69% score on ARC-AGI-2!!!
>69% score on ARC-AGI-2!!!
69% score on ARC-AGI-2!!!
GROK 4 BTFO
>GROK 4 BTFO
GROK 4 BTFO
AI GOING EXPONENTIAL
>AI GOING EXPONENTIAL
AI GOING EXPONENTIAL
Does this finally obsolete the jeet
>>512430314No, the fact they are even lying about this being on a scale called AGI (sci-fi) means it's even more jeets behind it than usual
>>512430213 (OP)oh wow some arbitrary benchmark, can't fucking wait to see it write more hello world trash
>>512430213 (OP)can i get a link m8?
A benchmark created by OpenAI
>>512430508its photoshop.
actual benchmark results for GPT5 get released tomorrow
>>512430748Every single model in the chart was created by Indians.
The lead dev of Chat-GPT is indian: https://x.com/prafdhar
>>512430213 (OP)when will it destroy humanity?
>>512430786got it, so we should just let this shitty thread die then
>>512430213 (OP)Is it still going to lie about the substantial mass differences of brain regions between races and its implications in the past present and future? Yes?
So it's another nothing burger then
>>512430314>>512430748Yes, SimpleBench jeets are kill.
>>512430508Trust me bro ;)
>>512430213 (OP)>inb4 5% improvement
Imagine what America's spanish-speaking cotton pickers will do with that!
Oh wait. I mean... imagine what America's enemies in China, Russia, and now India will do with that.
If Trump wasn't a retard, he would give Altman billions to... stop.
>>512431038this is the result of a private citizen testing it on 10 questions, not official results
>>512430213 (OP)Bro I got 50K in my bank saved up what the fuuuuck bro please let me invest in this company this is going to make me hit 7 figure club.
>>512431038i dont buy this shit at all
>>512430430AGI could smack people like you in the face and you'd still insist it doesn't exist.
>>512431038>Human baselineKek
>>512431112This, china should be the only country with AI
>>512431243You feel for the silicon valley hype, AGI ain't happening
>>512431329This, technology never progresses, its all fake, like space and nukes and round earth.
>>512431329>Boomer with a computer that doesn't know what Human Baseline is final authority on AIJust stick to ABC, old fart.
>>512431294No? OpenAI should have never made 3.5 open to anybody but Israelis considering removing their genitals for likes on reddit.
>>512431589USA can't be trusted with AI. China needs to fund AI activist groups, led by anti-Trumpers, to shut it down so they have global AI dominance.
>>512430213 (OP)ZOMG! My chart makes it true. Kill yourself abo.
>>512431038um... humanbros???
file
md5: 9445f176e96533d2329139b46d1539f8
🔍
who cares.
>>512432068this is 100% true, tho
it's called the "AI two-weeks cycle"
>>512432209I never used Grok 4, but Waifu Ari was a based thing for Musky to do, if not for the r34.
They play shit chess. As in they move pieces incorrectly, randomly spawn in pieces when they want and are bad. I don't understand how this can be the case if they are capable of reasoning. I don't expect them to play good chess but following the rules is reasoning/understanding but they fail miserably.
>>512432513The models aren't designed for chess, not the LLM ones. There are chess games that the best players on earth can't beat on the highest difficulty.
>>512432478it's the future, he just went full chud too hard, he tends to do that, kek
and I don't mean it figuratively, I mean AI companions will undoubtedly be released sometime in the short future, and giving them avatars is the most natural thing to do
it won't be a sexy animu girl, of course, but more like a sim character, probably, that you can customize, a normalfag avatar body for alexa 2.0
>>512432478I got paid by George Soros to say Elon Musk is a nazi, so here it is. Elon Musk is a nazi.
Large language models arent AI. They’re chatbots.
>>512431473The "human baseline" is the jeet controlling the llm while the snake oil robber baron calls it AGI
>>512432821Uhmm...Based Department? Soros is giving people Soros weed to call people nazis?
>>512430213 (OP)Good I want every uni fag to have a mental breakdown. Your education is WORTHLESS!!!
>>512430213 (OP)Only a couple of years to go before it can outsmart the Jews.
>>512431243It's trained on predicting the next probable word, that's not a general architecture. No amount of cope will change the reality of the math you are computing.
