I never woke up after my surgery. I'm still in the coma and my family plays youtube videos for me. This is the only possibility.... I can't believe any other option with how stupid people are....
@@bernardthongvanh5613 It would be like if someone took a picture of your bike and put it on a poster they were selling. Which, still isn't a big deal I think this becomes a problem when they start using personal information and faces and stuff like that. I think people are afraid of AI (which they should be) but they don't know what to actually worry about or how it works so they complain about superficial stuff like scraping pictures no one would care if anyone looked at in the first place, because they can't articulate what is actually wrong with any of it but have a gut feeling and understanding that it is dangerous, and it is going to take jobs.
@@nickmagrick7702 tech always takes jobs, that is the whole purpose of tech. that's why there is no ice sellers but fridges now and no horse seller but car sellers, and so on. All we can do is adapt. I don't think any job is safe up untill I retire, that times was centuries ago.
I’m a digital artist and I am beyond furious that our jobs are threatened by deep pocketed companies stealing our collective work. I was told a story of a billionaire once who stole patents, grew his empire, and payed the fines on the patent infringement, which were negligible. I thought this was so smart when I was a kid. Now I realize how disgusting it is. We shouldn’t let people and companies get away with this kind of thing.
@@PrismaticCrafter I am a motion graphics artist. My job is absolutely at risk. Maybe not as in a single LLM is doing everything that I do. But significant portions of my industry tasks are being cut. Lookdev/style frames, character design, vfx element creation, material/footage development, etc. Lots of parts of our creative workflow are being sourced to “AI” and entire jobs are being lost. There have been massive layoffs in our sector over the last year and a half. I’ve been doing this for over 21 years and I’ve never seen anything like it. I see the tools being applied and I know the artists getting sacked. Entire departments of lookdev artists being let go. And it affects me directly. I specifically work in title design so things likes the titles for Secret Invasion actively eat away at my specific industry. Lots of my friends are losing work to automated online AI apps that animate explainer videos. This technology was developed scraping artist’s work and now is replacing them.
If you're actually talented at your work, you can create social media posts on any platform, edit the videos in a satisfying way, and get thousands of people who would want to buy your art. Your job is not being threatened, neither is your creative process. People just want something to complain about, and I don't say that in a patronising way...that is just human nature.
It is executive policy to comply with all current legislation. They take the responsibility to adopt processes that ensure all legislation is complied with. This means taking responsibility for ensuring that all relevant information is flowed both up and down to the relevant people.
@@OctagonalSquare C-suite level should really know this kind of shit though. In short, your CTO 100% should not be making whincy "I'm not sure" faces when people ask technical questions so close to your product.
I love how she keeps saying publicly available or licensed. Publicly available is still subject to copyright law. And most licensed media still comes with restrictions like not creating derivative works for resale.
I think she us referring to licensed data as licensed for a specific purpose (like their deal with shutterstock), not just 'I paid for a subscription to gettyimages'. Although likely the latter is because getty will sue
If you watch something publicly available, it may provide inspiration for you to make transformstive works, which is allowed under copyright law. Sora videos are certainly unique enough to be transformative. Though I imagine as was the case with dalle sora may be able to reproduce fragments of the videos it was trained on. That is something they will no doubt work to address, and obviously the value is in creating novel content not redistributing inherently worse versions of original works.
copyright laws only apply to distribution, once a model is trained, it's completely original and unique with no trace of any source materials, it can't even be considered a derivative work you'd need entirely new laws to make this illegal, as of now it's perfectly legal, they like to keep it vague anyway because of public opinion there is legal precedent of this, machine learning is decades old, it falls under same legal jurisdiction as statistical analysis
@@danieljost5881No, because we're not talking about humans, we're talking about a machine that gets fed data and then compresses that data into noise. Whatever you type, it will then try to restore that noise based on your prompt - which is made of keywords associated to the original data. Given the sheer mass amount of data connected to the same keywords, that's what creates randomness, although overfitting, aka almost perfectly replicating the original data does happen. However, this does not happen with purpose. Because it's a machine and does not think. The image generators are not inspired. They don't know what they're doing, or what anything even is. It's an image generating slot machine. Aside from that, as he pointed out in the video, if you take something and use it to replace/outcompete the original work, it cannot be considered fair use. And all AI shills do all day long is maliciously talk about how they're gonna replace the artists, etc. who the data was taken from.
