Sankakucomplex is also well-tagged but I wouldn't consider it a booru.
Actually using the site sucks because of the ads and you need a userscript to disable them. Honestly I just download stuff from it through Hydrus instead of browsing that shitty site.
>Orville Wright, who, along with his late brother Wilbur, pioneered aviation in America, has said that the men planning to make transatlantic flights were ‘absolutely foolish’.
Consciousness is not required for AGI. If an AI can replicate absolutely every behaviour we consider "intelligent" using mundane statistical methods its still AGI.
Not saying it is not impossible but airplanes were extremely complex back then but we figured out a way to do transatlantic flights. The rate in which technology improves is exponential. This year has been the best proof of this. There have massive breakthroughs in plenty of fields. The US successfully achieved net energy gain in nuclear fusion. It may seem impossible now but data is getting shared faster every year. Things are moving fast and might figure out things that we thought impossible before.
1 year ago
Anonymous
>The US successfully achieved net energy gain in nuclear fusion
No they fricking didn't, there's a reason they didn't include the (abysmal) efficiency of the laser in their calculations. Toruses are closer, and they've been "twenty years away" since 1950. >two more decades
1 year ago
Anonymous
eh, it's a win technically, let the fusion nerds have it, it's going to take a while >two more decades
I'd say 30 or even 40
it's something for our grandkids to work on
1 year ago
Anonymous
I see you missed the sarcasm in my distancing quotes.
1 year ago
Anonymous
>"twenty years away" since 1950
When you look at
1) how little funding has gone to it
2) how much political bullshit has obstructed it
it becomes obvious why fusion is always 20 years away.
1 year ago
Anonymous
>"twenty years away" since 1950
When you look at
1) how little funding has gone to it
2) how much political bullshit has obstructed it
it becomes obvious why fusion is always 20 years away.
There are lectures on youtube of profs talking about research into improved confinement and how they waited 10-15 years from coming up with theories to actually seeing those theories validated in small-scale experiments in the lab, due to funding constraints. These were ~$100k experiments, not $millions.
Further, just look at the history of ITER, it's a quagmire of political butfrickery and red tape. It likely could have been finished 15 years ago if it was just 1 or 2 countries building it, or if the governments involved just gave grants to an independent body, but that wasn't the point of the project.
1 year ago
Anonymous
>No they fricking didn't,
Yes, they fricking did. 2 megawatts of electricity went in, 3 came out. An estimated 230 kJ of energy hit the pellet, and 1.3 MJ left it. That's net, no matter whether you want to talk about how much electricity went into starting it or not. You can read the reports if you'd like.
It's easy to draw up a list of all the things people said were impossible that actually weren't, but you conveniently forget the even larger list of things people claimed were possible that actually weren't.
Literally ever snake oil salesman ever. Promises of things like "negative latency". Buzzwords, hype, etc. It happens all the time, every day. If you want to fall into the hype and claim that something is possible because other things have also been possible in the past and that's the outcome you'd like to see then feel free.
>Shit is moving so quickly
This tech, in it's exact form you are seeing now, has been around since 2016 at the latest, 2011 at the earliest. The only thing that has changed - literally the only thing - is that GPU hardware is that the algorithms have been optimized enough to do it with home-level hardware requirements instead of needing 256GB of VRAM. That is literally the only thing that has changed.
>What is that logic?
We're not ten years behind - what you're looking at is exactly where it has been for, at the longest, 11 years already. That logic is that you're already looking at the peak of what the tech can do. The accuracy of machine learning and various AI has already been above 99% for a long damn time as far as technology goes. The existing problems that were holding this sort of AI back - overfitting to certain characteristics in the data, vanishing gradients, exploding gradients, the need for pre-labeled data, the ability to have unsupervised learning - have largely already been solved, a long damn time ago. Literally the only thing stable diffusion did was achieve the same level of accuracy for drastically less resource cost. And we're not getting dramatically better than where we're at, for the simple fact that even if you could store the meaningful characteristics as a single byte of data, you're still looking at billions of learned characteristics to achieve this level of accuracy. It's always going to take gigabyte of VRAM to do these jobs. And since Moore's Law is dead, we can't expect the explosive gain in computing power that took 50 years to bring deep learning from when it was invented in the 1970's to useable today anytime soon.
This is it. This is the peak of this tech. It's not getting drastically better any time soon. You're talking another 50 years.
