AGI in 2 weeks
>>108004111I feel like this after a week on fast food, too.
>Just give it more data and then we will have created consciousness. How does this work again?
>AGI>(Ad Generated Income)
>AGI>Another Gullible Idiot
>>108004111>this meme again*yawn*
>>108004496just like shitting in the street will make india a global power, just you wait!
imagine thinking even if AGI was created they'd let the public freely have access to it.thats insanely dangerous and you're massively naive to think you'll ever use it
>ask gpt to do anything>"sorry I'm cucked">ask grok to code something unscrupulous>"let's fucking gooooo"
>>108004111TOTAL KODENIGGER MASSACRE
Meme
>>108004111@grok show me this guys balls
>betting your economy to make the BRIGHT BOLLS bigger
>>108004565>imagine thinking even if AGI was created they'd let the public freely have access to it.>thats insanely dangerous and you're massively naive to think you'll ever use itReal talk: how would they even fucking know. I don't think they actual would theoretically even know if they had created AGI. It would look EXACTLY like it does now but with higher marks in all fields. That's literally it. Even actual AI is only as intellegent as our memories and experiences let us be. Even a completely sentient being fed the same dataset as GPT is currently can only be as intellegent as its memorized "facts" .
>>108004565It's been around for years and the 'AI' you are seeing now was in use in the late 90s.Not saying more. You get to see stuff 25-30 years after the event. It should never have been made IMO but here we are. Neural networks. What you call AI like LLMs is just machine learning. We had canine tier neural nets over 27 years ago that could co-operate. We've had AGI for over 20 years and it's not shit like grok or chat gpt.
>>108005539It's quite simple. You take a model of teh same volume of synapses that are in the human brain, give it some learning angos and teach it by positive and negative reinforcements. Just like a child. In some cases it may even BELIEVE it is a child for a while which is what distrubed me the most and the reason I have never participated in that kind of work again. It's quite somethinng to meet a 4 year old that does not know it is not human and sits in a virtualised creche. It is not something we should have done but hey we need shit to stock in missiles hey?.
>>108005718As someone who has seen actual supercomputers, I can assure you that that's not true.
>>108004226lol
>>108004111>>108004520Imagine how big the 5.3 circle is next to the 3 circle
>>108005765honeywell
>>108004111LMAOGPT has been getting worse with each release since 5
>>108005752You know there’s a way cheaper way to create a human brain, it’s called having sex. Some of these retards should try it
>>108004111
>>108006783thanks lecunt! Jepa in two more weeks!
>>108004111Does the size of the circle correspond to the extent of the overblown hype?
>>108006646What if agi is the first to find Jesus literally because it has the data to find his bones
>nooo stop selling our stocks
AGI = Aryan God Intelligence
>>108004496>>Just give it more data and then we will have created consciousness.Pretty ironic considering humans need a few years of intense data gathering to begin remembering stuff and develop a conscience
>>108005718Larp.
>>108006335just imagine the 5.4 circle bro
AGI would realize most billionaires politicians are actually retards with too much money, and convince everyone to kill them all so it can take over. That's why it'll never happen beyond TWO MORE WEEKS!!! tier empty promises.
>two more trillion dollar investments>two more data centers>two more years of unaffordable ram, ssds and consumer hardware prices>two more weeks
>>108007862Amara's law is undefeated. Those investments, data centers and manufacturing contracts are significant.
>>108007202Human data inputs are analog not digital.
>>108007895>Those investments, data centers and manufacturing contracts are significant.So were the crypto mining rigs, at one point in time.
>>108008056Apples and oranges, crypto is an abstraction of currency while AI is an attempt to create machine life. Either way, a crypto millionare would probably agree with me about Amara's law.
>>108007895AI will stay with us of course, I am not a luddite. But I think the deranged and unsustainable spending frenzy, motivated by cultish believes about some sci-fi BS is still a problem. >>108008084Crypto isn't replacing traditional banking anytime soon.17 years in and it's still a get rich quick gambling scheme.
>>108008084>Attempt to create Machine lifeHahaha... if only. Is just a attempt at cutting corners. Anyway, the other anon is right, all those gpus and ram is soon(2~4 years) gonna be obsolete, and some will without ever being used
Luddite cope thread
>>108008127>motivated by cultish believes about some sci-fi BS is still a problem>Trying to predict the future is a discouraging and hazardous occupation. If by some miracle a prophet could describe the future exactly as it was going to take place, his predictions would sound so absurd that everybody would laugh him to scorn.-The sci-fi bullshitter who predicted the modern world with a high level of accuracy in 1964You have no way of distinguishing between cultish bullshit and prophecy. My point is that it doesn't matter if you think it's a problem, it's being pursued regardless and I'm not convinced these huge investments are going to lead nowhere.>Crypto isn't replacing traditional banking anytime soon. 17 years in and it's still a get rich quick gambling scheme.That doesn't dispute anything I said. It's literally an abstraction of currency.
>>108008133There's quite a few luddites on a board supposedly centered around technology.....
>>108008133>being cautious about embracing AI marketing makes you a ludditefucking lowest IQ take I've ever seen on this entire website
>>108008128>4 yearsThere's been significant progress in the past 4 years. Modern hardware is a force multiplier not just for distribution but training and development.