I don't really see how they're going to be affected. This'd be like a reaction channel being angry that there are other reaction channels. Whatever someone does that's popular, other's will copy it and that's how trends start. It's how music genres start. All the rappers there are today are inspired by the music they heard from other artists, and they're immitating that style. If they were born in the 1800's they wouldn't have came up with the genre themselves. They'd be copying what's popular in that time. An AI can only immitate the exact same way humans immitate. If they'd never heard/seen it, they wouldn't be able to immitate it. Like imagine if we did a real life Truman Show and there was no music in that false reality. The real subject(s) wouldn't start inventing all the different music genre's themself. They wouldn't have anything to go off. Everything is inspired. If any artist has seen further, it's by standing on the shoulders of giants.
I am ok with "not going into detail" ... but "I'm not sure" ???? She is the CTO, how can she not know. That is not a small detail buried inside the code.
@@RavenMobile Does that mean Elon will also be suing himself over the pickup that was supposed to start at 40k, the full self driving that was supposed to be able to drive coast to coast in 2017, the 30k robo taxi that would generate 30k /yr, the rockets that would lower the cost per tonne to orbit, and so, so much more? OpenAI are a bunch of knobs, but elon calling them out is the lump of coal calling the kettle black.
Funnily EU just passed law yesterday that compels AI companies to publish summaries on the training data, so from now on they are compelled to tell what data was used if they wish to operate in the EU region.
Still easy for them pull the "sources say" by stating that they use a mix of data from differing views etc.. when they used .5 seconds of one political source and then just make the rest the side you agree with. Technically you did use data from both sides. But you also didn't. Unless they're forcing them to explain how much they used of each thing.
She basically pivots to "well... it was just lying around so..." The same morality as someone who sees your front door open and just figures "hey, I guess this is a house full of free stuff until someone tells me otherwise".
@@AnataBakka - The book Diamond Age actually talks about this. It shows a basically scarcity-free universe because of nanotech that can decompile and recompile matter - shuffling the elements that make it up around into other forms. Part of that novel shows that there's still a market for artisanal versions of the same things that people can get for free. People talk it up as an ideological thing, or insist that the handmade version is better, and they pay money for it. But the protagonist points out... that's great if you can afford it, but you'd never be able to sustain this business if you had to sell to people who aren't already rich. I dunno, seems relevant. My guess is that generated food would be, like, the unbranded painkillers that cost a tenth of the price of the branded ones, but which a lot of people won't buy even though they're literally paying for a molecule which can't possibly be better just because it's more expensive. Prestige pricing is weird and I don't really get it but I guess some people like the feeling of spending more money than they have to.
'publicly available' only means they didn't outright steal them from hard drives. That's not the blanket protection she wants it to be. I hope a class-action lawsuit is on the horizon.
I think either they get sued by Alphabet or several of the companies that own copyrights to movie trailers and music videos uploaded on RUclips, or there's nothing that can be gained from legal action. It's unlikely that the legal teams of Disney, Universal and Sony are going to miss something that's obvious to a class action lawsuit lawyer. Going to have to wait for a commercial release of this AI video tool to see what happens though.
I taught myself how to fix something via RUclips... Then I fix it for other people.... Now people ask me how did I learn how to do that and where and did I pay for it or not? This is a stupid legal case
It really highlights how much we've already given away to massive corporations. For instance, Reddit made a deal with Google but there was never even a possibility that Reddit would compensate its own users for the same data.
Back in the 1820s, the south did bad stuff, but we should worry more about that the north is industrializing. I don't care if big evil corporation is stealing my paycheck. I only want people who have no money to buy my artwork
@@Kevin_Street I'm not sure if it's still this way, but iirc a few years back social media such as Instagram and Facebook had clauses in their terms of service stating they reserve the ability to use anything you post as they see fit. Honestly, what I'd guess is that OpenAI and other such companies are purchasing data sets that happen to include works that users signed away their rights to. This is the unfortunate price of not reading TOS, or not understanding it. But instead of going after the companies repackaging and selling our data, they're going after the companies making legal purchases.