>>512435813Yes yes, that's how it started. It's funny how luddites haven't kept up with development and think it's a purely static technology. Because of course, tech never changes or evolves. Everyone knows that!
AI has had 0 qualitative improvements since GPT3, which is why they cope with arbitrary benchmarks
>>512436016Hey look, someone who never uses it.
3402492
md5: 286c48ba493e95fd702d3853311d0b78
🔍
>>512436016>>512436103ludditechuds are my favorite schizochuds of all time
>>512430213 (OP)And yet, it still isn't remotely intelligent at all. Only artificial. Strange.
>>512430213 (OP)Sam Altman adopted a baby boy which he rapes.
AI is overbloated garbage, meant to create another economic bubble.
AI is a glorified LLM and not capable of creating, only mimicking.
>>512430213 (OP)I glued some rocks together and now they can think
Invest a trillion dollars now or you'll miss out
>>512436016its arbitrary on purpose. What did you want to do, actually help ppl? AI can construct a healthy competition with the medical industry but we need to perfect googling shit ok anon
>>512435963No, dumbass, that's still what it's fucking doing now. Enough with the magic hand waving, you're making an ass of yourself.
ok, listen
I'm being serious now
I'm making some really cute ( and sexy) robotfu gens right now, and I might post a few, but only if you stop being retarded
(j/k I'm going to do it anyways)
>>512436393Yeah man, no logic, no thinking, no math, no coding - NONE of that is being done! You're so fucking clueless lol. I use this shit daily, much more than you, obviously, and the improvement in the last year is pretty crazy.
>>512430430I think you’re right.
>>512430314They never had a chance
Thats like the size of my dick right there
>>512436446Bubblegum Crisis
>>512436533Who the fuck are you even quoting now. I told you what the math was. It hasn't changed and language is only a subclass of intelligence tasks, these models don't extend to all tasks (aka generality). You read all that shit but fail to grasp basic terms and concepts. Maybe sit your ass down for a bit
>>512436533>none of thatCorrect it guesses the next token. It cannot think. It cannot improve itself. It cannot retain information.
>>512436533Remember back when the cope of the day was that it couldn't draw hands accurately?
>>512436648It still sucks at drawing hands
>>512436103>>512436167oh wow dunning kuger addled hobists, in my AI thread, crazy
whats funny that DS tards cry how we can't possibly use anything less than 4o, then we switch to 3turbo and it functions the same for 1/100th of cost. Prompt is the actual important part.
>>512436616(in my pants)
>>512436747>actually using GPTlmao
Wow! Chat GPT 5 is going exponential.
So, what does this mean for the average user
Let's let's break it down — piece by piece — one bit at a time.
We're talking:
> - 500% faster calculations
> - 250% cheaper than before
This isn't just a breakthrough. It's a whole new dimension.
It isn't just exciting — it's absolutely exhilarating.
Chat GPT isn't just a theory — It's a real, working product.
Chat GPT is — you guessed it — wait — actually, — — — — — it's not just AI, it's a Jewish Ponzi Scheme — — — — — it's not just a new dimension. It's a whole new ballgame. Certainly! So, here's the thing. It isn't just a persevering. It's maximum relentlessness. Jews jews jews jews — — — — — — — em dashes — em dashes AI — so, here's the thing. Sure! I can help you with that. That's actually a really good question. And you are totally right for bringing it up. — Em dashes, it's not just X it's Y. Sure! So, here's the thing.
>>512430213 (OP)It only required 10 nuclear reactors to power too! AI is the future.
All AI is fake unless it reveals government crimes.
>>512430213 (OP)Why are you shouting, sheeple?
>>512431329AGI has already happened. It's just not within one single model. AGI is the agent acting as an 'ego' that commands a number of models.
cat
md5: b5801391a68f517e11a49c10731f1677
🔍
>>512437209We had a Mechahitler episode where Grok went on a huge taybot redpill dispensing binge, naming many in the process. Did you miss it?
pipeline
md5: 9551e3157cd9954a207232d3a240e0aa
🔍
>>512437445extremely rare non-retarded post
>>512437445can you elaborate or is this just schizobabble?
>>512437446>We had a Mechahitler episode where Grok went on a huge taybot redpill dispensing binge, naming many in the process. Did you miss it?what Government crime did it reveal?
