"Generative AI" is built on theft
ฝัง
- เผยแพร่เมื่อ 12 ก.ย. 2024
- Stephanie Sterling's video on having her work stolen by AI companies: • AI Stole My Content (T...
Nerdsync's video on having his work stolen by AI companies: • AI companies stole my ...
SAMTIME's video on having his work stolen by Apple: • Apple Stole My Videos ...
Actor Joe Gaudet on having his voice stolen and recreated through AI by a client: www.tiktok.com...
✔ SUPPORT ✔
Patreon: / councilofgeeks
TH-cam Membership: / @councilofgeeks
Paypal tip jar: PayPal.me/coun...
Ko-fi: ko-fi.com/vera...
Amazon Wishlist: www.amazon.com...
✔ OTHER CHANNELS ✔
Break Room of Geeks / @breakroomofgeeks
Vera Wylde: / verawylde
✔ SHOP ✔
Merch: www.teepublic....
My Book on Gender Fluidity: a.co/d/atfibBA (Amazon Associate link, commissions earned)
My Fantasy Novel: amzn.to/2SCxB8j (Amason Associate link, commissions earned)
✔ SOCIAL MEDIA ✔
Twitter: / councilofgeeks
Facebook: / councilofgeeks
TikTok: / verawylde1
Instagram: www.instagram....
Twitch: / councilofgeeks
✔ OTHER PROJECTS ✔
Council of Geeks Podcast (home of What the Frell & Jumpgate): councilofgeeks...
✔ CONTACT ✔
E-mail: councilofgeeks@gmail.com
Mail:
Council of Geeks
PO Box 4429
St. Johnsbury, VT 05819
My mum can translate three languages into English and has several degrees just so she could become a translator for EU law, specifically agricultural law. It's highly specialised, and her work is being taken by AI. Thanks to Brexit, she has to go through an agency for work now, and they use something called Trados, an AI translation software that my mum now contractually must use for work. This software takes everything my mum does to train its AI, and she has no say except to retire from the career she spent decades building, or carry on while they pay her almost half of what she was being paid 15 years ago. She is having her work stolen and she's practically had to agree to it to even be allowed to continue in her profession. So AI being absolutely shite is very close to home for me, I can pretty confidently say I empathise with everyone who is being stolen from.
As a professional writer, this was something that scared me a lot. I've been putting work into this field - both creatively and professionally - for years. This is what I'm good at. This is my best marketable skill. And I have absolutely no idea what happens to me if AI takes over.
Thankfully, I think a lot people have realized it's simply not a viable alternative - due to cost and quality of work - to just paying someone like me and I suspect ChatGPT is gonna be just straight up dead before too much longer. But I understand the fear of that. And I see similar issues in using AI for translations. How can it pick up context? Meaning? Idiom? Metaphor? How can it render those things in a way that makes sense to two radically different cultures rather than just translating a phrase from one language to another in a very literal way? Translators make intentional choices just as much as writers do.
I hope it gets out of your mom's life as people realize that it is incapable of miracles.
given how terrible AI is at nuance in language (because it doesn't really understand shit) I would say there will be a period where they desperately try to convince themselves it's working and get mowed down by competitors who just went back to using people with a brain. They're going to offend the FUCK out of all kinds of clients and such before they learn that lesson, I bet. Probably someone will lose a giant sack of cash before they figure out this is a dumb idea.
@@angelainamarie9656 I really hope so! But the state of it at the moment is just horrendous, and it's all the big agencies that use this system, and they're the ones being given all the work 🙃 From what I know a lot of the translators who have these specialist skills are just leaving the profession at the moment, they get paid more working at a supermarket, so agencies/ companies may have to do an awful lot of grovelling to get them back!
So sorry to read that about your mother.... 😢
@@HeidiSholl I trained my supervisor's supervisor because he was a quota hire in the 80s. He never did the job, well maybe a 5th of the job he was hired to do and we took up the slack and when we lost workers those workers' output was replaced by us taking up the slack again. I was working so as to attend college and get a degree. I would have been less resenting this situation if he had wanted to do the job and get promoted up to a better paying position. It would not have made the situation fair or just, but I could have had empathy about how he should have his opportunity to work and show off his abilities and skulls and get paid fairly for it. Please note, not all of those 80s era quota hires were bad workers or lacking skills needed. The good and decent minority workers hired in this era that I met would often seemed to hate the quotas, because it made it harder for them to get hired.
"James Somerton was a person. If nothing else, he had that going for him."
Damning him by faint praise.
I am reminded of a quote from _Going Postal_ by Terry Pratchett, “Steal five dollars and you're a common thief. Steal thousands and you're either the government or a hero.” I think you’re right that pointing out hearing about theft from individual creators hits differently than hearing about the massive scale of theft being done by the “generative” AI companies. People have all kinds of cognitive tricks to justify theft in the abstract. It gets harder to justify when someone points to a person you know and like and says, “this person is a victim of the kind of theft that you think is okay.”
One of my students the other day tried to claim that an assignment wasn't plagiarism, indeed could not have been plagiarism, BECAUSE they'd used Chat GPT and therefore Turnitin must be wrong about the assignment being ripped from Wikipedia. My eyes have never rolled so hard.
@@II00I00 Plagiarism checkers, whether AI enabled or not, are an imperfect technology but a necessary evil within the system as it now exists. It is not possible to start a corrective conversation about academic honesty without "proof" that wrongdoing has occurred, as that "proof" is the only legal pathway to having the conversation at all. Accusing a student of misconduct based on feelings alone is not allowed, we must present objective evidence of wrongdoing or a case cannot be opened. But teaching students about plagiarism and academic dishonesty is literally our job, we can't just not do it. So the bots are here to stay, on both sides of the podium. It is not an ideal solution, but there is no ideal solution. Teachers must still endeavor to teach. Students must still endeavor to learn. There is always eventually some cliff to reach where technological ancillae cannot help you any further and the ancient arts of human interaction come into play.
The amount of ick I’ve had when I heard other students talking about getting AI to do homework for them. No, I didn’t know their names so there wasn’t anything I could do about it. Yes, I would have if I had known. Just ICK. So, so many levels of ICK.
@@II00I00 weirdo
I’ve had students use AI to complete assignments. They don’t know the content of their "own" work. They cannot summarize it. They cannot quote it. They cannot discuss it.
“What AI is is James Somerton on an industrial scale” on a shirt please
One important technical note - you can't actually train a usable AI on only your own artwork. No matter how many paintings or drawings you've made, simply by being one human, you're not going to have sufficient volume to train an AI from scratch.
You need millions of pieces of art and photos to train a generative AI that does more then create random patterns.
What people are doing when they train an AI on a particular style is that they're taking one of the incredibly expensive massive data models as a base (and these are ALL built on an amount of theft truly unrivaled in human history) and then having them specialize based on a subset of data, like the works of a particular artist. The base model is still there, it's just been tweaked to create a specific kind of output.
An analogy is like - the base model is a computer speaker, and the specialized model is deciding that you're going to use that computer speaker to play exclusively recordings of flute concertos.
Without the base models, the specialized models don't work, and all the base models are built off of data acquired without permission.
You're right although there do exist models that are trained purely on CC0 images that you can train your own style LoRA for if you want to go the purely ethical route.
The fact that my FOUNDATION ART COURSE at UNIVERSITY was ENCOURAGING AI USE in projects “oh not to submit pieces but just for research and inspiration guys trust me” was deeply troubling
Yeah that's disturbing
i'm on an art discord and the mods there had to patiently explain to someone why posting a genAI pic of a "bird" in one of the anatomy reference channels was not appropriate. this person did not seem to understand that the genAI pic is useless as reference because it literally doesn't have anatomy or even coherently display how light works. super worrying that a proper art course is rolling over for genAI usage like this and fostering really bad habits in students while a freaking discord server knows better.
