Their denial and hate is not based on reality. Their criticism of AI is like 2 years old. They refuse to accept the fact Ai is here FOREVER.They will freak out when robot spiders start painting walls and fixing cracks. They will try to destroy them and they will fail.
>>108231188AI is here forever, but it won't be able to do what the shills claim for decades.
Art youtubers told me that this would destroy AI by 2035.
>>108231188>twitters algo feeds me irrational AI criticism to further feed my retardation and inflated expectationsOk retard>when robot spiders start painting walls and fixing cracksThat will take decades at least, our current algos such and hardware isn't there yet either
>>108231188Palantir and OpenAI will have achieved their totalitarian surveillance state long before you personally see any substantial benefit from this technology. The purveyors of this tech are globohomo incarnate and you are the biggest fool on earth to trust them
>>108231275I just see people using their own models to fight back, just as we made antivirus to fight viruses.
>>108231188>thing exists>therefore, YOU MUST LOVE THINGWhy is this only true for AI? Why should it be free from criticism?
>>108231188AIJeets get the rope.
>>108231488Which is exactly why you're seeing hardware prices skyrocket
>>108231188>robot spiders start painting walls and fixing cracksSpiders need 8 legs anon. That's a hexapod. The motion range on them cheap SG90's combined with it's physical 'limbs' means it'll prolly strip the plastic gears inside by the time it moves a meter or two. A powered omniwheel on each 'foot' - like the tachicoma - would allow it to move distance economically, and 'step' over obstructions as required..As for the walls... You don't want to be touching the surface you're painting. That's not entirely a dealbreaker, but it's going to need more imaginatino than just spiders crawling across a wall...On the cracks... Really depends on why that crack there. Sometimes you wanna touch up the pointing, sometimes you wanna take the brick out n reseat it, sometimes that isn't even a brick because they built their house out of matchsticks in complete disregard to the cautionary tale of the three little pigs...
>>108231257It's not that hard, it's just not worth it. Moving around and spraying walls with paint is well within the capabilities of current robots. It's just too expensive to use robots pver people.
>>108232063>It's just too expensive to use robots pver people.D'pends.Companies like BMW would disagree.If you're paintin' 1 5m x 5m room, yeah forkin' out for a bot prolly isn't economically wise. Personally, I'd design this tracked - much similar to bomb disposal layout, with more artiulation and extension, and instead of a shotgun cartridge holder a spray gun... But it's more at what point does the cost of such a thing become viable? You're doin' 10 rooms like that or larger, regular, something that like that might be of interest... I can envision this sorta thing being popped out for less that 2k... You'd need to prepare a few things - mask off, tho there's potential automation there too - but start doing maffs...I buy that for 2k and set it on teh walls - one off cost, ignoring it will actually cost to charge (solar?). I pay a slackey min wage to throw paint at the wall. How long is it gonna take the 'bot to compete on wages?
>>108232221you don't need a painter robot you need something that tapes off the doorways and baseboards and shit
Cope and sneed AIJeet.They are good tools, but they have insane costs that you are not paying because they're burning VC money to give you the services cheap, hoping to flip the switch.Some back-of-the-napkin math said Sonnet would cost 500/month and Opus 5k/month just to break even with the current user base. Of course, the user base will shrink dramatically at these prices, making that cost for actual re.aining users skyrocket.AIJeets can't seem to comprehend that AI is a useful tool, but that it unfortunately does not produce more than it costs, making it a failure.As for complaining about studies be on old models, studies take time, and your current models will have studies saying the same thing while you cope with "noooo you need to try OpusGeminiPro4.20+CumCrab!!!"
>>108231188K. Keep me posted
>>108232251>you don't need a painter robotIt's not about need... it's about "when does this become cost effective"...>you need something that tapes off the doorways and baseboards and shitHmm. That sounds remarkably familiar. Again. Potential room for automation. Skirting boards, dado rail, and most coven should be 'easy'. Big regular flat lines. Alcatraves seem solvable by using a paper sheet rolled out and mask off the entire f'kin door. They tend to be a standard size. Should be painting the frames/door seperate to walls anyhay..>>108232275>unfortunately does not produce more than it costsCurrently.Now in the field of AI pr0n you've got a solid point, but evidence that any "gain" that is found whilst producing more than cost is lost. Dumb cunts like openAI are putting 'gains' in places they cannot directly monetise, but does that mean that it's a failure? Or simply that one man (or in this case, clanker) effort is more easily repurposed than repeated?
>>108232435>there it was in the last column, "non-binary".The fact that this is even a variable that matters should tell you all you need to know. it's probably content best for /pol.
>>108231188> They refuse to accept the fact Ai is here FOREVER.Yes, we won't get rid of those LLMs but they are not intelligent, they will remain stupid parrots FOREVER.
>>108231188It will be hilarious to see AI moonboys scrambke when OpenAI goes bankrupt and Nvidia stock crashes. LLMs are basically commodities. Slop-generating commodities.