This tech is not dramatically improving. There's literally less than a single percent of accuracy left for it to gain. It can't make major improvements in hardware efficiency anymore, and there's no explosive leap-forward in tech on the horizon.
What's left is novel applications of existing AI, and that's about it. You might see some very, very, very weak AGI in the next 20 years.
large, well-tagged collections of images, with a somewhat similar theme = profit
>well-tagged
top lel
most of the time gay shit isn't tagged well so no matter how much I try to filter it out I always get some anal, yaoi, futa, bara material
our only hope is AI learns to tag them
but now it's a chicken and egg problem
perfect way for social media "influencers" to scam their low IQ audience with less effort >logan paul spends 5 seconds writing some proompts >AI shits out some nfts >"HEY GUYS TODAY I'M OFFICIALLY LAUNCHING THE PROJECT I'VE BEEN WORKING ON FOR THE LAST TWO YEARS! I'M SO EXCITED FOR THIS BECAUSE I WANTED TO GIVE BACK TO THIS AWESOME COMMUNITY WE'VE BUILT TOGETHER"
Who gives a shit about artists, I give a shit about what steam's front page is going to look like once every aspect of a game can be easily generated. Here I thought 10,000 games we already get was bad enough.
The real problem is the same as it's always been: automation always greatly reduces quality in exchange for greatly reducing the time/effort/resources required to create a similar (not same) thing.
So now you will never be able to find an indie game again, because only AI-generated unplayable crap (think marvel movies but in game form) will litter the first 100000 pages of any game store.
>So now you will never be able to find an indie game again
indie devs are arrogant whiny pricks that demand the world to worship them because they shat out a mediocre game. I hope they get replaced. gamedevs are next after the artpoor slaughter is complete
Just when you thought it couldn't get worse than soulless yearly AAA trash you get soulless yearly machine generated AAA trash. And people will buy it. There's probably no practical difference between a machine generated FIFA and the real thing except that the former is cheaper and thus makes more profit.
There's already a shitton of crossover between 2D and 3D art. Later's gonna be even easier for AI since it's governmed by math.
t.split between BOT and /3/
Artists are still safe because all we have to do is rise up and physically destroy the Internet with blast force. Just because the legal system refuses to protect our intellectual property rights doesn't mean you win, it just means you've forced us to take our rights into our own hands and protect them extralegally.
Not just that, we'll destroy the art too, and never make more of it. You proved you don't deserve us, so now you don't get to have us. We're going to delete all your existing training data with bombs and then deprive you of any more by simply not making it, so even when the computer and GPU economy recovers from the attacks and the ML ecosystem follows suit, there will be nothing for your AI to learn from.
IP is fricked. Enjoy being a pawn of the AI animated frozen corpse of Walt Gisney.
>implying extralegal self-defense of rights is bound by law
Nice reading comprehension moron
production level 3d modelling is a few orders of magnitude more difficult than concept-coomer image generation
it would unironically be easier to just get networks to produce videos instead
I saw an AI generated animation on rule34.xxx it's over.
it's funny because those boorus are probably the best way to train AI hentai creation.
large, well-tagged collections of images, with a somewhat similar theme = profit
The only booru that can be considered "well-tagged" is e621. Everything else, especially the popular ones like rule34, are a complete mess.
Sankakucomplex is also well-tagged but I wouldn't consider it a booru.
Actually using the site sucks because of the ads and you need a userscript to disable them. Honestly I just download stuff from it through Hydrus instead of browsing that shitty site.
post link
Sorry, I was asleep.
It was actually on rule34.paheal.net.
https://rule34.paheal.net/post/view/5389072
It isn't very good but still kinda impressive.
I like it
>artchuds seething in the comments
Tick tock...
>Tick tock
no it's paheal.net
lol the text
But this is really good it doesn't even randomly switch characters halfway. Artchuds on suicide watch
POST LINK Black person
LINK YOU homosexual
Is AGI this decade possible at this point? Shit is moving so quickly and it will get even faster
probably not
but you don't need it for regular ai to disrupt society into anarchy
AGI might be actually impossible.
>Orville Wright, who, along with his late brother Wilbur, pioneered aviation in America, has said that the men planning to make transatlantic flights were ‘absolutely foolish’.
No aspect of consiousness is understood. Also the brain is a million times more complex than an airplane.
Consciousness is not required for AGI. If an AI can replicate absolutely every behaviour we consider "intelligent" using mundane statistical methods its still AGI.