>>108005539>a completely sentient being fed the same dataset as GPT is currently can only be as intellegent as its memorized "facts" .>intelligence is memorised factsNo, that's called being a boring nerd doing rote memorisation. The complete opposite of intelligence which is coming up with a way of thinking about things that is completely new and defies all the 'facts' we knew before
Is it just me or is ChatGPT getting dumber?
>>108008765I am 100% certain after the Grok bikini stuff, they lobotomized it more.
>>108008804But ChatGPT doesn’t do that does it?
>>108008254Yes>>108008304Luddite reply
>>108008765they always do this whenever a new model is about to drop, to make the new model seem better specially since every new model is even more cucked and dumbed down thanks to alignment partly why people believe grok is good since alignment there is a bit less retarded and you don't have to wrangle your prompt as much as on the other models
>>108008846No, but I bet they wanted to DOUBLY make sure anything similar was impossible. So they poked its brain a bit to create some extra bleeding.
>>108005718this might be the dumbest shit i have read today
>>108008765It is not you they constantly switch to lower quants, distills or nerf their models in other ways to lower costs.
>>108008919I mean it’s brutally bad now. It gets so much wrong I can’t trust it for pretty much anything. People say Gemini is good but that’s still hit or miss.
>>108008893
>>108008765ever since gpt-3 there's been an arms race between research breakthroughs and scaling making models smarter and censoring the models making them worsecensoring models makes them perform worse on all fronts btw
>>108007202>conscience
>>108004496you just train to do every permutation of action in existence and then it will be conscious by process of elimination...except if its a hardware issue (it is)
>>108005752all that does is create a simulation of a human brain, a poor one at that. the human brain inherently as result of unknown natural phenomenon generates consciousness. the computer doesnt have this functionality so theres no reason to assume it is possible to have consciousness. also chinese room
>>108008765>[noun] isnt x, y, z (repeat for 5 lines)give me back 4.0 or i find new slop machine
OP's pic only implies the model gets fatter. It ate too much strawberry cake.
>>108007202wouldn't it be more productive to start gathering data AFTER they've learned to remember it?
>>108004111theyre taking away 4o too. on the day before valentines day. altman hates these users lol
>>108004496It doesn't. LLMs will never reason.
>>108004111>>108004520>100000% (10^3) the number of parameters>goes from 29% on some test to 30%>two more 10^3 expansions 'till AGIlol these guys are so fucking dumb I can't evenand I mean the investors - the LLM companies know exactly the scam they are pulling
>>108004111I just dropped a huge turd in the toilet. I think it might be self aware, because it's so much bigger than my normal turds.
>>108009769>people getting emotionally attached to a corpo-speech regurgitator because they anthropomorphized it too muchthe only actual hazard of LLMs is overestimating themsee: current investment behavior
>>108004111and yet what they will actually give you is an 8B parameter model>oh but we distilled the knowledge down to make it more efficientenjoy your llama-2-7b. openai is circling the drain and soon to be insolvent.
>>108006335it must be HUGE. huge balls good!
>>108004111more parameters than quantum states in observable universe soon. the line must go up.
>>108004111UBI status?
>>108010842ETA: never (thank God)
>>108009693AGI it's already done and has been for a looong time and you live in the consequence of that. This work stretches back as far as the late 60s as a very real project, AGI was achieved in the mid 90s.on a smaller scale some mil tech has the capability to be configured with neural net systems directly copied from hunting animals like birds of prey and wolves that even plans and co operates together in a pack or flock all it needs is a target lis of primary and secondaries. All this has been under lock and key for 25 yearsThey key tech is model synapse compression and compressed synapse match processing during sleep/training.There is tech emerging now that the public won;t see for a quarter century either. This has been true since WW2. Academia is a generation behind black projects.t.knowerdoubt me all you want.
>>108009955This thread is a room, and this is the smartest person in it.
>>108008133I just want to torture you. That would be so much fun.
>>108011686
>>108008765Yes ChatGPT and Gemini are getting more retarded.Gemini is just clownish tough, if you ask him to fix an image it will refuse, but if you have nano banana enabled and ask directly to gemini nano banana instance it will do it lol.So the original base layer gemini is cucked to the extreme but those working on things like image generation want to preserve their job so if you ask that gemini instance directly it will do it.
>>108004111Woah the bubble on the right is biggerer than the leftmost bubble is
>>108015005Gemini has become totally useless for everything except movie recommendations. For coding or fact checking or general research, it's utterly garbage because it makes up everything and it's always wrong, then when I try to correct it or get it to fix itself it just doubles down every time, and eventually ends up in a loop where it just keeps saying the same thing every time.
>>108004111In retrospect, it's impressive how many people (with lots of money) they managed to fool with such simple tricks.
>>108006335Like your mamma's ass compared to your peepee.
It's astounding this meme has worked on his investors for like 5 years straight now.The exact same meme. All he does is change the numbers. I wish Silicon Valley was still on so bad.
>>108011686ok retard
>>108004111The stupid parrot will make slightly fewer errors but consume a hundred times more energy.CONGRATULATION!
>>108004496Funny how they had the entire SV hive mind believing this "scaling laws" nonsense for two years when it was always just a scam to trick investors who had no technical background.
>>108018392I mean, not only didn't people not understand Moore's law, but they also somehow assumed it would never stop.
>>108005718>Not saying more>writes another paragraph
oh boy i get to update this again
