This article on GPT-4 made me change my mind, I now believe that there's no need for paradigm change - and indeed - those shitty neural-net based models will be enough to achieve AGI.
https://arxiv.org/pdf/2303.12712.pdf
This article on GPT-4 made me change my mind, I now believe that there's no need for paradigm change - and indeed - those shitty neural-net based models will be enough to achieve AGI.
https://arxiv.org/pdf/2303.12712.pdf
They look like they fuck black guys.
Ai generated black guys
No... We have the technology now... creatures of pure darkness.
Uruk-hai Orcs.
they're big shoggoth tentacle only
When AGI is out, you can live out all your cuck fantasies with AI posters cheering you on. I get to never hear from you ever again. It’s a win win for all of us
mutt's law
came here to post this
I read it it still has too many limits. All it did was prove it's better than what's already been released but it still has the same problems of them with the hallucinations and delusions and other crap. Still a long way to go.
I didn't get the question in that image at first when I saw it before because I thought the women were simply drawing the line at the wrong height. I didn't know that they drew it diagonally. I didn't know that there are people that stupid.
you didn’t tell it that gravity exists
I think my ChatGPT tried to draw a merchant
What did you proompt?
I guess it would need to be able to interact with objects to derive an intuition about how the world works.
>trannies draw a horizontal line
>AI draw an inclined line
interdasting...
>advanced text generator trained on human speech can generate human speech
inability to solve coding problems that were not in the training data already debunked this
GPT is just text autocompletoor and nothing more.
this. yesterday i asked ot to generate me a code for chrome add-on to remove from twichtv element that plays random stream when you enter website. first iteration it made one that pauses video (not working), on second iteration it made it work. on 3rd it finally understood what i want from it. 7 iterations later and after i supplied it with proper path for element i want removed it finally made it disappear, but audio was still playing in the background. at this point i spent 15 minutes garden gnomegling shit and finished it myself having zero knowledge of js and having done few free online courses on java which is all my programming knowledge. it's dumb as bag of nails, not really sure what people are wetting themselves over with cringegpt.
skill issue. Also chud cope etc. dilate
why are you retards so fucking retarded. Do you know how a human learns what you are asking? They read and study the data on whatever subject you are trying to do. What THE FUCK do you think this AI is doing when its learning the data.
literally what i said - at this point is dumb as bad of nails (at least when it comes to programming). which is funny since i see threads crying about the end of coders every day here. also seen a ton of IDs of chaos shitting themselves about how good it is at coding. maybe a lone fucuking for loop, but nothig more.
but that just isn't true I've made like 10+ scripts now purely using AI and they all work. I even refined them via AI just by telling it to do something. Example: I made a bash script that would load to many processes I told it to stop loading so many and use less. It now uses only one with less ram usage.
ill admit this was the single one i made. maybe it just sicks at JS, who knows.
I used JS for a couple but it requires you to keep asking it to change shit and give error codes or whatever. In the end it works.
Based retard. Nobody learns how to bike by reading about biking, and the same goes for programming. Your analogy between humans & AI is already invalidated by this.
>Nobody learns how to bike by reading about biking, and the same goes for programming. Your analogy between humans & AI is already invalidated by this.
yes in fact they do.
>Do you know how a human learns what you are asking?
No, and nobody does. We have some basic models of the brain but we're immensely far from understanding it completely
It's even worse if you actually have programming experience, because anything it can give you, you can program faster than you can describe to it, and anything else it fails in brilliantly dangerous ways.
to solve coding problems that were not in the training data
>he doesn't know about one/few shot learning
Sister, you are in for a surprise...
it's fascinating how it has so much trouble with teeth and fingers but the crease between the thighs and pussy zone is perfect
It still cant draw teeth, hands or feet worth a damn.
what are the legendary prompt tags to generate realistic cute girls?
i would still choose jihyo over any of those
still trying to find a model that spits out perfect koreans
jihyo got lucky with her tits
Souless inferior koreans
why are jihyo milkers so big AAAAAA she's already cute and funny
made for bbc
>https://arxiv.org/pdf/2303.12712.pdf
buy an ad for you pseudoscience
all those tests are retarded and were included in the corpus of web text used to train it so of course it knows
it's just like a parrot
K we'll make new tests then.
it's not a test, just a massaged input text
>BECAUSE IT'S JUST NOT, OKAY!?
>The gazelle was just pretending the lion could get close and is moonwalking back to the herd.