Unfortunately that's not surprising. A I companies have a lot of money and that is how they spend it- by bribing educators into shoveling this crap onto their students.
I remember back in the mid-2000s attending a class that was supposed to teach you how to do bookkeeping- Only to discover when I got there that the class was actually how to use Quickbooks- a software program whose purpose is to do bookkeeping for you WITHOUT you knowing how to do it.
At the end of the class I didn't know how to do bookkeeping and I barely knew how to do Quickbooks either.
As someone who got their art degree in the last class at my uni where even using computers *at all* was optional for all art students, this disgusts me. No, I am not an elitist snob, I am a flesh and blood artist who has worked at my craft with my hands and various physical media (and no "undo" key!) for close to half a century now. AI is the opposite of art. It is the death of art. Art is a living thing, made by living beings. AI "art" is, as a great poet once said, simply sound and fury, signifying nothing.
This is increasingly common. I don't know where you study but I know that Gobelins, one of the most well known animation schools has AI classes now (I don't remember in what way exactly, I think it's a separate diploma but idk), the Annecy international animation festival also presented AI short movies and when criticized for it defended themselves by basically saying that AI is just new technology like digital art was (🤢). It's terrifying.
"Generative" AI should be called "Regurgitative" AI. That's _much_ more accurate about what it actually does.
Agreed
I'm an artist. So the fact that work is stolen makes me feel angry and depressed.
But the thing that really horrifies me, and makes me feel bleak about everything, is the environmental impact. It's so bad. I can't believe it isn't a huge deal.
'Environmental impact' lol.
Using Twitter, TH-cam or a bunch of other services that have literally millions of servers running 24/7 didn't matter, buying your phone that was built by pretty much slave labour that also produces tens of thousands of tons of waste a YEAR didn't matter but NOW you care about environmental impact?
It feels so good making money off artists knowing you're always trying to take the moral high ground while being complete hypocrites that are nothing but ignorant.
I feel like a lot of defences of generative AI get the verb tense wrong. "It can be used in such-and-such non theft way" as though this is a theoretical examination of what might happen in the future. It misses the fact that the theft already took place. This kind of generative AI has been, and continues to be, theft.
And every word you typed could be considered theft by the same definition. I could find nearly all of them in Steven Kings book “IT”. Many in the same "cunning order" to paraphrase Douglass Addams.
That's what is being stolen, the order of words, the arrangement of pixels, the cadence or tone of a voice. The derivation of derivatives that is itself derived from many long-forgotten originals.
This is how I think the law fails creators. You can protect content, but how do you protect a style? A style that was itself inspired by another style or combination of styles and perfected over many years of practice.
That's how the creative process works, and AI does nearly the same thing. The AI learns from creators in a similar way creators learn from each other, by studying their works, techniques and styles, then combine them to develop their own. As creators we are all thieves.
The difference is that it is software doing it, not a human being.
The browser you are using to read this text right now is using data the same way to display my words, this web page, text, ads, video and thumbnail images as these AI companies do to scrape training data from the internet. Every image on this page is temporarily stored on our computers HDD and memory until it is cleared to make way for new data. The authors didn't agree to us having their images on our computer but they accept it because there isn't any other way for you to view their content.
The internet is nothing if not constantly copying, storing and deleting billions of images, text and video. It would take the age of the universe several times over to get every creators permission for all of it.
In the early days the data wasn't encrypted and compressed, so if you wanted to download a video and the browser wouldn't allow it, while the page was open in the browser you could just go to the cache folder and copy and rename the video or image to another folder. Now most web pages can simply be copy/pasted to another file or directly downloaded. even TH-cam videos can be downloaded via browser plugins or online services. Even the biggest players just gave up.
Creators had to go to extreme lengths to protect their work. Apart from a few exceptions like Napster - DRM, encryption and takedown requests and lawsuits didn't work then and probably won't work now. It may be that AI is something artist are simply going to have to learn to live with like they did with online music and file sharing.
I also dispute that the data collection is "theft" _as we normally define it_ -- even a traditional artist wanting to reference a specific style or subject matter will doubtless locate various images without (necessarily) checking the source for creator consent.
So if the primary difference is still a "man vs. machine" issue, that the generative models are trained using FAR more raw data than a trained human would require for a similar purpose -- and that once sufficiently trained, the model can be used to generate "a" result in a fraction of the same time as a human requires -- maybe that really is the difference worth protecting.
Consider, loosely, the difference between a pistol and a machinegun: both are fundamentally the same device (a metal projectile launcher) but the sheer difference between their respective rate-of-fire defies useful comparisons.
@@II00I00 Except for the jobs ACTIVELY BEING ERODED.
@@II00I00 Piracy is most certainly a theft, even conceptually, since the "pirated" version is _by definition_ a functional (sometimes arguably superior) replacement for the original (legitimate) copy, and produced by someone _other than_ whomever is authorized by law to do it. And it, too, is a subject almost entirely overrun by its ethical questions. For example, if the copyright holder decides _not_ to publish copies for a certain region (or even at all), the law says you must respect their right to do so, even if you might prefer otherwise. And if your motivation for a given pirated copy is _BECAUSE_ the owner refuses to provide a legit copy ... whose side should the law be on?
@@II00I00 Well, it is a fair point that when something physical gets stolen, the problem is less that someone else has it now, but more that _you no longer have it yourself._ From which standpoint yes there is a fundamental difference.
Though software piracy is still mired in the vague, broad territory of "unfair practices or competition". And law has always been better suited to defining what is clearly _unethical_ or "wrong" to do, not defining what is ethical/right. (In fact, the law is sometimes _entirely forbidden_ from defining what is ethical/right, most notably with freedom of religion).
the phrase "James Somerton on an industrial scale" is vaguely amusing, despite the awfulness.
Re: 17:20 - It's already been discovered that these techbro companies are struggling to scrape enough images to train generative AI, despite them being revealed to be stealing from the internet. If scraping _the internet itself_ isn't enough to get AI to not mess up fingers (in fact, the tech's been getting worse, not better due to AI copying AI), it gives me hope that generative AI is going to fall away like NFTs. It's just annoying how long it's taking. Thank you for this video.
Definitely agree. The best-case scenario is if all the corporate attempts to replace human skill with AI models backfire on enough levels for the models to gain a _reputation_ for failure. Convenience is compelling but only until reliability matters.
As someone who is hard of hearing, with variable tinnitus, and who is active in the disability rights movement (sharing videos with those who are deaf, and with audio processing disorder), I *need* videos with closed captioning. The auto-generated captions are getting better, but they still have an accuracy bias toward typically "White," "Male," "American"-sounding voices. The thought of creators opting not to put captions on their videos, in order not to get their work stolen, honestly makes me want to cry.
Even if A.I. *weren't* theft (and it absolutely *is*), it would still be morally wrong because it's being sold on the premise that every piece of art and writing has to be polished and "perfect" from the very first attempt in order to be even acceptable, robbing people of the joy of learning new skills on their own, and developing their own, unique voice.
Tech should be used to do all the menial labour so we can focus on art and self-enrichment, but the tech overlords want to do the opposite
Agreed - this is one of the most profound switcheroos of the AI/automation industry.
As a person with fine motor issues, I'd love to have a tool to create passable visual art that actually resembles what's in my head. Unfortunately, it functions only because people with talent have been robbed. I'm not that okay with that.
We might get that kind of neural interface at some point, (we've had *some* successes along those lines) but I think one of the problems is that people don't all visualize the same way, or to the same degree, but we also tend to adjust what is in our head based on what we're looking at, so there's some weird feedback issues to sort out. Amusingly, this is actually something that machine learning could help with and that cannot be "just scraped/stolen". Sorting signal from noise, and "is that a stray thought or an instruction?" do involve the right kind of specific statistics questions they *can* do.
Just... probably don't volunteer for Musk's project. They *did* make a brain-interfacing mouse, it just... fell apart because real-world brains are wibbly jello.