Even if you're not a content creator, stuff like posting selfies online or family pictures can still be scraped to get those very realistic-looking images. And as someone who, ignorantly, helped with data scraping for a writing AI before they knew how bad things already were, stuff like Wikia pages also got put into data to help teach the AI about fictional characters. Whenever I asked the developers and leaders if they were getting the licensed material legally, like novels, they dodged the question and said it was fine. That really should've been my sign to stop helping, especially as a creative myself. The hype train is very hypnotizing, and either you see the damage the train is doing and get off or you continue to ignore it for the feeling of contributing to Cool Stuff and The Future.
And then how much clean freshwater those datacenters need that gets trashed with toxic chemicals and isn't available for people to drink or use in agriculture
Yes, The Future is important. You’re going to stop technological advancements because they got information from the internet, oh noes. That’s insane. You’re insane. Thank goodness the majority of people working with AI are making them better and better and not abandoning them. I’d fear for the future of the world if everyone felt like this. We’d be back in the stone ages.
They're untouchable at this point. They know it. You said it exactly. They're growing fast while capturing all the market share, all of the users and it's the same thing drug companies do. Your product may cause harm, oh well, the government just has you pay (not even) pennies on the dollars that you've earned. It's a tried and true model at this point. There is no punishment for going fast and loose when you're that size.
They saw it worked with Uber. Uber went around knowingly breaking laws, knowing that if they spend enough on lobbing or get enough market share people will just change the laws for them. They were even caught saying that with Australia. And... it worked.
In classic tech startup fashion open, AI has decided that it's better to ask for forgiveness than to ask for permission. Although Google's lawyers might have something to say about them using their websites, videos for training data
@@klaykid117 what’s wrong with using free, publicly, available videos? As I mentioned before, I’m allowed to train myself on how to make videos by watching videos on RUclips, especially if what I come up with, is my own creation from it.
@@menjolnoyou’ve said this exact same thing twice now and it means nothing. Yes copyright law is fucked, but that doesn’t mean you should be able to just take peoples work and then monetize it. Even if the ai makes a new image or video or whatever none of it would be possible without the people who made the data.
US laws are fuzzy due to fair use rather than an explicit list of what is allowed but generally speaking any jurisdiction with without fair use laws will have laws stating Data Mining and training models on any public data has always been allowed. Common Crawl dataset is from 2008 and google has been training models on all internet data since inception to power search. Copyright as the name implies only really gives exclusive right to create copies but doesn't grant other control over the work. ie. If I wanted to train a model that analyzed how amazon prices items or accuracy of their descriptions, they/sellers have copyright over their webpages but they can't stop me from doing anything I want with their data as long as I am not distributing copies. Perhaps laws should change for generative AI but I am pretty sure they have strong precedent on their side legally with current laws for training on public data.
@@pushpindersingh785 this is a good point and I think that the most frustrating part of the explosive popularity of AI and the visibility into data collection is the lack of pre-existing laws that adequately define what is and is not allowed. People feel taken advantage of for corporate profit. Companies are arguing that they aren't doing anything wrong because there hasn't been a high number of similar legal battles to establish precedence. I'm don't know about you but I'm American and the absolute biggest flaw in this fight between people and LLMs is that Americans think of the Internet as if the US sets the rules and/or that other countries citizens can't break them. Even if our government cracked down on scraping and data mining to the extreme, there is a massive world of Internet that isn't US content. What prevents a US company from building an LLM from any sources outside the US. There aren't world-wide laws and punishments enforced by global police. Just the same, a company outside the US could create an LLM from US hosted/created content and there is nothing we could do. Hell, a US based company could even buy that LLM data set and exploit a loophole. It's a really out of control problem that naive people think can be controlled by a single country or small group. It's unfixable.
or fine them, and forde them to remove that data from the set. as it can't really be done, without regenerate everything, which is expensive too, and needs time. and if they don't comply, double the fine
unironically there's definitely some GPL/MIT in openai's code, most major advances in machine learning are academic and opensource, and even openai themselves started with open source (their name used to mean something)
The point you make in the last two minutes is so crucial. Content creators/users should have a say as to whether or not their data gets used to train AI models. I personally don’t feel comfortable with that, but sites/companies/entities that might have my data may not care, especially if they’re given a fat check or other incentives from OpenAI.
