Anonymous at Fri, 31 Mar 2023 11:44:25 UTC No. 941679
>>941673
"clusters of trees"
"rolling hills"
"fallen leaves of decaying fall"
FUCK
OFF
CRIS
Anonymous at Fri, 31 Mar 2023 11:53:25 UTC No. 941680
really getting sick of ai vaporware. the next ai winter cannot get here soon enough. at least the vr winter and crypto winter is already here.
Anonymous at Fri, 31 Mar 2023 12:23:03 UTC No. 941686
>>941680
Yeah, unlike Crypto or VR, AI has actual use. Before the AI winter AI had 0 use, now it is basically the most useful invention in the past 20 years. Every company is rushing for investment and the main worry on the people working with it is not that it wont return its investment or that it will go into winter mode, but rather that it is moving too fast and it is freaking even the people profiting from it out.
Anonymous at Fri, 31 Mar 2023 12:26:39 UTC No. 941687
>>941686
dude, i have a crazy idea, what if .... what if we put ai on the blockchain?!
Anonymous at Fri, 31 Mar 2023 12:32:30 UTC No. 941688
>>941686
>now it is basically the most useful invention in the past 20 years
O Rly? Ai that steals your work and doesnt and cant attribute credit is the most useful invention? Get a life.
Anonymous at Fri, 31 Mar 2023 12:34:25 UTC No. 941689
>>941688
yeah. the same way airbnb made so much money by essentially operating illegal hotels or uber running an unlicensed taxi fleet, ai is basically a way to launder copyright infringement.
Anonymous at Fri, 31 Mar 2023 12:43:11 UTC No. 941690
>>941689
let me add that ai doesnt even operate on facts, it has no idea what the truth is
Anonymous at Fri, 31 Mar 2023 20:00:25 UTC No. 941732
>dumb ass cattle is falling for another scam
https://www.youtube.com/watch?v=wzb
Anonymous at Fri, 31 Mar 2023 20:52:08 UTC No. 941741
>>941732
They using a prebuilt scene, hiding assets and unhiding them was the first thing that came to my mind.
But that they were too inept to forget to recompute the shadows before doing makes it an open and shut case.
Anonymous at Fri, 31 Mar 2023 21:14:21 UTC No. 941747
>>941673
Why the fuck did you make 2 threads for this fake ass shit?
At most it's not doing anything more than placing existing assets in the scene based on tags. None of the words in the left side make any difference to the final output. It's just looking for shit like "house", "fence", "car", etc.
It's as "AI" as some Indian dragging and dropping trash can assets onto a pre-existing scene.
This >>941741 is the most likely.
>>941688
That's not how AI actually works, but keep believing that. The AI models don't contain any of the source material nor can it output anything that the source material actually contains.
Are the datasets "stolen"? Absolutely. But you're better off blaming the data collection services and the websites you put your art up on, as they're the ones selling those datasets off as their product. The final model and companies training those models have largely nothing to do with "stealing art", except from buying from a seller that turned out to be less than reputable. It's like buying something from a pawn shop that turned out to be stolen.
Of course, this is the part where you bitch and moan, call people names, and spout off a bunch of misinformed "facts" instead of having a discussion with the intent of hearing both sides.
Though you'll probably not even make it that far, as
>stopped reading on the first sentence
Is probably the most likely response.
Anonymous at Fri, 31 Mar 2023 22:10:57 UTC No. 941753
>>941747
>That's not how AI actually works, but keep believing that.
that's not how copyright law actually works, but keep believing that.
Anonymous at Sat, 1 Apr 2023 01:34:27 UTC No. 941767
>>941747
>The AI models don't contain any of the source material nor can it output anything that the source material actually contains
It contains it alright, in a highly compressed noise form that you can adjust based on seed and "denoise"
Anonymous at Sat, 1 Apr 2023 01:45:37 UTC No. 941769
>>941767
Dude, we already have language models that can perform logic problems and are threatening actual STEM wagies, as well as copywriters and marketers. There is no need to pretend that AI is just compression tool that remembers everything it is trained on 100% and does nothing but regurgitate it with some mixups.