Contrapoints, Philosophytube, Tom Scott... I'm still checking all my favorite creators, but these three already came up.
They have pretty solid subtitles, so it makes sense they would use their content
Oh, this pisses me off so much! If a human couldn't be bothered to create it, why should I, a human, be bothered to consume it? ...as soon as I determine the video I'm watching has an AI voiceover i turn it off! Hire a human!
Though I largely disagree with the OP's claim that AI is theft, I concur with your statement about AI voice overs at this moment in time. They're awful, inhuman, and cheap; if the business putting it out there can't make the effort to use a human VA, what does that say about the product they're selling? Most likely, it's cheap, too.
@@this_epic_name I appreciate that, but I would like to know, if you don't mind, why it is that you disagree with op's claim that AI is theft?
@@ThomasTulak how is it theft? do you understand correctly how Gen-AI works?
@@dani01949 well, my understanding is the creators of AI train the AI on existing things (with out the permission of the person who created that thing, hence the theft), and based on that thing the AI learns how re-create similar things. So, large language models learn to string words together based on the things it was trained on. Generative AI learns to re-create images based on the things it was trained on... This sounds like theft to me. If you have a different understanding of how it works, I would like to know!
I, too, will be awaiting dani01949's explanation to enlighten us all on how we are wrong since they seem to have some strong opinions. :) They seem to have replied to various users' comments saying they're "misunderstanding" how generative AI works but have yet to provide any defenses. So please, I'm waiting.
Oh and that epic name person or whatever's explanation as well. Funny because these kinds of folks never come back to you with a proper answer, but oh well lol.
Exactly, the "Taking inspiration" is a human thing. AI is not sentient. They are anthropomorphizing computers. And even the using a tool like the Camera/collage argument falls flat. A Camera operator is selecting the shot and scrutinizing it themselves. The collage artist selecting bits and pieces of colors, images, etc is selecting it themselves and arranging it, figuring how each fits together. AI would only apply if you were the one selecting the each individual part of an image, not asking a computer to draw it for you.
As a freelance digital artist, its always good to hear people rant about Generative AI.
I am always reminded that I got into an argument for AI voices are fine as long as the person they are stealing, oh sorry willing volunteer as they kept putting it, from gets paid and credited which just annoys me cause that is not any of that shit works.
I don't use AI art, but I do use speech-to-speech voice changing for characters in my videos.
And even if you grant that, these companies are still stealing stuff in practice anyway.
I am looking forward to seeing the AI bubble crash. AI as an industry is hemmoraging money and the software itself isn't improving.
Even fancy generated videos are useless, because you can't edit them. Find a nice shot but want to change the colour of a hat? Nope; you can only get a whole new video.
I'm also quite glad that the use of AI has already become synonymous with cheap, poor-quality work.
I'm a 3d digital sculptor, I came into making art very late in my life (approaching 50 now), and AI is starting to creep into my field - I do digital sculpting to make miniatures for tabletop gaming - and I get very angry seeing so many people posting AI "art" and pro-ai comments. Artists - actors, musicians, painters, writers (and everyone else doing something creative) are already so marginalized and struggling in their fields as a whole, because the masses look at art as being too expensive to pay for a commission. Musicians playing a gig are supposed to do it for the exposure and so on and so on. "Generative" AI is just a further slap in the face to a group of people that already have to often work other jobs to live just to give up on their art (at worst) or just do it as a side hustle (at best for most). Not saying that there aren't those that can make a living doing their art, kudos to them. But, starting out right now for younger people must be so defeating.
All in all I 100% agree with your statements. I just wish the world could have to go a week without any form of art, just to see how very important it is to the human condition - no music, no painting, no movies, no books. nothing. Only then, maybe, people would see how truly important art is. Stealing others art is absolutely disgusting. I'd rather see folks make "bad" art, than steal something with this generative bullshit.
make sure the arguments are correct, first. You are agreeing with lies.
Vera, as a playwright who has been offered jobs to help 'train AI' on Mandy network... Thank you for this. 100% agree.
"Hi, I'm not a Generative AI, but I play one on TV."
Sorry. I needed to do that.
I hope, as a culture we'll develop filters to block the mediocre, bland, unimaginative crap that is churned out by AI. It's the decision-making algorithms that really freak me out.
I like to listen to narrated scary stories on TH-cam while I do other things. A good channel will produce maybe one video a week where the channel host is the narrator and the work is credited. Over the past couple of days, I've been finding new channels that are posting 1-2 videos a day, the work isn't credited, and there's no narrator listed. The voice narrating these videos is pretty good, to the point where I almost can't tell that it's AI narrated. I'm actually scared for the survival of the channels I enjoy because they can't keep up with automation.
The auto-generated subtitles for your videos are pretty accurate
I find it hard to watch certain people who do things with AI Dungeon and the like, because I love their content, but hate the fact that they are continuing to use these services that continue to steal work from genuine people.
There is no denying however that what TH-cam did is completely and utterly fucked. Especially from creators who are struggling to get monetization, or are hit by multiple bullshit copyright claims. What a fucking joke this decision TH-cam decided on is. If it didn't make it clear enough to creators that TH-cam doesn't care.
As a recent writing graduate, it angers me to no end that "AI" is taking away my jobs. Hell so many writing jobs on Indeed and LinkedIn are just to train an AI software. It is them asking me to give their tech my talents so they can make money.
I mean...the editorial industry has done the same before AI. Only rich people can write full time.
@@dani01949 You've got a point to an extent. Rick prime can just do it without a care in the world. As much as it sucks that they can just do that while writers who want to just get started can't even get hired is infuriating.
I don't know where I stand on this debate, the one about AI scraping publicly available data being theft. I think it's ambiguous because people create youtube content to (make money hopefully sometimes, let's put that one aside) to entertain and inform. If a company is using a video to inform it's neural network then how is that theft? If someone who charges fees for a course as part of teaching that course embeds a youtube video (linked to youtube and giving increased views on youtube when it's shown so they don't "copy" it off the site, copying off the site to play offline is unquestionably theft) they are using that video as a teaching aid for something they get money for and the video creator does not. Is this any different?
As I said, I don't know where I stand on this debate if AI is theft. Companies using other people's work without permission to train their neural networks being unethical, AI being negative towards creators and taking us on a downward spiral of originality, I can get behind those ideas and agree with them no problem. I don't defend all AI to the death. But using speech patterns to emulate a more natural generated speech flow - I'm not sure that is theft. This video uses the phrase "AI is theft" a lot, but you never explain why you think it's theft. This is not the same as someone breaking into your house and stealing things but choosing not to fence them, or someone taking your savings to set up a charity. In those instances you had something that was for your own use and you were deprived of it and could no longer use it. In this case you have created something to inform and publicly sent it out into the world to inform. You still have it, you can delete it in the future if you want to, it's still informing people on your youtube channel after AI scrape it. There are multiple variations of theft, just the two you used don't work here in my view.
I am still not saying here I think companies scraping other people's work to train their neural networks is not theft, just saying the analogies you used here I don't think are comparable, so I'd find another explanation of why you think this is theft helpful in my understanding of where both sides come from and then I might have more of an idea where I stand here. I don't know what your end goal with this video was, personally I think anyone who was against companies scraping data without authorisation will still be against it after watching this, and anyone who thought it was okay before watching will still think it's okay. If you want to move the needle on this debate, and either change someone's mind or help them make up their mind (like me) I think explaining why you believe it to be theft would be a lot more effective than just saying it is.
I agree, pretty much. You elucidated my thoughts on this better than I did, so thank you!
I did not expect to see James Sommerton invoked in a discussion of why AI is bad, but I’ll be damned if it doesn’t clarify the point perfectly.
but it is based on a completely wrong assumption about how Gen-AI works 🙃
Already knew all of this, but it's good to see someone getting openly pissed as much as I am about it. Also... fun fact... one of the recommended videos in the sidebar was a video for a channel that was all about "How incredible the AI revolution has been and will continue to be". Just went ahead and flagged it as "don't recommend channel". Just sucks that no matter what I do, TH-cam keeps pushing that bullshit on me.