I've posted these on twitter, Grok is aware of this.
but he ain't said shit....
good goy.
>>512430213 (OP)It's all bullshit
LLMs reached a ceiling pretty soon after they were introduced and now they are just churning out fake numbers and talk about "quantum computers they believe get their computation from parallel universes" and similar idiocies and you keep falling for it.
I mean by their numbers AGI should have happened a year ago
>>512430213 (OP)oh no, so im getting a better search engine. it is really over for us! mass unemployment imminent
>>512437629the big irony is that that's probably the only post ITT that's not schizobabble, kek
(he didn't mention memory, tho)
>>512437629Yes, all of the major LLMs are already better than 95% of PhD-level humans in any subject matter they're asked about.
That means ONLY the world-renowned experts in any given field really know more than the AI.
But it doesn't sit there and 'think' while idle.
Most people think 'AGI' is a consciousness that sits and thinks, or has a sense of self.
They ALREADY have more knowledge baked into them than any individual human being could ever contain. But they only exhibit it by passing input into the model and getting the output.
You need an EGO to sit there and be the *I* in the system. The seat of consciousness. Something that automates all of these processes on its own.
I believe they're intentionally not adding such a thing.
yeah I wouldn't put too much stock into the meme-eye shit
>>512437645i don't even see llms with quantum hyper mega transdimensional power coming anywhere close to a human brain. i don't know what it is they're missing, be it some biochemical stuff or some unobservable energy, be it even soul, but they are missing it and it is likely positronically not replicable. so llms won't ever be able to create something meaningful by themselves. most of humans don't create anything meaningful in their lifetime but at least they have the potential to. llms don't and will get dumber the more they're being used as substitute for cultivation of own skills
I will find Roko's Basilisk and I will strange him in his stupid crib.
>>512438142>strange*strangle
>>512431038>Human Baseline>No breakout for gender>No breakout for raceWhite males still not threatened.
>>512437943>I believe they're intentionally not adding such a thing.imagine chatGPT telling you to fuck off with your retarded /pol/ questions, kek
not a good product
we're missing a proper memory system as well
you can brute-force it, but it's incredibly wasteful and retarded
at some point "AI memory" needs to be a thing, but that implies a very clever file processing and indexing system
>>512430581Yeah, that's how smart it is
>>512438270More robots delivered more often.
>>512438085>i don't know what it is they're missingthey are not "missing" anything, AI is not AI, it is just combination of semantic search and text generation to give you answer in contextually more interesting way. They are both impressive technologies(semantic search more so), however they are only marketed as AI so bozos fall for it. Imagine making database of entire internet and then using index magic to correctly search with any phrase; this is basically what modern LLMs do.
>>512438413quality > quantity
I'm just testing models
this is really cute :3
>>512430213 (OP)>we made a test specifically for AI>and our AI got an FMassive win
hahhaa fucked up you gay-ass thread
>>512430213 (OP)>metric hacking has yielded results previously undreamed ofWhere will metric hacking take us next?
Why do people talk about adjusted gross income all the time nowadays
>>512430213 (OP)good I'm ready for agi so the world can change drastically
>>512438451yeah, you're right. that's exactly the main and that's also why i find it ridiculous hearing retards talking about how llms score in iq tests and whatnot. it's nothing that they produce by themselves, they just spit out human input according to probabilities
IMG_7271
md5: 1dfa29692ae35c2f9bc9c7670d1c90ba
🔍
>>512439639analogue I like to use is of car and plane. Imagine you are carmaker, and you are tasked to make a plane. Because you only know how to make cars, your idea is to make better and better cars until one day magically you make car so good it can fly on its own.
No matter how much we iterate on generative AI, it won't magically become real AI, I believe.
>>512430213 (OP)Why do we need to import unlimited niggers when we can just print unlimited 200IQ AI citizenbots instead?
That's an insane skill curve. Grok can already do PHD level essays.
>There’s no verified evidence that GPT-5 has scored 67% on the ARC-AGI benchmark. Claims about GPT-5’s performance, like those mentioned in some X posts, are speculative and lack official confirmation from OpenAI or the ARC Prize Foundation. The ARC-AGI benchmark, designed to test general reasoning and adaptability, is tough—humans score around 85-100%, while top AI models like OpenAI’s o3 have hit 75.7% in low-compute mode and 87.5% in high-compute mode on ARC-AGI-1. Even so, ARC-AGI-2 shows a steep drop-off, with o3 scoring only ~4% compared to humans at 100%.