Daily reminder to anyone who needs to hear this: _publicly available_ and _in the public domain_ are not the same thing. Just because it's on the internet, doesn't mean you can copy and redistribute it without permission.
@@JGnLAU8OAWF6they’re directly profiting from the work done by RUclipsrs, using it for a public platform, no credit, no compensation. You can’t just reupload a video and monetize it. Legally it’s close. It’s also scummy as fuck morally. They should ask for the data. Data IS the commodity of the future. It costs money.
@@beekydogg but i can learn from RUclips for free without credit or compensation. Re-uploading is different, they do create NEW videos that didn't exist anywhere. Just like every other person in the world taking inspiration from other content that was before theirs.
They are not redistributing it. They are using the fundamental videos, without even looking at most of it, to train the models. As an IT person who uses Python, I don't care if they are using my videos for this.
It's deeply concerning to think about how user generated content could be being used without explicit consent for training AI. The lack of transparency from AI companies definitely adds to the unease. Seeking clarity on this issue should be a priority for all content creators and users alike!
@codex5080 i read actual legal opinions on the matter, not just some random youtubers’ and angry artists’ opinions. These generative machine learning models do not save the material itself they learn on, they extract mathematical relationships and they save those. That is transformative, because it makes something very different from the video/image/text. If you use an AI tool to “recreate” something it was trained on… well unless you can prove it that it regenerated the “original” by itself and it wasn’t you tweaking it hard… well that is the same as recreating a copyrighted work in any drawing program. The end user who creates the artwork is responsible.
An additional fun revelation to add. So a couple months back i applied for an AI data training position. The application process included multiple video and written content tasks that involved using ones voice, face and reasoning per se. And i mean alot of it. I was abit thrown off but since the company claimed to be GDPR compliant I didn't think much of it. An entire month after the application and automated interview rounds no response. I reach out to the company and they claim they are still going through the applications. Another month goes by and still nothing. I reach out again and ask them to at least inform me if i was not selected, they rephrase their previous response which was very shady. So i logically conclude these guys don't want to end the application process since the application agreement states my personal information, biometric data and additional provided info is to be used by the company for the entire application process. I respond to them asking that my application be withdrawn and to have my entire data redacted as per GDPR regulation. They respond and they then admit that someone else was selected already and the position is no longer available but they express that keeping my profile with them would mean I would be shortlisted for the next opening. I respond and clearly state that i no longer want to have my personal data kept for any upcoming openings and i request that my data be deleted/redacted and to recieve an email confirmation informing me of said compliance. They go radio silent basically refusing to comply to my valid GDPR redaction request. Obviously i could take this to court and hold them accountable. But i am not financially nor mentally prepared for said process as I'm focusing on my job search atm which i am sure they know is the same case for every person applying for said position. Moral of the story is that companies are now posting fake jobs to simply harvest data using the perpetually open application legal loophole. It's amazing how the corporate mindset finds a way to capitalize on our most vulnerable actions.... .
@@ZiggyozI feel for you mate for the public data, companies should be able to use it as making a functional AI is already insanely hard, requires lots of capital, and making profitable is hard. but that should translate to the AI being open source, the company being non profit, or have the AI being public. Open AI was all that, but not forever. they turned into for profit, wont make it open source. the AI will prob not stay public when they burn mircrosofts cash. the lawsuit is justified, we're letting so much unregulated shit going on. but they should be able to use public data if the company makes a compromise. I think even non share holders should value an AI product being properly made. that will give society so much value.
The entire interview was right out of LA Noire. Even after she said "I'm not sure about that" 1:25 you can audibly hear the interviewer "pressing X to doubt!" 😄
Real talk - I'm not sure how useful something like Sora will be for anything but boilerplate corporate videos. The problem I see with gen AI art is that it's a total black box, you only have very abstract control over the process in the form of your prompts so you're kinda stuck making generic slop or inferior copies of actual artists.