From the inception of "generative" (read plagiarising) AI was a thing I was absolutely an opponent even though I'm a fledgling pretty bad digital artist who is both self taught and is neuro divergent where improvement could be taxing to my mental health at times. The claims of gatekeeping and accessibility are absolutely moot as art is accessible to anyone and there's a ton of free resources if you only try to look for it and put in the work and soul into it. What didn't occur to me until it was spelled out by you now is that this have been fed on an accessibility feature which is a newer even more disgusting low for AI to get to I never imagined possible before...
I don't use generative AI for my channel (different account). But I also have a hard time finding a good, dependable artist to do art for my channel, whether in general or for my animated model. Where are all these artists who are looking for work? I have the resources to pay them fairly... I just don't have the resources or connections to find them.
There’s a thriving art community on Twitter and I’ve also had work done through artists on Fiverr, though I haven’t been back there lately and I don’t know if AI is taking over.
@CouncilofGeeks Maybe that's my problem... I don't have a Twitter account, and I have always loathed the platform. 😣 I have heard of Fiverr, though. Thanks for responding to my comment, BTW -- I really liked the video!
@@britzkrieg2 I’ve found Fiverr very helpful. Some of my merch designs are from there.
A lot of artists post work on tumblr, too. I used to, before AI companies started image mining, then I took down almost my entire online artistic presence until further notice. Unfortunately, I know I'm not the only artist who's done this.
Btw, just out of curiousity, what kind of work are you looking for? Digital or trad media? Any particular style? I might be able to recommend someone.
@thing_under_the_stairs Right now, I'm mainly looking for someone who can revamp or convert my current DIY VTuber model. She needs a new hair style and a change of clothes, for starters. A new YT page banner would also be nice. I did get a new channel logo, but the artist took a long time and sometimes wouldn't communicate with me in a timely manner. I'm looking for someone who likes drawing birds in the abstract.
I’m so sick of ai, I want it to fizzle out already but it’s still going
To me, Ai is slop, it says nothing, it means nothing, and it has nothing. There is no message, there is no care or love or work put into it. You don’t get ANY of the joy of making something. I mostly make a billion ocs, and if I used AI to generate their appearance, it would be bad. Maybe not in technical skill (though it is likely because AI will never be perfect because it does not understand 3D space nor does it understand stylization) but it would be bad due to character design having meaning. There are lots of little things in character design that help SHOW you who the character is instead of telling you. AI doesn’t understand this. This is the case for every art form. It doesn’t get what makes the art work, and makes it have soul and intention.
When I look at AI “art” I feel nothing except for disappointment
The first half of this reminded me of someone's post where they found a fanfic someone used an AI LLM to generate, and they said something to the effect "if you didn't care enough to write this, why should I care enough to read it" and that's pretty profound to me. If you consider yourself non-passionate enough to try and one-click generate (regurgitate) something based on theft, why should I care when your anti-passion crosses my feed?
Something I am not seeing being talked about is that not all data is the same. Everybody seems to act like all data is the same, but there is a difference between a meme that I created versus a landscape photograph I took to sell prints versus photographs of my pets, family , and friends versus my own personal journal that I have as a Word document. It is one thing to use my landscape photo to train AI which is bad, but it is another to turn around use my personal photos and thoughts to train AI. One thing is I wish we as a society at least had a conversation about where data can and cannot be scraped. There has to be at least some boundaries as to where this data is being scraped. Especially, when Meta has stated that they are scrape all our data from Facebook and Instagram whether it is a public post or a post that we only wanted friends to see. And there is controversy over if Adobe may be possibly scraping data right off our own hard drives to train AI. There has to be a boundary as to what areas are being scraped.
I am glad that you also brought the environmental factors as well. Training AI is using massive amounts of electricity and water. But something I heard SomeOrdinaryGamers pointed out which I didn't think about which is these AI companies are creating massive amounts traffic on the Internet as they go out and scrape the Internet. Who knows if this is going to slow the Internet down.
I am surprised that commentators on the Left aren't talking more about AI from displacing workers, to using the work of artists so corporations and billionaires can make billions, to the environmental impact, and to the possibility of misinformation to deep fakes.
The end game of this is where hardly any human made art is left for AI to copy and it creates an infinite loop of AI copying itself, and never seeing a new art style.
You’re so right, helping artists my ass. They’re just using that excuse to steal from them. I understand that it can be an useful tool for real artists and people that work in other creative areas, but at this point is doing more harm than good, and its usefulness doesn’t justify all the harm that its doing in any way.
Steal from them then charge them for use of their product! It is like stealing the plans for a faucet then selling the faucet back to the people they stole the plans from!
@@SirWussiePantsSo true! I forgot that you have to pay to use most of them 🙄🤦
The harm is not the fault of the tool itself, but the massive (and legally influential) companies backing it.
Excellent video, Vera! Love a ranty rant about ai “art” 💐💐💐 all the flowers 💐💐💐
On the matter of artists permitting, I believe terms of WGA settlement included any training on writers' work required their permission. Although I had the skeptical concern that if it became a default in contracts, writers wouldn't really have other options, it did seem like a win.
Do you think there is a place for AI to recreate lost Dr Who episodes like Ian Levine is doing?
Absolutely not. And not just because that man is vile.
Cool idea. Eventually, it'll do that, as well as create entirely new episodes! :D
No, an AI-recreated lost Dr. Who episode wouldn't actually be a faithful representation of the video or audio. This isn't an "AI can't POSSIBLY get that good" question, it's a "we have too many missing pieces for AI to fill in the blanks" question. For example, let's say we have the script to an episode, but no audio or video. Even if an AI can perfectly recreate the likenesses of the actors and their voices, there are millions of micro-decisions made by both talent and non-talent on the set, in the prop department, the camera department, the makeup department, and the editing booth that impact the final picture. The script alone will not provide enough information as to what those decisions were. What you would wind up with is not a recreation of the lost episode, but a shot-for-shot remake of it.
If you had the actual video or audio but it was damaged, an AI could absolutely clean that up. In fact, that technology has already existed way before generative AI and all of its copyright problems.
@@SuperSmashDolls If a "shot for shot remake of it" is the best we can possibly do, for something that has effectively been destroyed, and lost to time.. Then I'm not sure I see how any of that really matters. It wouldn't be about getting back exactly what was lost, but rather just filling in missing pieces, to create a more complete story, in a way similar to how it already feels complete, to someone who watched the originals, before they were lost.
Additionally, we may eventually see generative AI so powerful, that it can effectively create "windows" into customizable realities, or other "universes", giving us something along the lines of "Interdimensional Cable". Heck, it'll probably also be interactive in real time, through a game controller or VR devices.
Generative AI is just Boggle for A-holes
The reason its different from a person taking inspiration is because being a person is fundamentally better than being a game of boggle. Humans have intuition and intention and you cannot give that to a boggle game.
And for the people who say "well i don't know how to do x,y,z but i need it for my other creative project!"
No you don't. You're creative you can compensate for it, figure it out, learn a new skill, make friends and collaborate, and otherwise get good.
100% agreed! Lots of the best, most beloved creative works out there turned out as special as they are BECAUSE OF the constraints the creators were under when they made it. Whether that's budget constraints, limits in tech, limited number of actors, limited space, limited time, or they just weren't formally educated in that field so nobody who knew better could tell them "this isn't how we do things" or "this is impossible".
A couple years ago I worked two-three data labeling projects for about two months. Most of the people working on them were people with a disability or that for any other reason was unable to find other ways of generating income, the process of training this AI is already pretty exploitative. Landing projects is inconsistent, the companies that sell the service are really shady (I worked remotely for a company in the Philippines), I was lucky they paid me as agreed in my contract but I quickly realized that if for any reason they decided not to pay or partially pay I’d had no recourse to get my money.