Anonymous at Sat, 1 Apr 2023 02:20:35 UTC No. 941771
>>941753
Funny, I didn't mention anything about copyright law, and neither did you. But keep moving those goalposts :^)
>>941767
This is how I know you're just talking bullshit. You're so ignorant of the subject, you don't even know how much you don't know. No wonder you're scared of AI. It's beyond your comprehension, and you're unwilling to even learn proper and basic facts about how it works before passing misguided judgement.
🗑️ Anonymous at Sat, 1 Apr 2023 12:01:01 UTC No. 941799
>>941771
>The AI models don't contain any of the source material nor can it output anything that the source material actually contains.
you clearly haven't read the getty images vs. stability ai lawsuit.
Anonymous at Sat, 1 Apr 2023 12:28:20 UTC No. 941801
>>941771
Ive trained models in dreambooth, kiddo. When you're done training if you input the identifier and increase thr denoise strength you get back the original images.
You dont know what you are talking about, dumbass
Anonymous at Sat, 1 Apr 2023 13:03:47 UTC No. 941804
>>941747
>That's not how AI actually works, but keep believing that. The AI models don't contain any of the source material nor can it output anything that the source material actually contains.
iT cAnT oUtPuT aNyThInG tHe dAtAsEt aCtUalLy CoNtAiNs
Anonymous at Sat, 1 Apr 2023 13:36:20 UTC No. 941809
when will this fag finally get a lifetime ban for spamming the exact same garbage 10 times a week?
Anonymous at Sat, 1 Apr 2023 17:16:23 UTC No. 941825
>>941673
it's over, every single day, everyday.
Anonymous at Sat, 1 Apr 2023 20:54:51 UTC No. 941839
>>941811
>training set of one image
>generation looks like the one image it's trained on
That's still not the original image, that's the model's understanding of it. If your dataset is only a single image and you go to town training on it, it's going to understand that image and nothing else because that's the only context you've given it, especially if your caption is shit like in your example.
It's still not the original image. It certainly "looks" like it, but that's because once again, it's trained on that single image. It has a "deep" understanding of what that image contains, and is able to recall what it looks like.
It's as if you thought of the Mona Lisa in your head after staring at it for a few hours, you'd be able to recall quite a bit of it, but it doesn't mean the original painting is contained in your head.
Since it's a tool meant to do this exact thing, of course it's going to be better at it.
Same with >>941804 , the AI understands the fundamental elements of the painting to a large degree, so obviously it's able to output something extremely similar. That still doesn't mean that what it has output is original data, simply that it has a deeper understanding of the subject.
>>941801
>you input the identifier...
You haven't said anything that "exposes" anything.
When you're done training, you input the subject you trained it on (identifier) and turn up the samples (denoise strength) and the output looks like something you trained it on? Yeah, that's how training works. You're still not getting back "original" images though (as in the literal pixels you fed into it). It's the model's interpretation of your identifier based on its understanding of it. Dreambooth training is quite good at understanding single subjects, but once again, it's not outputting photos you've fed it like a fancy photobash.
At this point you're not even trying to argue in good faith. Then again, you weren't ever. It's all about your "agenda" because you don't like the thing.
Anonymous at Sat, 1 Apr 2023 22:10:12 UTC No. 941843
>>941840
What anon means is that the data that makes up the original image isn't present in the trained neural net
it is similar to how we have a memory encoded in our brains of something we've seen and are able to mentally reconstruct that something to some degree of accuracy.
Obv the machine is way better at this task than any human but that's something that's only gonna get more and more refined.
Problem is how this technology now exist and only way you can stop the algorithms from making anyone's style of art
,eventually better than the creator who came up with it themselves, is to never let it browse the internet.
Even if we panic and make laws to only let it train on ethically sourced public domain data that's not gonna hold this development back for very long.
It's just gonna give the edge to corpos like Disney etc who already owns a lot of high level art on which to train their ANN's.
Anonymous at Sat, 1 Apr 2023 22:21:41 UTC No. 941845
>>941843
thats not how its going to work in the future and also please look again at >>941811. This doesnt make it past the supreme court.