@@Colddirector Find me an old buzzfeed article where the images couldn't have been replaced by AI. I'm not an artist or a photographer but I bet most of the money come from people who need "generic slop".
@@Colddirector All the AI models today are too small and aren't even using all of the optimal tech, because it comes out so fast. Claude 3 Opus is by far the best, but it still is small, wasn't trained on 3D environments, and does not have Gemini 1.5s context window. When models have all of these with the kinks worked out and 100 trillion parameters, AI will be more than useful
Thank you so much for covering this, Coffeezilla. We need more voices like you out there! I stand on the side of integrity. And there is nothing about generative AI in it's current state that sells the idea of conducting business with integrity intact.
OpenAI - opensource btw - oh wait no, they turned an open sourced company into a private company and now they charge money. Money for the publicly available data.
OpenAI believes they have the most valuable technology mankind has ever seen. Of course they are going to hold on to that for power and wealth. They probably think they will become gods among men 😅
@@darekmistrz4364 The copyright and licenses relating to the technology she is responsible for as chief technology officer. She's not some standard 9-to-5 sysadmin who makes things work in the server room and takes orders from her boss. She is the boss. She's supposed to be aware of not just the technology itself, but everything surrounding the development, licensing, and logistics of said technology.
@@tatherva7387 I'm not on their side but the way I view it is that she doesn't have to be aware of anything. She is responsible for technology on paper and that is it. If people under her supervision are doing 100% of needed work, there is no need for her to know anything. That is what a good manager does: delegate. Also if they have some "data gathering team" then it's responsibility of that team to respect the law, licensing and copyright. And being in such role in the past, CTO doesn't know anything about laws and you need to go to lawyers for consultation about laws and licensing. Also the interviewer didn't ask the correct question because "Videos on RUclips?" is a huge set of data where there are videos that are in public domain, there are videos where they are published on RUclips license, there are reuploads that are illegal. Same video that is uploaded as public domain on RUclips might be uploaded to Facebok under different license. Some videos might be available outside RUclips on public domain license but on different license on RUclips. You shouldn't invite CTO to a meeting about licensing and law interpretation discuss. She could have been prepared for technical interview where they would talk about technical pieces that are building blocks of OpenAI. Same as not inviting CFO to talk about technology or CTO to talk about finance.
Using copyrighted material is generally legal. Copyright law is a bit crazy and covers lots of things that have nothing to do with copying. But using something is entirely different than copying it. I read copyrighted books, and watch copyrighted videos. I also create things using what i've learned from that. This too, is generally legal. Accuse them of copying if you like, but using is not copying.
i think coffee is a 100% right on what he's saying "recklessly train the AI on every facit of data then take the slap on the wrist then profit" what the punishment should be instead of a financial fine is the deletion of all data revolving around the copyrighted works & if it can't be figured out what is and what isn't copyrighted then the full AI should be deleted
But no one knows if this is a violation of IP. Data mining and analysis qualifies as fair use, which is the same reason that you can't use Google for indexing your side and showing it in search results. It's totally plausible that openai can use the same argument.
@@cyropox8235citation or calling BS. Actually calling BS. Many websites have warnings at the bottom about illegal electronic copying (scraping). And if you go to ChatGPT right now and say ‘create a python routine to scrape data off xyz site’ it will but it will warn you to check to see if it’s legal to use.
So, what I'm hearing, is the new EU laws mean they have to release that information to the public. Sooner or later. Then.... oh wait, copyright law is catching up to them. This is so interesting to watch in real time.
The entire training dataset should be discoverable in any lawsuit against them. Of course, they probably "didn't keep records" of what was used, specifically so it couldn't be discovered, which is both an admission of guilt and a deliberate attempt to frustrate the court's ability to make determinations about copyright. The courts should force them to restart training from scratch, with explicit limits on what can and can't be used and a requirement to document everything.
You should find someone who has access to Sora and prompt it with, "Show me a video of Coffeezilla, or another random RUclipsr, doing jazz hands." If it can reproduce any of them, then Sora committed an insane level of copyright infringement. Not a good look for the Albanian-born CTO raised in a culture of stubborn corruption..