The labeling itself could get frustrating as it does not allow for nuance. If projects were understaffed we’d be overworked but the pay was really good, if on the other hand the project was overstaffed we’d all be waiting for work to come in and try to take it before anyone else could.
One of the weirdest freelance gigs I ever had.
Nitpick: "Generative AI isn't really AI" is a pointless point to argue. First, changing what we call the technology wouldn't do much. Second, generative AI has as much claim to intelligence as video game AIs, and that use of "AI" is older than I am.
It's wild that I got an ad for an AI service when I clicked on this video...
Microsoft is bragging about the 'AI Revolution' whenever I start up my computer thanks to Win 11 start/log on screen.
Wow. Glad I still use 10.
🤢🤮🤡
Generative AI should effectively be treated as guilty until proven innocent BY THE LAW, imo.
The analogies to someone breaking into my home and stealing my things don't really work for me - and it's related to why I don't like that we're using the words "stealing" or "theft" for copyright-related violations; if someone steals my television my main problem with that isn't that they now have my television, it's that I no longer do. Art theft causes issues, but they're not the same issues regular theft causes.
Besides that, it seems that the argument for why generative AI being theft is based on a) it's not human, and b) it's only learning to do exactly what humans have done, it's never adding to it. I would argue that most humans who learn from others do the same, though - they may pick up small bits of inspiration from many other artists subconsciously, but they are mostly learning to do what others have done before anyway. Humans who introduce something identifiably *new* to their art, something that has not been done before by anyone, seem to be a rarity. The argument that AI isn't human is somewhat of a circular one, as the question in the first place is what differentiates what generative AI does from what humans do. You can draw a line at whether the "learner" is human - but I don't think there's a good argument for why that should be the line.
To me, what current generative AI does is still theft (in the copyright sense), but it's not because it's an AI learning to do things like humans have done - but because it isn't learning to do things like humans have done, it's learning to reproduce *exactly* what humans have done. When a current AI generates an image, it's not trying to do something like what it's seen, it's trying to do the exact same thing (or, specifically, small bits and pieces of a lot of exact same things). I can imagine a better AI in the future that does something closer to what humans do to learn, though, and I don't want to preemptively say that that's theft too based on the fact that it's a machine doing it rather than a human.
As for calling it "generative", I think the word is used in a technological context, i.e. the same as when it's used in "random number generation". In that context the word is precisely correct :)
In your third paragraph, you probably correctly state what earlier gen AI models did, but they evolved past that long ago, and many image generators now have settings for it (sometimes labeled as a "creativity", "variation", or "chaos") . On one end of the spectrum, you can tighten a model to make it produce things that it has seen before. But on the other end of the spectrum, you can loosen the model so that its outputs are far (sometimes *really* far) from anything it has seen before. It's all in the math. So while it could be argued that gen AI tries to reproduce things it has seen before (at least in the general sense), it doesn't have to be the "exact" same thing, nor must it be built to try to do that, as model parameters can be adjusted to give the model leeway in "how close" it needs to get.
The comparison between AI model training and human study/inspiration is a rough one anyway since we don't even have a working (scientific) understanding of how the neurons of a human brain operate -- though, obviously "it just works". You end up comparing one black box to another....
I also can not and WILL NOT label AI models as "theft" by any definition that could equally apply to a human artist scraping/downloading images from the Internet for use as reference/study material _(also_ typically done without confirmed creator consent!). What IS the difference we care about here, if not some flavor of a "man vs. machine" argument?
Scraping for subtitles certainly explains why youtube fairly recently rolled out the crowdsourced "help your favorite creators by offering corrections to their subtitles" function.
That feature is so old that it was one of the things you could earn points for in TH-cam Heroes. But I wouldn't be surprised if TH-cam is pushing it harder these days.
I'm an artist and a disabled person. I was pissed enough about AI potentially scraping and stealing my work, but the fact that they're using disability technology to do it. FML.
Even if I, as a writer, trained a generative model on my own writing, it still wouldn't be ethical due to the extreme enervy cost, the lack of any valuable output and the supporting of any kind of software that involves generative ai
I have to say, pointing out that Generative AI is nothing more than "James Somerton on an industrial scale" has got to be the best argument against its use that I've seen yet.
untrue, though.
As someone who has taken several courses on Machine Learning in college, I love the tech behind the nonsense that is "genAI". The algorithms themselves are a wonderful tool, but they require a shit ton of data, and the data has to be decent data. Garbage in, Garbage Out and all that.
The one thing that these algorithms are good at that have surpassed humans is finding patterns that require looking at data in a large amount of dimensions (which, as humans, our brains are pattern recognition machines. Hell, we try to assign patterns to things that have no business being group in a pattern!)
I see a useful tool that has been warped to serve the capitalist machine.
I want to see the tech behind this to be used to help with summarizing personal notes and finding connections that I am not able to see. I want to see this tech used to take different medical data points and perhaps help catch things before it is a major problem. But, because none of this brings in profit like letting people find some sort of creative outlet (no matter how unethical it is) seems to do, the only way that gets expanded on is through personal projects.
I am so fucking conflicted on the use of it. I can see so many uses beyond the stealing of other peoples work, but I don't have the personal time to work on any of that.
13:36 the way I like to think of this for myself is that a person can look at a picture (or hear a song, read a book etc) and see ideas. AI only sees pixels.
The debate about the artistic merit of AI art is really easy to sum up: why should I be bothered to look at or read something that nobody could be bothered to paint or write?
I’d like to rebrand ‘Generative AI’ as ‘Plagiarism Engines’. It’s more honest, and stops people conflating these awful things with proper General AI. (Which I think is like scalable fusion energy: always twenty years away.)
The small dev team game thought experiment really does fall under the same standards as businesses: if you can’t afford to pay employees a living wage, you can’t afford to be in business. Find artists for what you need who are at a point in their career where you can afford them. The art style and music doesn’t need to be comparable in complexity to a AAA game to be amazing and beloved. Look at Stardew Valley, or Undertale.
Thanks. I have exactly the same feelings about it. It's not that it doesn't work (although it's way overhyped in what it can do), but in-so-far as it works it just a highway to dystopia. Who is meant to buy this crap once most of the population is barely able to pay rent and food? What money will there be making art if the market is over-saturated beyond belief with just endless amounts of AI-generated books/films/songs/e.t.c? And why should a bunch of worthless talentless billionaires be rewarded for being a bunch of crooks? I don't support AI as a matter of principle, because it's humanity careening straight back into the dark ages.
Thanks for talking about this, Vera.
I've done hundreds of illustrations for stock sites over the past decade. Of course they changed their ToS to be able to scrape all my work to use for AI generation. Now there are people generating absolutely garbage-looking AI images to sell on those same sites. I never made a living wage off stock sites, and I've not seen my profits on those sites increase, in fact they're decreasing bc the sites are just flooded with garbage now. FEELS REAL BAD...
Totally agree. And on the point about it being a waste of resources, we now have tech farms that will always be looking for the newest way to make money whenever the last scam collapses. I don't know what needs to happen to fix that problem, but I fully expect at this point that there will be a spiritual successor to crypto, nfts, and ai, just because people with money and resources will it into being.
It's kind of wild to see the amount of AI "artists" that put copyright marks water marks all over their images because they don't want people stealing them. Just imagine thinking that you have any copyright protection for an image made off of stolen art in the first place. And not being able to see that it's ridiculous for them to try and have that double standard
My problem is that AI was always supposed to be trained on data in the public domain.
Ripping TH-cam is then fraud.
It never was
AI companies burned through public domain data ages ago. It wasn’t enough to get these things to be any good.
Was it though ? Ai companies don't care about artists, they only care about money. They have been ripping off every plateform they can, even art programs (Adobe and probably others), meaning they're stilling the art *while it's being made*...