Not saying it is not impossible but airplanes were extremely complex back then but we figured out a way to do transatlantic flights. The rate in which technology improves is exponential. This year has been the best proof of this. There have massive breakthroughs in plenty of fields. The US successfully achieved net energy gain in nuclear fusion. It may seem impossible now but data is getting shared faster every year. Things are moving fast and might figure out things that we thought impossible before.
>The US successfully achieved net energy gain in nuclear fusion
No they fricking didn't, there's a reason they didn't include the (abysmal) efficiency of the laser in their calculations. Toruses are closer, and they've been "twenty years away" since 1950.
>two more decades
eh, it's a win technically, let the fusion nerds have it, it's going to take a while
>two more decades
I'd say 30 or even 40
it's something for our grandkids to work on
I see you missed the sarcasm in my distancing quotes.
>"twenty years away" since 1950
When you look at
1) how little funding has gone to it
2) how much political bullshit has obstructed it
it becomes obvious why fusion is always 20 years away.
There are lectures on youtube of profs talking about research into improved confinement and how they waited 10-15 years from coming up with theories to actually seeing those theories validated in small-scale experiments in the lab, due to funding constraints. These were ~$100k experiments, not $millions.
Further, just look at the history of ITER, it's a quagmire of political butfrickery and red tape. It likely could have been finished 15 years ago if it was just 1 or 2 countries building it, or if the governments involved just gave grants to an independent body, but that wasn't the point of the project.
>No they fricking didn't,
Yes, they fricking did. 2 megawatts of electricity went in, 3 came out. An estimated 230 kJ of energy hit the pellet, and 1.3 MJ left it. That's net, no matter whether you want to talk about how much electricity went into starting it or not. You can read the reports if you'd like.
The educated skeptic or the naive buffoon.
Taking bets?
It's easy to draw up a list of all the things people said were impossible that actually weren't, but you conveniently forget the even larger list of things people claimed were possible that actually weren't.
>larger list
can we see it?
Literally ever snake oil salesman ever. Promises of things like "negative latency". Buzzwords, hype, etc. It happens all the time, every day. If you want to fall into the hype and claim that something is possible because other things have also been possible in the past and that's the outcome you'd like to see then feel free.
>Shit is moving so quickly
This tech, in it's exact form you are seeing now, has been around since 2016 at the latest, 2011 at the earliest. The only thing that has changed - literally the only thing - is that GPU hardware is that the algorithms have been optimized enough to do it with home-level hardware requirements instead of needing 256GB of VRAM. That is literally the only thing that has changed.
Absolutely none of this is new.
What is that logic? Ok cool that means we’re 10 years behind and they can probably replace us in most jobs
>What is that logic?
We're not ten years behind - what you're looking at is exactly where it has been for, at the longest, 11 years already. That logic is that you're already looking at the peak of what the tech can do. The accuracy of machine learning and various AI has already been above 99% for a long damn time as far as technology goes. The existing problems that were holding this sort of AI back - overfitting to certain characteristics in the data, vanishing gradients, exploding gradients, the need for pre-labeled data, the ability to have unsupervised learning - have largely already been solved, a long damn time ago. Literally the only thing stable diffusion did was achieve the same level of accuracy for drastically less resource cost. And we're not getting dramatically better than where we're at, for the simple fact that even if you could store the meaningful characteristics as a single byte of data, you're still looking at billions of learned characteristics to achieve this level of accuracy. It's always going to take gigabyte of VRAM to do these jobs. And since Moore's Law is dead, we can't expect the explosive gain in computing power that took 50 years to bring deep learning from when it was invented in the 1970's to useable today anytime soon.
This is it. This is the peak of this tech. It's not getting drastically better any time soon. You're talking another 50 years.
This tech is not dramatically improving. There's literally less than a single percent of accuracy left for it to gain. It can't make major improvements in hardware efficiency anymore, and there's no explosive leap-forward in tech on the horizon.
What's left is novel applications of existing AI, and that's about it. You might see some very, very, very weak AGI in the next 20 years.
you are a fricking moronic the fact we can now run it on home systems means that it is ten years away you fricking idiot. have a nice day.
Nah AGI wouldn't possible
Carmack thinks it's 7 years away.