“Data” is doing a lot of the work of deflecting from “a creative work” posted for consumption, but not for training an AI model to reproduce it for commercial purposes. I mean, the Facebook or Instagram or RUclips question is important when you think about what access to someone’s social media profiles would allow these AI models to do … make a digital replica of you saying or doing ANYTHING and without your permission. Radio personalities are already terrified of this, people who do ad reads now and have control over what they endorse could now be forced to endorse anything. These AI companies went from text, to images, to sound, to video so f’ing fast the law hasn’t had time to catch up with the fact that these companies know they are likely to be found breaking the law. And the ramifications are becoming pretty terrifying. Deepfaked porn is arguable just the start.
Great business model these companies have. Steal a bunch of content to train and then make products that will compete. It's like a grocery store that is stocked by products stolen from other grocery stores. Just another way that companies rip off the public.
Fair use. You can use copyrighted materials without permission under certain circumstances. Transformative use is one such thing, and an ai model is transformative.
They already knew they could. No companies making LLMs have faced consequences proportional to the money and user base they can generate from the stolen data
Gotta love one of the other arguments that popped up...the "well if we didn't do it then someone else would". Imagine the bank robbers in court: "well your honour, if we didn't do it then someone else would" 😂
Cop: "Is that you on the CCTV with the bloody knife in your hands?" Perp: "Weeeeell, I'm not going to go into that right now. But if it was, I would only use a knife someone else left lying around."
Here's the main issue of Open AI, no one would care if it was still under the umbrella of being open source and being a free boon for everyone. But the fact that it's a company with a profit motive now means that whatever they do will be fought and hit at with all their might. Mean while smaller AI companies will swoop in and just be more sneaky. What Im saying is that OpenAI will most likely collapse and be reformatted with there being better options like Claude out there. But they will be the shield that other companies hide behind
I'd like to agree with you but in reality the times would also sue them if they were still open source, because they would still be threatening their business using their own data
Apparently there are programs that can completely brick AI’s if they scan an image with a certain filter on it, and I really hope this will be applied to all videos and images in the future to prevent theft.
@@alfred9805 what if I had a tool to scan the bike and make a new one instantly? would that be theft? cause its not the same bike, The bike is the Data I required to scan to get a new bike for myself. Did I need to ask the owner of the bike for permission to scan? Does it matter that I scanned without permission? What if the bike owner was actually a company that sold those bikes, Would it be theft? Im not physically steal or even touching the bike, My tool just scans it and recreates a bike out of thin air
@@G0DKILLER_ I can't tell if this is a joke or not lol but the answer is: Yes, its theft from the company that owns the design and the rights to reproduce the design. It would not be theft from the person with the bike however because they only own that instance.
While we take for granted that any videos uploaded to youtube belong to it's creator, I wonder if youtube doesn't have some ambiguous clause deep in their ToS that they can use to partially 'own' your videos and lend them to these AI companies or to google's own AI.
@@Julez60This is actually even more dangerous, especially since Alphabet has its own competing models. Remember how reddit and Twitter shut down their APIs and effectively killed all third-party clients? Given their war on ad-blockers the last thing we need is for OpenAI to give RUclips a reason.
No thought is truly independent. We read books, watch movies, gain experience at work. We then process this and create new patterns of thought, which can be unique to us but not without inspiration/seed material. I think if the material shouldn’t be used in training data, it shouldn’t be shared at all. Of course, the problem is that we now have a system which industrializes this on a scale and pace never before seen.
@@sdavis7096it’s different than the main channel, not better. It’s better for casual content, but for in depth investigations, this channel is worse, because it doesn’t do that (nothing wrong with that)
It really does. It kind of does things sort of. But those things.... aren't really popular. Go look up "ai video" and look at the view count. So many books written by ai, but... They suck. They're also not accurate (so hard to use them for things that need precision as they're only 80% accurate) So..... how does this "create value". Only thing it seems really good at is grifting.