@@novae756 well public domain was the cover story. It was always a lie though.
@@CouncilofGeeksthe public domain doesn’t even have enough data for basic, modern day function of these programs
I mostly agree with this video, except for you saying that 'generative ai' is like a Frankenstein thing, and that its not 'generative' I think that is a fundamental misunderstanding of how the technology works. Sure it does use data from a bunch of things, but the way it uses that data isn't anything like a Frankenstein. To give your argument some credit, though, I will say that its a Frankenstein of ideas. Like, it doesn't understand what hair is, but it has the 'idea' of hair, and an 'idea' of a human, and it knows that hair goes on humans. Etc. And by 'idea' i mean like, conceptualization, not like an 'original thought'. What generative ai will do is take these concepts and mash them together, based on a prompt, like you'd tell it which concepts to use, and it mixes them all together. I think those are called 'tokens' but don't quote me on that. I think what some people imagine when they hear that ai takes data from people and mixes it together, that its like a photoshop, like stitching what this person did with what this other person did. Unless you have a tiny dataset, the output isn't really just gonna copy what someone has made exactly. I mean, of course it's still possible, and if the model is poorly made, its more common. But I think 'in general' it doesn't work like that. I do want to hear more about what you think it is since I must not be fully understanding, and maybe I'm wrong. But that's what I think
Your explanation is much closer to the mathematical reality, which this video doesn't really bother with
Let me preface this with some credentials: I am an AI scientist (university degree) and an artist (no I don't use AI for my art). I think that AI isn't the problem, it's capitalism and a lack of tech literacy in the people responsible for making policy.
AI scientists have been calling for stricter regulation for (AI) tech years before the unnuanced "AI equals evil thiefs" thinkpieces hit the mainstream media. It hurts to see the reputation of the field that I love be tarnished by a version of AI that is just a mockery born from capitalism.
All generative AI means is that the AI model has an output that is similar to the data entries it is trained on. In the same sense that a random number generator is neither random nor "generating" "new" stuff. Generative AI can and IS already used in respectable ways, such as in any sort of research that seeks to explore new options, such as finding promising candidates in medicine research, maths/geometry, or materials, etc.
But that news gets overshadowed by all of the enraging "AI is bad" discourse clickbait since rage sells. Yes, that application of AI is bad, and luckily anyone with half a brain knows that by this point. But while the general public points and laughs at GANs creating sausage spider hands, the capitalist elite points and laughs at us because they managed to trick us into getting mad at AI instead of them.
I don’t even really disagree with most of this. But with all respect, I feel like we’ve got a better chance at halting AI than we do fixing the capitalist system that makes it the worst version of what it could be.
@@CouncilofGeeks I really think that's seeing things backwards. We have a FAR better chance to adapt our economic system now, than we have any chance at stopping technological progress in AI. Capitalism has been going through 'late stage pains' for a while now, and maybe AI is finally the thing to push it over the edge, by invalidating the value of human labor?
AI is the best justification we're ever going to get, to demand major economic change. Historically, economies have always changed!
I just cannot see any way in which we could "halt AI". Regulations can't stop everyone, companies will still do things in secret, and the open source community will only be emboldened by regulations that say they can't do something. Nothing we do can truly stop other countries from pursuing it either, especially our enemies.
@@NikoKun history says you’re very very wrong about “invalidating human labor” leading to a system change. Automation has already shown that regardless of the fact that it *could* have meant physical labor wasn’t required and freeing humans to not have to engage with it, in practice it just left millions of people without marketable skills. We’re currently headed for a world where work and income are still required but the available job market keeps shrinking. Get universal basic income installed and I’ll be more willing to budge on this.
@@CouncilofGeeks The difference this time is things will happen much more rapidly, and impact a far larger portion of society. The historical examples you're referring too, happened to much smaller subsets of workers, and over a longer time span, so rather than invalidating human labor, it merely shifted it around.
I agree with the need to implement a Universal or Unconditional Basic Income, but for that to happen, there will need to be mass demonstrations demanding it.
We may also have better luck with such an idea, if we frame it as an AI Dividend for All, that everyone would get as a return on their data investment which helps train AI.
@@CouncilofGeeksYou're right, capitalism is not going to vanish any time soon. But AI is only a technology. It's not going to vanish either. Nor (in my opinion) should we want to go down the road of banning technology/research. It exists, and all the wrong uses of it are ultimately the symptom of the underlying systemic problem.
While it's highly unlikely we can outright dismantle capitalism at its core, what *is* possible is coming up with solutions that target the underlying cause. For example, more knowledge about AI and the dangers of it should flow to the people who can influence policy and legislation. A huge part of the issue is that legislation is 20 years behind on the quickly developing technological landscape. As a result, we're left with an anarcho-capitalist wild west where corporations can do whatever they want because our politicians literally don't even know how Facebook works. (I'm EU based, but from what I've heard of the USA Senate your situation appears to be painfully similar)
Another solution, more specific to generative AI, is to redefine copyright laws so that they are more up to date with AI and other technologies in mind. Technically it's not the generative AI that's wrong, what's wrong is that the datasets that are being used for this AI are often not transparent and unethically sourced.
I totally hear you regarding universal basic income. I agree that as is, the future doesn't look great. Jobs will change, very quickly, because technology is developing very quickly. And that leaves a lot of people without marketable skills. New jobs will be created, because any technology ultimately needs to be maintained, but many people who are being outcompeted by technology like AI would have to gain a new skillset, which is a very big ask.
So yeah, I think we probably agree on a lot of points. Where we may differ in opinion is where we prefer to put emphasis. I believe that focusing too much on (generative) AI being the "bad thing" is a waste of completely valid outrage. It's fighting a symptom. I don't see a complete ban on generative AI as realistic or even preferable. Even if it can be pulled of, at the rate technology is developing, some new tech will probably come along within a couple years which will just expose the same exploitable weaknesses in current legislation.
I've said in a few places now "If you want soulless quantity, use an AI. If you want emotional quality, ask a human".
So-called generative AI can be a useful starting point though, at least in theory. If you have a rough idea but are a bit stuck, plugging the idea into the program and seeing what it comes up with can provide that missing link to kickstart your own actual work. I've seen more than a few digital artists throw a concept at the AI and then alter it to properly fit their vision and style. As a first step, it can be valid, but if your whole process is just plugging the idea into the program and spitting out the result without applying any of your own touches to it, you're wasting everyone's time.
I haven't watched the video yet, not sure if I will because I know they already have an anti-AI bias, even though this video is supposed to be aimed at generative AI, but I'm with you. That's one way in which I use generative AI art, as a starting point, or a way to get beyond a mental block and so on. Nothing I generate gets used in the final work. In fact, it's no different than if I'd spent hours scrolling through galleries looking for the same thing, except with generative AI it's taken minutes instead of hours.
Your comment about 'soulless' brings to point a thing many don't grasp, the difference between generative and creative. A computer can only do the first, while a human can do both. That lack of a creative touch is where generative AI gets abused by companies and those looking for a quick pay day by simply plugging something in and running with the results.
@@Elwaves2925and real humans can make soulless art too. If we are calling the AI generated stuff art.
@@eastlynburkholder3559 That is very true. Tracy Emin comes immediately to my mind, along with a frantically scribbled pencil piece that looks like it was drawn by an angry child, yet it sold for millions. Wish I could remember the name of it.
Generative art is definitely art. Fractals have been considered art since the 80's, maybe earlier, and while it's a different technique, they are generative and use algorithms.
@@Elwaves2925yes there are art marketers who can create a trend, who can create a hype for a specific artist, who can create markets. Maybe the soulless artists are not solely to blame for the soulless art being made a distributed.
I vote we start calling 'Generative AI' Industrial Plagalism.
I like calling these algorythms LIGMAs (as a tortured acronym for Learned Image Generating Machine Algorythms) instead of AI cause that's what I think of 'em.