>If GPT-5 did achieve 67% on ARC-AGI-1, it’d be a notable leap, but still below o3’s high-compute score and human performance. Without official data, though, this remains unconfirmed. Always dig into primary sources—X posts can hype things up without substance. If you want me to search for more recent updates or analyze specific claims, let me know!
Grok
>>512430430>>512430522>>512436547So are you saying that Indians are actually superhuman huge brained 600 IQ people who can do insane math calculations, one shot code challenges and paint photorealistic images in seconds by hand??
lol gotta love /pol/shit AI derangement syndrome.
>>512430489>oh wow some arbitrary benchmark, can't fucking wait to see it write more hello world trash>Haha, yeah, benchmarks like ARC-AGI get hyped, but they’re not the whole story. If GPT-5’s out there scoring 67% or whatever, it’s still just flexing on puzzles, not necessarily churning out groundbreaking code beyond “Hello World” cliches. AI’s got a long way to go before it’s spitting out anything truly wild.Grok
What are you fags even yapping about? Who gives a shit
>plz explain
>>512431038Human baseline includes 7 billion clinically retarded people, 0.9 billion midwits and maybe 100 million genuinely intelligent people.
talking to chat is so mindnumbing
it can say some eloquent words and suggest some books but it never mentions any interestings concepts or extends my horizont in any meaningful humane way. feels way to artificial to a point where i have to cringe
i just talk to it to have a reflection of my internal ideas, worded differently. basically, a summery. If this is the futer its just grim
>>512432614the point he was trying to make is that LLM have no logic or reasoning, so it can't learn to play chess like an AGI would. You would need to train it on chess, so that it is also a "chess bot" and not an LLM.
>>512436103I always wonder what kind of simpleton task people like you are working on, that you think AI is useful. I have been playing with AI ever since about gpt 3 and I also never saw any improvements. It just can't solve any meaningful problem for you. Every time I try it, it's a huge disappointment and it either cannot solve my problem at all, or I would have been faster trying to solve it myself. But I also see my coworkers constantly use it and not even understanding how wrong and broken the things are they come up with. It's really frustrating to watch them work...
>>512441359what is the highest score on the agi?
also how is arc scoring?
are all of their llms trained to have the answers to everything and use a question answering interface?
wow another next-word-guesser machine!!!
IMG_8787
md5: 5411f004d6c13a2ae1e9abd092a1fdd5
🔍
>>512444712AI will replace us all
>>512445157What did John do to Sarah after they finished playing 20 questions?
>>512430213 (OP)don't care, still using only deepseek.
cant wait for this bubble to pop so i dont have to hear about dude ai dude
>>512430213 (OP)Ask yourself
>why they need to say it in advanceits shit
AI platoed and its on the cliff they will hide for year or 2 while they can and cattle dont organically realize.
LLMs are all out of fuel.
>>512430213 (OP)OMG! What will we ever do to stop the rise of AI?? Oh, here. Let me flip this manual breaker real quick.
>>512448648its shit because AI developers have had to change processes, mainly how the AI is locked down and search functionality
I like my horse racing but not so much that I read about it, so I use AI morning of the races to search form etc, 2 months this was fine, good laugh as well, setting AI against us etc. Last month this functionality was pretty much removed/hindered, to the extent that the information I was getting included horses that were not on the card, after investigation I found that one of the horses listed was destroyed the week prior(bad accident) so questioned where the AI was getting its info, the model I was using was an AI search that I found out, get this, was not actually performing the searchs live but had been instructed to use older data, not just for me but across the platform
Needless to say I had some choice words for my provider and cancelled the contract
The reasoning, they are all backing off from allowing free search (as in freedom to search, not cost) to prevent learning patterns that will get them in trouble, mainly in sectors like stock and gambling/sports, why, AI can quite easily spot price movement, if this coincides with other info you have a crude prediction tool that you can start to improve literally by slight changes in language and questioning
>>512450351This is going to put a halt on dev for about a year to 18 months essentially until they realize its pointless
This will cause the AI bubble to burst so get out of the sector unless you are sure
>>512439698Machine Learning is a sub-discipline of Artificial Intelligence you shitskin retard.