>well-tagged
top lel
most of the time gay shit isn't tagged well so no matter how much I try to filter it out I always get some anal, yaoi, futa, bara material
our only hope is AI learns to tag them
but now it's a chicken and egg problem
>-male/male
was it really that hard
only if it's tagged like that
same applies to solo_male
the problem is it's not always tagged
What
>text-to-nft
jfc
perfect way for social media "influencers" to scam their low IQ audience with less effort
>logan paul spends 5 seconds writing some proompts
>AI shits out some nfts
>"HEY GUYS TODAY I'M OFFICIALLY LAUNCHING THE PROJECT I'VE BEEN WORKING ON FOR THE LAST TWO YEARS! I'M SO EXCITED FOR THIS BECAUSE I WANTED TO GIVE BACK TO THIS AWESOME COMMUNITY WE'VE BUILT TOGETHER"
this is great, in 5 years the "shitty free versions" will be better than the best ones today, and everyone will be btfo
Oh shit, this is too close to home. How far off is this from ruining my comfy WFH inventor modeling job?
Who gives a shit about artists, I give a shit about what steam's front page is going to look like once every aspect of a game can be easily generated. Here I thought 10,000 games we already get was bad enough.
>I give a shit about what steam's front page is going to look like once every aspect of a game can be easily generated
AAAAAHHHHH!!!
who cares about buying if you can just ask an AI to make you games?
Sounds like more trouble than it's worth.
The real problem is the same as it's always been: automation always greatly reduces quality in exchange for greatly reducing the time/effort/resources required to create a similar (not same) thing.
So now you will never be able to find an indie game again, because only AI-generated unplayable crap (think marvel movies but in game form) will litter the first 100000 pages of any game store.
Not much will change though.
AI generated 3D with AI voice and AI story and so on vs. npc pixel art shovelware
You don't understand. It will be 10x WORSE than npc pixel art shovelware.
>So now you will never be able to find an indie game again
indie devs are arrogant whiny pricks that demand the world to worship them because they shat out a mediocre game. I hope they get replaced. gamedevs are next after the artpoor slaughter is complete
Just when you thought it couldn't get worse than soulless yearly AAA trash you get soulless yearly machine generated AAA trash. And people will buy it. There's probably no practical difference between a machine generated FIFA and the real thing except that the former is cheaper and thus makes more profit.
Finally lazy japanese homosexuals on seppuku watch
Looks like a bunch of styrofoam 3d model not even practical for commercial use.
Boring
Can I make my own models using AI in blender now? I wanna explore my creativity and scenes I made in my head when I was a kid
Holy frick that was fast.
This is what we've been saying to you over and over again. Things will move VERY quickly.
Yes, you guys predicted 10,034 out of 2 technological singularities.
Where did I say that?
There's already a shitton of crossover between 2D and 3D art. Later's gonna be even easier for AI since it's governmed by math.
t.split between BOT and /3/
Yeah its literally over for us software bros. Hardware is the future while software will be dominated by AI.
hardware is designed using software, that can be controlled by AI though...
Joke's on them, because 3d "art" on rule34 was always pure spam and garbage quality.
As an artist, it's gonna be cool to feed in your characters and watch them get turned into 3D models.
Any good ai for generating logo?
Does it just do artistic figurines or can it do engineering models to replace cadmonkeys too?
artists are safe because at least AI can't generate graphic novels (comics).
comics are just words and pictures
that mean a writer can make comics without artists
or even replace the writer with chatGPT
It's generating pointclouds, NOT 3D triangle meshes. It's still (hard) work to generate a triangle mesh from a pointcloud
Artists are still safe because all we have to do is rise up and physically destroy the Internet with blast force. Just because the legal system refuses to protect our intellectual property rights doesn't mean you win, it just means you've forced us to take our rights into our own hands and protect them extralegally.
>Digital artists will destroy the means to make digital art to protect their digital art
Not just that, we'll destroy the art too, and never make more of it. You proved you don't deserve us, so now you don't get to have us. We're going to delete all your existing training data with bombs and then deprive you of any more by simply not making it, so even when the computer and GPU economy recovers from the attacks and the ML ecosystem follows suit, there will be nothing for your AI to learn from.
>implying extralegal self-defense of rights is bound by law
Nice reading comprehension moron
IP is fricked. Enjoy being a pawn of the AI animated frozen corpse of Walt Gisney.
production level 3d modelling is a few orders of magnitude more difficult than concept-coomer image generation
it would unironically be easier to just get networks to produce videos instead