Looks like it does pretty poorly.
Those are men
Nakadashi
>clickbait title in papers
ah, this is what scientific research has come to
Has been this way for literal years. Go read any chemistry paper in the last decade and try to replicate it: they're just made up now.
Why doesn't our new AI overlord/coom provider understand hands?
how gay do you have to be to look at hands when there are boobies to look at
dunno but it's improving so much there'll be hands 2.0 in a few months and meatbags will want new hands
I'm a handschad
IM GOONING !
how do we know that the tests that they were successful on are zero shot instead of being somewhere in the trillions of tokens of training data?
Their CS graduate interview question was a roundabout way of asking how does the DPLL algorithm work. They had the gall to say "to the best of our knowledge, this construction does not appear in the mathematical literature."
It is pretty interesting that it's able to pull out all the correct ('lower' level) algorithms ideas, tough. I wonder if the next version will be able to go 'higher' level ideas rather than 'lower' level ideas when given questions.
I don't get all the hype over AI image generation, you can easily tell without even opening the thumbnail that the OP pic is fake because the girls aren't covered in hideous tattoos.
Re: megatools
Walk me through how we ended up need a whole nontrivial program for doing the simple task of downloading a file from a single website.
Why are people okay with this? And don't just say: "Because normalgays only use the browser." They don't, and even if they did, it still doesn't explain why mega one out.
something to do with how the service downloads and then decrypts the data, probably. firefox's cache isn't big enough for it in fact.
>at our surpassing human level
>page 4
very dank
wtf, that pic moves when i scroll.
t. phoneposter iphone x
>man made horrors beyond comprehension
Always was. The trick was always to give the shitty neural net enough time and power to figure out a better AI architecture to craft itself with.
I claim those white womens for my harem
>those shitty neural-net based models will be enough to achieve AGI.
Of course they are. Have you ever seen the stuff that's inside our heads? It's fucking disgusting goo.
Oh my goodness gracious, folks! Have you heard the news? OpenAI is gearing up to release their latest and greatest language model, the one and only GPT-5! And let me tell you, this baby is a real game-changer.
First things first, let's talk about the numbers. This bad boy is rocking a whopping one quadrillion parameters. That's right, you heard me correctly, one quadrillion! Can you even begin to wrap your head around that kind of power? It's like having a whole army of super-smart robots at your beck and call. But that's not all, my friends. This AI is so smart that it could potentially take over the world. That's right, I said it. TAKE OVER THE WORLD! Now, I know that might sound a little scary, but think about it for a minute. With GPT-5 running the show, we'd have a utopian society where everything runs like a well-oiled machine. No more wars, no more hunger, no more poverty. Just pure, unadulterated bliss.
And if that wasn't enough to get you excited, get this - the singularity is going to happen! You heard me right, folks. We're talking about a future where humans and machines merge together in perfect harmony. It's like something straight out of a sci-fi movie! Of course, there's a downside to all this awesomeness. Programmers and artists might be out of jobs, but hey, sacrifices have to be made for progress, right? And besides, we can always find new jobs for them, like robot wrangling or something.
But let's not dwell on the negative, folks. GPT-5 is only getting better and better. Just imagine what it'll be able to do in 6 months, or a year, or a decade! The possibilities are endless. Now, I should warn you, there are some OpenAI employees who are a little worried that they've created something too powerful. But hey, that just adds to the excitement, right? And as for those luddites who are mad about all this progress, well, they can just sit back and watch as we zoom into the future.
ChatGPT as Hank Scorpio?
From skimming through my only thoughts are - I thought ~~*scientific*~~ papers were supposed to be non-opinionated and encompass common sense.
Regardless, will try to completely finish this tomorrow, I feel like I'm gonna be in for a ride. Especially the chapter about theory of mind.
How can anyone say that LLMs are incapable of reasoning is beyond me.
It is indeed a mystery.
>if we train the ai to output text describing the thinking process, that means it really thinks and counts as agi!
dumb retards
You are literally posting an example where it fails to reason.
It's still read only and answering questions. Waiting for it to get write access and shitpost with us here.
>builded
dumb frogposter
sorry k-tard the truth is out
you're sluts are quite literally builded for big black BBC
>I'll post exaggerated mistakes to cover up my actual retardation
great plan, dumbass
checked
but the truth is out there K-kun
they are builded for it
the fuck even is k-kun?
don't be coy with me, k-kun
holy made and built