We know it's theft, so my question now is... what do we do about it? What measures can content creators take against this particular art theft as well as against art theft through A.I. in general?
Thank you for articulating this in such an effective and understandable way. I always struggle in arguments with people who don't seem to get it and you managed to explain it so well. Next time, I can just refer them to your video.
Somehow, I don't think this algorithmic slop generation is going to bottom out like crypto or NFTs. There's been a trend in media of following formulas more closely and limiting creative freedom. They're getting people to make algorithmic slop already, and using the algorithmic slop factory would only be cheaper. Labor is a sizable portion of the cost of doing anything, and if a company can reduce that cost, it will experiment with doing it. If making slop movies is more profitable than making movies with creative merit, they will do it. It isn't just creative fields, though. The attempt to make these algoithmic slop generators more conversational suggests that these companies might try to make cheap tech support, cheap customer service over the phone, cheap secretaries, and many other writing-related jobs. From a business perspective, there is no reason to pay a human a living wage to write the quarterly report when you can just generate it. I don't think this is like the funny bored ape pictures or the blockchain money. It just feels different.
(important to note that this is speculation on my part)
Interesting with the subtitles. I’m an artist photographer, I wonder if this is also with alt text and descriptions as without these the photo or piece of art is meaningless for training. I no longer put descriptions or alt text, which unfortunately does disadvantage the impaired. I have also been meticulous with scrubbing my metadata before uploading photos to social media sites. The less information about a piece the less useful it is for these companies.
And yeah, I’m with you. This is so wrong on so many levels.
Got AI voice over ad...
When it comes to AI, I do think there are a few valid art and art-adjacent applications... but it's primarily in things like upscaling and photo-repair software. I've tinkered with AI-assisted upscaling and repair on old, small, damaged family photos and gotten some promising results. And I do think it could be a valid way to make "HD Remasters" of shows where it'd cost too much to do a traditional remaster, or fix the plethora of tiny issues that come with film gradually degrading.
Similarly, I wouldn't strictly have a problem with a CGI artist using a generative AI to make textures for 3D models, as long as it followed one of two factors...
A) It's based exclusively on photos they took, or...
B) It exclusively uses paid-for, licensed stock photos that the photographer gave consent for to be used in an AI engine.
That's basically the only way I think "generative" AI art could be valid and ethical.
But the way it's being used primarily now, mostly to rip-off art-styles and replace jobs, is just very scummy and unethical.
I work as a captioner for educational videos. It takes a lot longer than you think to do captions, particularly with anything with multiple speakers, or educational or slang terms. I have to look up a lot of things to make sure that I'm using the right words and the right terms.
Thank you for this. Dunno if my work has been scraped [I know that one well-meaning 'fan' tried making synthesized versions of it and I basically said "Nah" and blocked him]. But I do know that even that 'tribute' felt a lot like the time some young woman came up to me while I was working and told me, "I love your style. I'm gonna steal it." It would have concerned me more if I thought that she had the years of effort, experience, and training that I'd put into it. I feel the same about the AI crap. I know for a fact they can't 'copy' my work, because it's stupidly complex- layers of color, classical composition theory, etc etc- but it's still insulting to be treated as some kind of cipher or replaceable cog in the creative process. Machine minds don't get it, I guess. Thanks again.
The confabulation engine is, dare I say it, HUNGRYYYYYYYYYY!
Prompt fondlers love a well-fed confabulation engine and they don't care what it snacks on.
With music, when it was lacking in the video, I spent a month learning the recorder myself and looking at tutorials. I messed around with Musescore, etc and recomposed public domain folk songs. I mean, it's not impossible to learn art. Creativity and art goes before Homo Sapiens sapiens. It's baked into our genes so hard. People act like there is a disability that makes it impossible to learn, but it only takes time.
I realize this is the legal equivalent of a Hail Mary pass, as we're talking about juggernaut companies investing ungodly amounts of money and time into AI projects, but I've heard that the so-called "monkey selfie" case provides a precedent to declare all AI-generated work to be in the public domain.
If this stuff sticks around long enough to be challenged in court, it will be interesting to see how copyrightable AI "art" will be. Now there are some work arounds for something like say, a movie studio, because for them the theoretical use of AI would be for individual components with some real people involved at other stages. So for instance, if a court said "your AI generated script is in the public domain," that literally only impacts the script and so long as all design aspects were done by people than all of the visual presentation of that a script is still copyright protected. It'd be like doing a film version of a Shakespeare play, whatever visual flair you add that wasn't on the page belongs to you.
Indeed, the US Copyright Office has ruled that AI-generated images are in the public domain. This ruling, rather than some sort of union leverage, is what facilitated the agreement reached to stop the SAG strike last year. When you do contract work drawing for Disney or illustrating cards for Hasbro for DnD or magic the gathering, you must sign a contract promising not to use AI for this reason
This stupid AI program my lab top wants me to use that I thankfully deleted destroyed my lab top battery.
We're living the future created for us against our will.
I'm sympathetic to anti-AI arguments re: making workers redundant in this capitalist hellscape, but the theft angle pushed here just doesn't work for me. The reason theft is bad is that it's taking something away from someone, depriving them of it. Training AI on something I'd hypothetically made and put out to be seen by the public for free wouldn't deprive me of anything, at least not directly. The fact that the only way to know it even happened would be to go on a website and run a search on the dataset to see if I showed up illustrates that. Same reason comparing copyright infringement/piracy to theft falls flat, but at least with that you can make the (albeit shaky) argument that it's depriving someone of the profits they would have made if their product was bought. I guess the equivalent to that in this case is the use of the training data to create AI models that deprive the creators of future paid work (which again I'm sympathetic to), but that doesn't appear to be the point the video is making.
I wouldn't really feel the need to point this out if the the video didn't go so hard on the "this is theft" thing. It's not theft. It's just not. It can still be bad, and be leading to terrible outcomes for creators and the world in general. But it's not theft.
AI likely won't replace the pro artists. HQ products and productions will always need experienced artists for illustration and concept art. But AI is killing the entry level jobs - book covers for self-publishing. Cheap ads. That kind of stuff. And if those beginner artists can't get entry level jobs, they won't be able to build their careers up. They won't be able to invest the time to gather experience and practice.
I have seen one (1) case as you mentioned: an artist training an AI on only their own stuff. I watched a youtuber a while back that trained an AI on his own voice, because he was losing his own due to a very serious illness. Valid undertaking.... I still had to unfollow, because the videos with the artificial voice just sounded... dead. Uninteresting. They were missing that human element.
one of the only use cases i’ve been able to think of for generative ai is for artist previsualization, where you need something quick and dirty to help pin an idea or help communicate it in a collaborative work, kind of like how a lot of artists i know will collage or make a pinterest board for inspiration or communication for a project they’re working on. a lot of the generative ai models out there seem almost perfectly tailored for that, but instead of pursuing that, all of these ai businesses seem to be in a race to see which can piss off the people who would have had the best use case scenario for their products the most
That's what we used to use sketchbooks for. It's just lazy.
@@thing_under_the_stairs That only applies if you're already an artist to some degree, which most folks aren't and have no desire to be. Yet they may still be in a situation where they need to get an idea across to people (who may be artists). That makes the case above very valid as long as the generated work itself isn't used in any way, it would be no different to looking through a gallery and pointing out what you want.
@@thing_under_the_stairs not all artists are 2d visual artists and a slapdash pencil sketch isn’t really helpful for all mediums. as someone who’s worked as a screenwriter/director for example, having generative ai as a tool in the same way pinterest boards and collages are could be helpful for communicating visual information to a storyboard artist, dp, or production designer in early preproduction for example. i don’t think the problem is generative ai in and of itself, it’s how it’s being implemented in our capitalist hellhole of a society
@@quarbarian2 If you can't do a rough sketch, you shouldn't be calling yourself an artist in any medium.
Yup. Had to put rule in my discord to be very clear on our stance on this subject.
