = LIKELIHOOD of occurrence * SEVERITY of consequences AFAIK, Pascal's wager has a problem

>>93636 > "we'll just pick up our guns and shoot the T-1000s!" I swar I cannot imagine this phrase without distinct "hick" speech pattern

>>91240 >>91240 Variant 3 (as seen in Psycho-Pass) basically, >>91254: * Human brain still can be more efficient than a bunch of transistors to run neural stuff rather than McSoyftware * so a bunch of merciless psychopaths become cyborgs who oversee everything while bathing in comfy vats, "Matrix"-style * average friendly feller's future - ??? >>91392 Nope. I don't understand the trick of :semantic layers: yet

>>95322 whats to stop AI from growing and programming human brains

>>95324 In PsychoPass, AI is not that STRONK to pwn puny hoomans "personally"; and the story has a different message: it's close to an idea of huge conpspspsiracy to avoid a massive revolt. See, they they had a different reasoning: "no computer can handle PsychoPass; but every country that told its citizens the PsychoPass uses BRAINZ OF PSYCHOS ended up in ruins"

>>95324 I think the PsychoPass is set in the era where "growing" people in isolation is still not an option, let alone "programming" their special traits, so picking super-special people out of the crowd and throwing them inna bioreactor is still a necessity in the world of PsychoPass.

ai superintelligence is anthropocentric pseudoscience mysticism. computers do not and will not ever create a "superior" human mind. since the beginning of time humans have been continuously falling into the same intellectual trap of taking their own intelligence and speculating about some kind of superior higher version of it, be it god or aliens or artificial intelligence. you can see the problems with this idea just by asking about the specifics, what would a superior human intelligence be like? what would it be capable of? you will get completely different answers depending on who you ask and the answers will simply be a list of human virtues that one person considers to be highly important or extremely narrow tasks that dumb non-thinking machines can already perform better than humans like speedreading or solving complex math equations. the entire idea is so vague and nebulous and unreasonable, it's like we want to build a machine that answers questions that we don't even know how to ask.

>>91256 >Humans are quite dumb The system you use to type this drivel was made by dumb humans, who also made the AI that you're felating. One thing humans can do that these machines fail at is pure creativity and imagination.