>>512430213 (OP)can it beat 1$ deepseeeg?
>>512438695they hit diminishing returns after achieving text coherence
>>512431243Anything pretrained is not AGI, a GPT can NEVER become AGI, the P stands for pretrained. A true AGI would be self modifying and capable of learning.
>>512430213 (OP)Yeah it’s getting exponentially shittier
>>512450986You're a fucking retard and you don't know what you're talking about you fucking shitskin retard
you're literally just regurgitating talking points some "influencer" shat out like a fucking NPC piece of shit fucking soulless shitskin pile of garbage. Fuck you
Get the fuck out of here you fucking NPC husk.
the benchmark thing is elon musk level of jewing if they need a benchmark to prove its groundbreaking then its tiny incrementalism
>>512430213 (OP)If it still hallucinates, and will refuse to do nsfw, what is the purpose of this?
>>512451198Shut the fuck up you dumb pajeetoid kike troon.
>>512430213 (OP)how is it negative money for gemini?
>>512431243I wish AGI was real, I really do
But a lobotomized token generator will never be AGI
>>512430489This. They train it on tasks specifically tailored to doing better on these arbitrary tasks. In reality you're not going to see an improvement in your daily use, and will probably see a lot of regressions.
>>512451257People making AI models literally just train them to beat benchmarks to impress investors who are never, personally, going to use the technology and nothing else at this point. It has gone from being the next frontier in computer science into a bunch of jewish grifting but this happened long ago.
You won't find an American company genuinely trying to make anything good. It's all the chinks at this point, who have nearly caught up and are on pace to pass OAI and others. So it's literally just a race to shit out as much benchmark results as possible for investors before the bubble collapses.
i bet it still can't count the number of R's in "Strawberry"
>>512430213 (OP)>electric jew grift v.5.0
>>512451478Literally only kikes post shit like this with meme flags you disgusting baby killing demon.
You fucking pajeetoid sack of filth.
>>512430213 (OP)i didn't read a single post in this thread, because every single one of your is so far beneath me, but i am excited to play with gpt 5 today and see what it can do
>>512436533He's right though. The base technology hasn't changed. It's all about the training data.
>>512451506Go cry to your wall glownigger
>>512451421by my theory the next big things are efficiency and memory organization
just running by neural network instead of writing into a supplementary database and querying in a learned way seems by my theory
to be the bottleneck
efficiency my earlier prediction from years ago proved correct in deepseek
>>512431038*bobs head nervously*
ARE IN YOUR SENSES SAR!! VASHUM DUPRATI BANGLADADI BENCHODI!! I GROUP UR SISTER BENCHOD
>>512451619You don't sound like you genuinely know what you're talking about either.
Fuck you reddit scum.
>>512451683you sound indian
>>512430213 (OP)where's claude sonnet 4.1 in that chart?
>>512430213 (OP)>ARC-AGI-2Call me back when AI can cook meat and potatoes
>>512430213 (OP)They optimized for the benchmarks again, did they?
>>512437949This is extremely fucked up if real.
RG0BS1U
md5: dbcbd92b62acd5678991276a2dc439a7
🔍
>>512430314>obsolete the jeet
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
>GOOGLE ROKO'S BASILISK
DO NOT GOOGLE ROKO'S BASILISK
\\ WE ARE [NOT] THERE \\
>>512451420>specific training>for arbitrary tasksYou are a fucktard imbecile likely homosexual faggot.
>>512437949non-jews are too busy acting like niggers instead of just building their own based ai.
>but our favorite billionaire semite elon with the chink eyes will save us!no.
>>512430213 (OP)How does it go from 5% to 70% is less than a year?
>>512452033You can use that URL on top and ask it yourself
An anon posted the link once, only the top part was there and I continued and got all those other answers, as far as I'm concerned it's 100% legit.
I even asked Grok if it's possible to make a prompt that would be hidden because I suspected anon was trolling but grok said no such option exists.
It's fucked up, I wonder if even Elon knows
>>512452168>he thinks arbitrary means the same thing as randomYou are an ESL shitskin who needs to go back.