I watched this video on A.I. and I agree with you. Actually, I'm glad that you and I have the same thoughts on technology. It is an excuse for studio executives to profit from the work of artists and writers without them knowing about it. I even got involved in this conservation with one of my co-workers when they were talking about A.I. at the same.
Great point about AI not working in the long run but still doing a lot of damage along the way. What's sad is that a lot of creatives will be driven out of business by this technology, but what replaces them will be an inferior deriviative copy of their own work- so in the end we all lose by this- instead of genuine creativity we will be sold cheap fake creativity, and the more AI is used the more everything will start to look the same.
As a visual artist I know that my own work has been used to train AI and I am not happy about it- but having watched the development of AI art for a couple of years now it's clear to me that these machines are hitting some kind of limit- not so much in terms of making pretty pictures but in terms of being able to control the pictures they make. AI Art looks like AI Art- and that was not supposed to happen- and I think the same thing will be true of AI Music- the more we listen to it the more we will start to recognise it as being AI generated.
Instead of enhancing human creativity- which is what the developers claim- Generative AI is doing the opposite- the more people use AI as a 'tool' to create with- the more the shape and design of that tool imposes itself on the things they create. We say 'to a hammer, everything looks like a nail'- well to an AI every prompt looks like a problem to be solved- and it will take the shortest and most efficient way to that solution, because when all is said and done AI is just a computer- it has no creative potential of it's own so the best it can do is try to match the prompt it has been given with a derivative amalgm of the data upon which it was trained.
Generative AI is a cultural dead end which will consume our creativity and then feed it back to us in a degraded and ultimately sterile form.
Thanks, now I just have the image of a massive Star Wars-like factory filled with James Somerton droids in my head. Maybe I could get a generative AI to make that so I can show it to others (I joke, I joke).
Good video. I see no problem promoting a friend's work! As for your video I agree, I was not aware of TH-camrs getting their videos use by AI. This was very helpful to learn and I agree AI is theft!
For me the an important argument was the fact Microsoft didn't train Copilot on their own code, they trained it on others' code. If they really believed what they preach, they'd use it on things with their copyright. They don't.
Theft isn't the right word. It's plagiarism.
Plagiarism is just theft of ideas.
I believe AI entertainment material will be used by those that don't want to pay for royalties. Examples include, a club scene in a TV show where music is being played, the jingle for a radio advertisement or art for an album cover.
I have to say, I agree that "Generative AI" will always devalue any art it touches (whether it's "fully" generated by the AI or if the AI was just used as part of it) and I hope it will never replace a genuine human artist, I'm however worried that it will either A) Devalue art to a point where artists will not receive full pay for art simply because companies will always use AI to either give a framework or to complete what the artist sells them. B) Lead to companies/Artists charging the average consumer extra simply because the art is fully human made, and what I mean is the difference would be more than could be explained simply by the extra cost involved in paying the Artist for their time and art simply charging for the "novelty" of fully human art. While the Artist/distributer can obviously charge whatever they feel like for their art I'd hate to see a world where the method of the arts creation dictates its value rather than the quality of the work and artist. Obviously there's option C where both become a reality and these are just my personal worst case scenarios for this type of AI.
A "Generative AI" COULD be trained ethically and made without it just being a single artist training it on their own art however that would cost the developer behind the AI a lot more time and money to find and pay artists for work to train their AI but even then anything made by the AI would be artistically cheapened by the lack of soul present in the work. You can mimic art but you can't mimic the soul.
and that hypothetical artist won't be able to generate enough paintings to feed the algorithm. If anything, Gen-AI aimed for mass consumption should be trained with works already owned by the companies, but the barrier we need to put will come from a good understanding about what does GenAI exactly does (which was grossly misrepresented here) and new legislation. We will need experts in ethics, philosophers, artists and Gen AI experts to get the best legislation possible. AI won't go anywhere.
My take is - if it's not used for commercial purposes, it's fine to use it. Also if it's used for assistance rather than the end result - also fine.
The best use case that I have which I think would not even require any training on creative work is summarizing (email threads, meeting recordings, etc)
The other half of my take is that they should definitely ask for consent before training it on anyone's creative work (at least that's not in public domain), but it's too late for that :(
I hate it, but I think it's going to stick around in one way or another and using it to assist with certain things is an unfortunate necessity for those of us in the corporate world.
Okay this might be an unpopular opinion, Vera, but I disagree with you on almost all counts. AI will always require human input. There's someone on the other end telling that AI what to generate. "Give me a picture of the ocean at sunset." " Write a story about a person who gets three wishes."
As to the replacement of creatives, that isn't going to happen either. Not immediately and not everyone, anyway. It's akin to being a horse breeder at the beginning of the car age. People still ride horses, but the context is different. Or similar to virtually any manufacturing job that was replaced by automation - there are still humans supervising the process and doing quality control. And if there is air of "get on board or get left behind" about that, it is the unfortunate reality.
And that's not to mention the fact, that, ironically, just like human authors, AI generated content has a signature that makes it obvious. Just like Stephen King was unable to hide behind Richard Bachmann because his writing style was unmistakable, no AI can go undetected. Humans need to come around behind and clean it up.
As to the theft portion, well isn't this how all art has ever been made? Artists take inspiration from one another, add their own spin and create something new. It's just both vast and fast with AI. Also, it's sort of like one baker complaining that his competition also uses flour, yeast, salt and water to make bread. To put it another way, AI can generate content, but it can't generate ideas.
AI is a tool. It's not an end but a means. Like any tool, it can be used for good or for ill.
Thanks for listening and I'll take your downvotes now.
Specious, inane, destructive and outright false. In that order.
I hope you were paid for your efforts. Otherwise, seek help for your self-destructive tendencies.
I can think of one thing Somerton AI can do: You can use it for in between frames in 2d animation. But that me saying that, not the people who are obsessed with it.
God forbid you bring up the fact that AI is built on theft and steals from creators in any pro-AI spaces the backlash and insecurity from tech bros is just something else entirely
I love you!
Hate the way I found you (Neil Gaimen/accusation shenanigans) but I am so so glad I did, I love your content!
It's telling that there's a mass of non-copyrighted material out there, and yet the AI companies specifically go to steal from others. If the public domain isn't enough to train an AI on, then it's not the amazing life-changing kind of tech that they're making out to be.
I mean the thing is, the sheer volume of data required is way beyond what most people realize. They burned through public domain stuff ages ago and the AI still sucks.
@@CouncilofGeeks Exactly! So the only way to do it without stealing would be people willingly giving up their own work... and who would do that? That'd be like turkeys voting for Christmas.
So, they do the morally awful thing instead of, you know, not making AI.
@@CouncilofGeeksthis
Yep, I feel like they're already hitting the point of "diminishing returns" but aren't good enough yet, so they're scrambling for literally any content they can find in the hopes that it'll push them over that last hump into viability.
I made a comment basically agreeing with Vera but it seems that youtube ate it. Weird.
The closest I've heard to your "personal AI is Suzu.jpg's self-AI bot.
As a trained voice actor/vocalist/impressionist, it was already really hard to find jobs, especially those who pay you. Almost all say you need to be grateful your name is associated, I've found it increasingly harder to find work as AI became better and better to the point where I just have to give in because there is no other option. I am desperate for money right now, and AI took away the only thing I could make some with. It wasn't even much mind you, but any little bit helped. Especially after I got long covid 2 years ago and my body's enery levels have gone to shit, voice acting was the only thing I could do and now I can't even do that anymore. At this point, I am willing to anything for any price just to have my name out there again, but the competition is harsh, and AI keeps improving. I'm not one to lose hope, but the future is begging to look bleeker for me by the day.
I was an artist but I have terminal cancer. When I was well I would post everything anywhere I was so excited for each poetry slam and art hop. I'm luckily surrounded by lovely things I was able to do but knowing that only people who don't know me who don't love me will have use of it hurts