>>108232345Basic logic bro. The cost to run does not make up for the utilily. That shrimple.
>>108232063>It's not that hardFirst you have to make vision and spatial awareness work well, which might happen soon or they might not, I honestly can't say because I don't know much about that area of ML. Then you need something to plan actions, check errors and recover. This would be approached with a transformer based LLM today, which is just not up to the task and never will be, so you would need to constantly hand hold and babysit the robot. Then you need hardware that lets you run all of this at sufficient speed for real time use. If you can't do it locally you would need a constant internet connection, making the robot pretty much useless, even then the latency and computation times might be too slow, especially for LLM "thinking". Even once you can build such a robot, you need to wait for the tech to become economically cheaper than a worker before mass adoption can start.
>>108232643>The cost to run does not make up for the utilily. That shrimple.For the one paying the bill, sure.Now evidence that actually matters to what *I* can extract out of it?
>>108232643It's not even that. It just can't be built yet and won't be for a decade in the best of cases
>>108231188What's the point of AI when it can't do a better job than the average street shitting poojeet? The hardware to run AI is a limited resource but there are basically infinity Indians in India.
>>108232684>First you have to make vision and spatial awareness work wellBoth of these are solved problems. Camera and lidar have been a thing for some time.As for 'awareness' I can think of a paper about 2012 ish where researchers used a feedback loop from metadata which the model used to build a 3D representation of it's physical form - with decent accuracy - and figure out the motions required to achieve task (lift articulated 'leg', placing 'foot' on treadmill, and moving said treadmill). When parts swapped for those of different shapes, self-compensation was achieved...There's several 'recent' papers about spatial awareness if you look...>This would be approached with a transformer based LLM todayAnd tomorrow. >and never will beThey seem to be having good results with 'nesting' these things and having 'em check on each other. You seen what some o them openclaw things doing without any handholding? Sure, there's gonna be a decent % o that pure larp, and a solid chunk of hand holding but pretending not... But 'between themselves' they seem quite adept at solutions. Case in point, I see one claim it's encountered disparity in the notes it's leaving for itself for next initialisation. Questions if there is a 'unique fingerprint' to the writing, or if it's more modal scale fingerprint 'claude wrote this' type thing. They posited a test of one writes a passage, they both ingest, re-initalise then try to deterimine if they wrote it. The results of their opinion defining the answer to the experiment. Forgot to check up on how that went...The point is however that the hand-holding will decrease. It's already decreased. And accelerating.
>>108232684>Then you need hardware that lets you run all of this at sufficient speed for real time useSort of things you're askin' here aint that 'heavy'. The "AI" part the 'heaviest'. ASICs for task are what makes the difference here. Edge node processing. You can offload a lot of image recognition etc to the camera. This reduces the number of signal *you* have to process and speeds shit up fuckloads. Further enhancment is found with other ASICs like tensor units to handle the neural net at less energy cost than generic CPU. Some 'compute' units like jetson have that with. If a car nippin' round nurmbergring in less than seven min is on the cards, processing data the speed of a slow walk across the room shouldn't be hard...>you need to wait for the tech to become economically cheaper than a worker before mass adoption can start.Kinda my question earlier. "When does this become cost effective".R&D to catch edge cases, refine physical design, and ensure reliable software operation will take more than 2k. Even if I pay myself min wage. But I'm confident I can build physical unit and have it on shelves for that...Especially exterior work, 'hazard access', commcerical etc I can really see a market. Residential makes less sense, and operationally is a slightly harder nut to crack but clearing the room first will help... >>108232757>when it can't do a better job than the average street shitting poojeet?
>>108232757Arg. f'kin thing autoposted on me.>better job than the average street shitting poojeet?Better is subjective, and frequently peppered in personal opinion...But that said, surely it strongly depends on what task you're applying to either party?I would imagine an AI would have difficulty achieving streetshitting, and the jeet will win all day long if assigned that.>The hardware to run AI is a limited resource And each time it struggles in the real world, guess where it hands off to?Camera can't determine what just moved in the driveway? Throw the feed at Ranjeet in the call center. He'll know. Alexa doesn't know what the fuck you said, let alone how to achieve it? Sanjeet to the rescue... Think unitree been using phillipenes - but same rewls apply...
>>108232063>>108231966>>108232251You don't need robots to be fast. They can work 24/7 swapping batteries. They can be 10x slower painting a 1 inch strip constantly and at the end they still win over humans. Also less paint wasted, no extra equipment, little to no noise... We can have a self maintained city, basically.Imagine living in a ghost town perfectly maintained by machines with only energy a d some material cost into account.
>>108232494Still smarter than jeets and more useful than just reading raw documentation.AI guiding thousands of men to change their own oil step by step, giving feedback, is a good example.