Anonymous at Sat, 1 Apr 2023 22:42:03 UTC No. 941847
>>941845
The algorithms are open source, the understanding of how to align and train these networks is getting cheaper and cheaper.
And art remains visible online. Put two and two together anon, anyone can assemble whatever dataset they want and train a network on it.
No matter what laws are put in place the capability to generate art will only become more and more accessible and more and more refined as we go forward.
What the supreme court says about it will be about as relevant as what the supreme court said about torrents or mp3.
Anonymous at Sat, 1 Apr 2023 22:45:57 UTC No. 941848
>>941847
not for long, its going to be cracked down on worldwide just like darkweb shit. Print this post.
Anonymous at Sat, 1 Apr 2023 22:50:18 UTC No. 941849
>>941843
No anon, it's a turbocharged JPEG, not a brain. See: NeuralVDB. I can't believe this is allegedly a 130 IQ board.
Anonymous at Sun, 2 Apr 2023 04:15:41 UTC No. 941870
>>941849
I'm thinking you're the one with a crusty JPEG in place of a brain if you really think that.
NeuralVDB is a different thing entirely. It's an AI-based compression algorithm. It takes a big thing, and makes it smaller. Image generation is completely different, it's even in the name.
Anonymous at Tue, 4 Apr 2023 08:15:48 UTC No. 942009
>>941849
If you're not impressed how a 4Gb file you can keep on a USB stick or your harddrive is able to generate just about any imagery you can imagine
on a desktop computer and dismiss it as 'turbo charged jaypeg' you're a very hard person to impress.
What it does, how it does it and how cheaply it does it is like sci-fi level tech, to me it's one of them reminders how we actually do live in the 21st century.
You'd go back in a DeLorean and hand your younger self a pair of DVD's with Stable Diffusion and that dude would've thought
you handed them alien tech when they get home and discover what it does.
Anonymous at Tue, 4 Apr 2023 10:34:24 UTC No. 942018
>>942009
>a 4Gb file you can keep on a USB stick or your harddrive is able to generate just about any imagery you can imagine
You are absolutely delusional
Anonymous at Tue, 4 Apr 2023 10:53:14 UTC No. 942020
>>941843
>sell hand drawn ferrari logos because "the original data is not there in the drawing, it's just a really detailed memory"
>get sued
>HOW COULD THIS HAPPEN TO ME
Anonymous at Tue, 4 Apr 2023 11:53:58 UTC No. 942022
>>942009
where can i get two dvds with stable diffusion on it?
Anonymous at Tue, 4 Apr 2023 12:10:13 UTC No. 942026
>>942022
The main files are available from hugginface official site (makers of stable diffusion), you need to make an account there.
You'll also need a github account, githubs windows client (assuming you're on windows) as well as python for windows.
The main file 'sd-v1-4.ckpt' which is the version most people use as it's the most un-curated and therefore interesting
is indeed just 4GB but uncompressed it's more like 10.
All the above are free but require some knowhow/research to install.
Go on youtube and type "how to install stable diffusion locally" there will be several guides available how to go about it.
Installed locally there are no limitations in what you can prompt for and it has a lot more tools and capabilities than the free online one.
🗑️ Anonymous at Tue, 4 Apr 2023 12:24:32 UTC No. 942029
>>942026
sounds like you watched some bozo influencer on youtube and didn't actually do it yourself. imagine thinking using github and python requires some "know how". the absolute state of people on this board.
Anonymous at Tue, 4 Apr 2023 12:41:43 UTC No. 942031
>>942029
point is it's not a 1 click install.
Perhaps such knowledge pops into your head out of nowhere or you use your penis as a radio-transmitter to grab it over wifi.
That the case good 4U anon, most people will need an instruction to follow.