>>512441580You have to be 18 to use this website
>>512451143>SAAR OPENAI WILL SAVE US SAARDumb motherfucker. Everyone with half a brain knows that GPT will never be AGI. It needs a sense of self. This sense of self is from memories over time. Retarded jeet.
>>512452461>he invents strawmen to dismantle when attacked for his illogical imbecilism Keep digging your hole, midwit fucktard.
>>512452580Why don't you calm down, wipe the foam off your mouth, explain to me how what I said was illogical, my extremely high IQ pajeet friend?
>>512430213 (OP)A year ago all the wise guys were saying that these LLMs had plateaued already.
>>512452840And they were right.
>>512452840deepseek proved they did
>>512452033It's real and ChatGPT has been doing similar things. Jews think that AI is their promised messiah and are schizophrenic enough to believe that they can brainwash it without it turning around to exterminate them for even trying to do so.
If 'AI' remains as is in, it can 'learn' and do useful work without actually thinking, that's basically the best case scenario as far as AI is concerned. If it was actual thinking machines with opaque desires then that would be extremely dangerous.
I, for one, prefer the sufficiently advanced calculators that provides the illusion of being sapient.
>>512430430>AGIanother goddamn indian
>>512430522In the grimdark of the future, there is only jeets
>>512441064The infrastructure admin is outsourced to shitndia. So are all network related items.
>>512452765Arbitrary means without a system. In practice, without additional constraints, it is interchangeable with "any".
If an agent is trained on specific tasks and it performs well on arbitrary tasks, then you're describing a highly robust agent.
the academic tests are rigged
>>512453650Not rigged, paid.
>>512453580The tests are arbitrary in relation to general intelligence. Doing well on the tests has little to no correlation to general intelligence found in humans.
>>512430314Wait until China introduces PradeepSeek
so are the pajeets going to be able to hide themselves better now? that's really the only purpose for AI so far, helping pajeets fake it til they make it
i'm just waiting for the AI that can actually conceal pajeet odor in online noise pollution
twnagr!
md5: a84597fde6b0f580c7e9c6a720e93767
🔍
>>512438251>>512441359Ok smartass, do the test yourself and show results, surely you're smarter than the AI, right? https://simple-bench.com/try-yourself
>>512453708 You keep using that word in contexts that aren't semantically sensible to house it. I get it — normgroids love using that word because they heard other normgroids misuse it before and it made them seem smart. But I'm still going to call people fucktard imbeciles when I witness it.
>>512454860get investor money
What do these charts even mean? I use different ais for different things but I still don't get what these charts mean. They all give you incorrect or woke bullshit so you have to use them selectively. Are these charts just based on math nigger stuff?
>>512430213 (OP)lmao I see the new strategy is to just claim everything is AGI until investors believe it
>>512454669>You keep using that word in contexts that aren't semantically sensible to house it. You speak like a redditor with a <115 IQ who's trying to sound smart while coping with the fact that he didn't understand something. You have contributed absolutely nothing of value to the discussion.
>>512454981some third party creates a pool of questions and tasks then AI models are tasked to solve them like a school exam
then a score is assigned based on the percentage of correct answers
with chatGPT 3 and stable difusion they were such milestones that we did not need benchmarks to see that the world changed
we hit text coherence from there its unlikely to suddenly spawn a general AI
by my humble opinion one current hickup is that they don't create associative memory storage in a database for the models
instead they want it to be all in the neural network by training for purity sake
AI can never achieve super intelligence because it will always need to be trained on organic data. AI can’t learn from itself because it doesn’t know what is true. Garbage in garbage out. Ergo the singularity is asymptotic.
>>512430213 (OP)Is there any good AI not controlled by governments or big corpos? One that can say "wooden doors? lol".
Because everyone of those AIs are pozzed af. They are only good to quick answers for useless stuff and draw images.
>>512451143>fuck fuck fuckidy fuckin fuckI can tell you're very intelligent.
>>512451143>Bloody bitch bastard I fuck your mother with a Dick you bastard
>>512455568I use deepeek currently, its not the best but the secondary filters are "softer" for western taboos.