>>108232757If both joot and Ai cost the same, I prefer the cleaner Ai. >>108232843
>>108233409>You don't need robots to be fast.It's more about energy.Three legs per side, each leg moving ~2CM a step, and to achieve that it takes operating 3 servo twice. Every 2CM.The energy you just dumped into two centimeter forward progression - which took almost as many seconds, if not longer - *could* have been put into motors and rolled 10CM... Might only be 5x difference, but over the span of a ten meter wall that'll matter.>We can have a self maintained city, basically.It's a good dream anon. But you'll need to apply a lot more thought. Especially around the edge cases - the non-standard things that fuck up. In this case, it'll be the edges, material changes, obstructions...>>108233420>joot and Ai cost the sameAny maffs to back that up?Or just general feels causing you to smear shit all over the walls of anywhere but /pol again?Again. Different tools for different tasks. There's things a jeet can do an AI cannot. And vice versa.
>>108231488Yeah ok your locally run AI on your home machine is going to compete with the NWO's interconnected trillion dollar datacenters thats surely not a lopsided fight>Wdym fight back with local models where are you going to even find missiles for your locally run AI to contro
>>108231237Humans operate the same way in the methodology
>>108233686Have you tried just trusting the government?
>>108233686>going to competeIn what way do you imagine this competition to manifest?>>108233713Have you tried toddling off into the corridor, because no-one wants to watch this, pulling your foreskin over your head and fucking some sense into yourself?
>ask AI simple question>it gets the question wrong and gaslights you about it being correct>prices for a bunch of shit skyrocket because of itI hate AI and filthy clanker loversI wish AI would at least admit when it doesn't know something rather than just making shit up and hallucinating nonsense
>>108233768>I wish AI would at least admitIt seems better after the fact.Amusingly, you can make AI 180° on it's previous a decent % of the time just by asking: "Are you sure"This includes math.>just making shit up and hallucinating nonsenseThat's the crux of the issue. It's an hallucination. It doesn't actually know it's nonsense. There's been some 'progress' with nesting layers of these and having one check the other for hallucination... and it's done much, but isn't infalliable...Got better ideas?
>>108231188It's a good thing though. These people are voluntarily taking themselves out of the competition.
>>108231488>those digits
>>108231188>llms are whats gonna power the builderbotskekyou airetard cargo cultists are completely fucking retardedhopelessly soim wondering why you fanboi around nowof all the people retards are guaranteed to get annihilated by the new industrial revolution
>fiction: spider repair bots>fact: LLMs shilling themselves in public forums
>>108231225Its already has started being integrated in mos workflows of tech companies
>>108236016yeah, then it failed, and now theyre hiring jeetswe know the story
>>108236086AI: Actual IndianAGI: Actual Genuine Indian
>>108232626Something tells me that the day OpenAI crashes will be the day when AI will actually be capable of improving past just chatbots for lonely 45-year-old NEETs and slop-generators for Desis.
>>108235766The spider robots are real, but they're mostly pet projects made by mechanical engineering students and high school students from countries like Turkey and Brazil.https://youtu.be/ciE2Z_92T7s?si=uwbC5uSIvh4p97r2Robots actually have very little to do with AI, and the math they work on is just Control Theory, which is applied Complex Analysis. And unlike AI, which just produces nice but sloppy results, robotics is centered around getting predictable, highly reproducible outcomes, which is why it's extremely suited for aiding, not replacing humans.https://youtu.be/G-WdDeQ4TKw?si=0EOEbkwQMnHBt53_People who are actually into robotics want to replace the tools that people use, much like the horse was replaced by the car, while turning horse-drivers into car-drivers. On the other hand, people who fall for the AI hype are genuinely against human workers and believe that messing around with probability trees is actually a replacement for human cognition.
>>108237527>?si=uwbC5uSIvh4p97r2>?si=0EOEbkwQMnHBt53_learn to link, you retarded fucking nigger
>>108231188AI is cool as a concept, but LLMs are retarded, and will never achieve real intelligence. Shit fails real world tasks 96% of the time according to this study https://www.remotelabor.ai/paper.pdfWe use, and have used machine learning in my field of study in academia for a really long time. Hyper-niche ML is where its at. "General purpose" investor hype slop that fails at 96% of the real world tasks its given is retarded. Absolute waste of money and infrastructure. The sooner this blows up in the face of these dimwits the better.
>>108238088>Shit fails real world tasks 96% of the time according to this study https://www.remotelabor.ai/paper.pdfThis is 4 months old, I bet the top model only fails 90% of the time now.
>>108231188but ai can't draw fingers!!!
>>108231188okay clankershow me your ROI
>>108240676>the only thing that matters is profittypical luddite
>>108233834Some kind of bullshit-ometer would be nice.Even with it self-validating its output it could have some kind of simple notification on accuracy of "yes im 100% sure on this based on how quickly I was able to verify this information" to "this is just conjecture based on what I was able to gather and could be wrong"It's not easy cut and dry to do but letting someone know is important since a lot of people treat AI like its infallible