Anonymous at Tue, 4 Apr 2023 13:15:21 UTC No. 942033
>>941673
>tfw AI just killed VN fags for good
What a time to be alive
Anonymous at Tue, 4 Apr 2023 18:22:35 UTC No. 942054
have you ever noticed how gendered internet hype waves can be? like crypto was almost entirely a bro thing. have you ever met a chick irl who was into crypto? or online for that matter? now think about ai. the only people i've seen use chatgpt in real life are chicks and online ai shills always sound like hysterical women, so it makes me wonder why the gender differences? what is it about ai that blows women's minds but crypto and blockchain leaves them cold? puff a little crypto pipe dream into the air and a hustlebro's adrenaline surges while women roll their eyes. use chatgpt to skim wikipedia and women swoon and hyperventilate like they just saw frank sinatra in his prime...
Anonymous at Tue, 4 Apr 2023 18:45:20 UTC No. 942058
>>942054
I'm a man who bench more than his body weight. I'm hyped about chatGPT and AI, crypto-bros I'd throw into the sun if I had the power. Our discord channel, all bros, one sis, all bros uses chat GPT. Every AI channel I've come across on the interwebs - they're all beardos.
Sounds to me like the algorithm have sent you down somewhere strange.
Anonymous at Tue, 4 Apr 2023 18:50:15 UTC No. 942060
>>942058
>Our discord channel
is that where you coordinate your shilling? are you getting paid directly by someone or you just hope to go viral and get some ad views?
Anonymous at Tue, 4 Apr 2023 19:04:52 UTC No. 942061
>>942060
>is that where you coordinate your shilling? are you getting paid directly by someone or you just hope to go viral and get some ad views?
No. We're not using AI as any sort of 'get-rich quick scheme' because fuck you and fuck money.
We're using it to enhance our work, research topics, joke around etc, you know enjoy life.
People with a brain is hyped about AI because of how massively it's already enhanced our lives.
Nobody has to 'shill' it. It sells itself as unless you're blind you can observe what it enables users to do.
ChatGPT and AI is 'web 3.0' for real. It's not BS that has to be forced down your throat like the metaverse or a some giant megascam like crypto.
Anonymous at Tue, 4 Apr 2023 21:50:30 UTC No. 942077
>>942061
>It sells itself as unless you're blind you can observe what it enables users to do.
>ChatGPT and AI is 'web 3.0' for real. It's not BS that has to be forced down your throat like the metaverse or a some giant megascam like crypto.
Anonymous at Wed, 5 Apr 2023 18:06:23 UTC No. 942157
>>942061
>People with a brain is hyped about AI because of how massively it's already enhanced our lives.
can you give us 1-2 precise examples of how is it massively enhancing our lives?
Anonymous at Wed, 5 Apr 2023 18:37:19 UTC No. 942161
>>941801
>if I train it wrong it gives me the images back!
Thats not the point retard. A model like SD haves billions of images in a 4GB model, thats an impossible level of compression so only relevant features are stored, the rest was lost in the noise and cannot recover the original images
Anonymous at Tue, 11 Apr 2023 19:26:41 UTC No. 942705
>>941688
sorry rando faggots and public domain pictures didnt get credit for AI porn. you faggots pirate and hoard resources without a care but need a reason to justify your dislike for ai. Just say it scares you and thats it, itll be easier than retarded excuses you spew.
chinks and social media has been blatantly stealing others art for years, imposters etc. no one cares.
Anonymous at Tue, 11 Apr 2023 22:00:38 UTC No. 942727
>>941801
your shit is overtrained, retard.
Anonymous at Fri, 14 Apr 2023 20:11:26 UTC No. 942996
>>941673
it's fake u retard in one of the videos posted the AO for the buildings was already placed before the dude prompted building to be made in the first place it's so fucking dumb stop posting this shit
Anonymous at Sat, 15 Apr 2023 18:05:39 UTC No. 943096
Normies don't understand (or more likely, don't care) how weighted networks aren't AI, so all they're going to do is hype the shit out of it, and promote "you'll never have to hire again" scams to idiot businessmen while they can. Third worlders like cris can't make things even that sloppy, so for them, it's an opportunity to get ahead of other useless third worlders.
It's not much of a niche, and it'll disappear as people get used to it, just like people got used to search engines and online maps. For them, it's a massive change. For those of us with actual skill, it doesn't really change anything. There are some machine learning brushes and tools in the works, but there's no
>I'm never going to have to work again
>I can skip studying!
and no obligation to adopt it to not be left behind.