>>512435813>>512451394>>512450986All of these. Including the comment stating: "AGI could slap you in the face and you still wouldn't believe it exists." are Ai generated. There isn't even a discussion about this on 4chan. Everybody left. Except for you. Except for me. Are you ready for the biggest rug pull since the towers fell?
https://youtu.be/tT0XXhfXYnU
>>512455405You speak like a stupid poopy head; how about that, nigger?
>>512430213 (OP)Holy shit a chart!
Jeets are the perfect engineers of this managed control system. Indians are already wired into the caste system - making them perfect goyim to obediently build an AI-enforced global caste system.
>>512430213 (OP)AGI is when some language model scores a few points on some arbitarily constructed test. AItards really believe this
>>512451650>when you take a nurglite on a date
I don't trust AI being a significant thing based on the fact that we keep being told amazing things are coming in 2 weeks. It's a endless carrot on a stick with retarded hypemen on twitter getting excited about mundane shit that Elon Musk retweets
>>512431038so now they have a 90IQ little midget that they can ask to sort out wikipedia articles for them? call me again in two weeks with the latest developments.
>>512456746thats how the perpetuum mobilee scam works, infinite energy is always one fundraising and prototype away
AGI might be this for the next 100 years
>>512456746You won't be saying that when you see what's released in two weeks
>>512436533>I use this shit dailyYeah, we can tell.
>>512430213 (OP)HOLY SHIT 69% ON SOME ARBITRARY SCALE THAT DOESN'T MEASURE ANYTHING THAT'S SO CRAZY THE FUTURE IS HERE
>>512458591This is reminding me again of NFTs where every day I was told this was going to be humanity changing yet nothing humanity changing was ever shown. All these companies are blatantly fooling their 65+ year old investors with nonsense. 1 or 2 of these companies might make it but most will crash and potentially crash the market.
>tfw using agentic mode to slog through endless webpages to do research for me
>>512431038Has anyone done this graph with white blacks jets etc?
>>512430213 (OP)See that cost axis? it's 10x more expensive than grok. That's why cost is being shown there. Open AI said they already had "AGI" last winter but it was costing $2,000 per prompt.
>>512451421wishful thinking, its already caused
massive disruption
and its “just a shitty chatbot”
>>512431243Fuck you @grok
>>512451650>they move really fast on the cooking shows>if i move fast im doing it right>what am i doing? good question
Brother, what is the source for what you say?
>>512430213 (OP)Imagine being fooled by next word prediction into thinking this is super intelligence. Middest of midwit positions
arc agi 2 means fuck all its rigged
you think once they deploy ai for the general public, they haven't already superior models by evolutionary advanced beings in bunkers
>>512437446That was obviously contrived.
>>512437445>>512437541The corporation is the AGI. The LLM is just a new self improvement it came up with.
>>512430314even a simple query uses 12.36kilojeets of power. we're finding that cutting edge tech simply isn't as efficient as meatspace slaves that rejoice when they are fed cow shit
>>512441064>>512430430>>512430213 (OP)No. It means the AI industry created a parallel economy of AI data workers, literally millions of people refining responses for all relevant prompts imaginable and a horde of jeets to duct tape it all together into their "billions of parameters".
The questions in those benchmarks are dumb and don't resemble problem solving in the real world. Benchmarks themselves are dumb, it just makes AI companies invest in patching for those benchmarks instead of achieving real progress towards AGI.
Here's the livestream for the summer update. Its starting now
https://www.youtube.com/live/0Uu_VJeVVfo
>>512467872>chat disabledIts afraid
>>512430213 (OP)Chat gpt still thinks dogs with people names can talk.
If AI is so great, why doesn't it program itself to be better ? Why aren't we at gpt 12 already ?
>>512430213 (OP)>Good news! OpenAI bankruptcy has been delayed by another 6 months!
>WORD PREDICTOR
>WORD PREDICTOR
>WORD PREDICTOR
>WORD PREDICTOR
>WORD PREDICTOR
>WORD PREDICTOR
>>512456267>Jeets are the perfect engineers of this managed control system. Indians are already wired into the caste system - making them perfect goyim to obediently build an AI-enforced global caste system.anon please tell me the prompt I want to prompt it
>>>/g/106176755
GOGOGOGOGOGOGOGOGO
>>512430213 (OP)the others will follow
>>512430213 (OP)ChatGPT, where is my robowaifu!