@@Saliferousno it doesn’t. I’m probably a 5x programmer with chat gpt than without it. That’s value. My company uses it to summarize massive amounts of data that would take our customers way too long to read. That’s value. We use it to filter text related things that are presented to us in multiple languages. It’s actually incredibly valuable but there’s also a lot of bullshit
@@nhanon67asif you're a 5x programmer with ChatGPT how good were you to begin with? Let's forget about the amount of bad code it generates but also the amount of actual vulnerabilities some of the code it spits out produces. Are you a junior developer? Because if you are, a 5x productivity boost makes sense but if you've been doing this for some time? That kind of productivity boost is... YIKES
@@Saliferousthat's not a very accurate take, ai isn't useful to a singular entity without an interest in what ai excels at. I know because I am a trench digging jarhead, I got no use for it, and to expect ai to somehow shovel for me just because some grifter said it would and when it doesn't I call it out as non valuable is just not a realistic way to look at ai, do you get my point if it even made sense?
“the ingestion of copyrighted works to create large language models or other AI training databases generally is a fair use.” - Library Copyright Alliance (LCA) I tend do agree - for the moment. First of all, ML algorithms do not copy pictures, they "learn", they are "transformative". If we do not want to limit artists from learning by analyzing copyrighted materials, we cannot outright stop machines from doing so. Secondly, the products presented by OpenAI are definitely innovative, they contribute to our understanding of machine "intelligence", so they can be considered "research", which is traditional exempted from copyright rules. That said, this technology is going to stay, and become more and more commercial, and less about pushing the barriers of human knowledge. So we still can and should develop mechanisms for rewarding authors of the source materials. In other words: OpenAI is currently an all too easy target for grudges, instead of focusing on the pioneers we need systemic improvements (in law and technology) to reflect the new realities
This is my thesis for my masters right now. Its scary out there. So many interests are invested here. the worst part is once the data is used to train the model you can't just remove it out. It will be so integrated into everything that it would be impossible to track a particular copyrighted content if the original owner wants his data to be removed.
There should be laws making models that doesn't have it's latent space coordinates annotated with licensing/attribution data illegal to use commercially.
I don't understand what you guys are talking about but lawmakers are discussing all this right now and are trying to understand the scope of potential risk before making a comprehensive law
@@ramarajuKannaa essentially everything a generative AI produces consists of different data points in it's latent space being sampled and remixed. If it was annotated and the algorithm collected a list of the influences all resulting works could at the very least be attributable the resulting work correctly or ideally pay a portion of the API fee to the author of the influences. Most of not all AI models are the result of wholesale acquisition of data that has not been suitably licensed. Current implementations do not do this sort of annotation partially to optimize performance, mostly to avoid liability.
I called this out at a company I used to work for, where we were using data but never asked the licensing question. We started to ask the lawyers, turned into a huge "pay us or no" and the project nearly got canned. That was several years ago before chatgpt, and when that came out, I shuddered. This whole thing is trending towards a disaster.
I just re-watched Terminator 2 last night for the first time since before all this ai sh** started going into turbo mode……… the parallels between Skynet and open ai is straight up unsettling……
That's because you have no idea what AI is. It's just programming. Basically it's not than different than math. If you look up how math evolves, there are questions that are fundamental that take centuries to solve. Same will happen with computer science and AI.
When are people going to learn: interviewers are like lawyers. They usually do not ask questions they either A. don't know the answer to or B. know you can only answer in a way they want you to answer. Either way, just say "no" to the interview lol
Even if Shutterstock did license the images and videos to openAI, that doesn’t mean that the original copyright owner gave permission or benefitted from the deal. Given that these models will completely and permanently destroy the market for real images and videos, making those original images essentially worthless, those copyright holders should have expected vast payoffs for the use of their material.
Nonprofits AI outfits might have an argument for Fair Use of content in training LLMs (research purposes), but when they're developing AI for commercial use I think the letter of the law and jurisprudence are pretty clear that licensing fees need to be paid to the copyright holder(s). It's probably an issue of creators not knowing about their work being infringed (which lots of these AI companies try to avoid through opacity and NDAs), but that's neither a responsible nor a sustainable business model.
She's a manager. She might have worked as a computer programmer or software engineer at some point but for years she's been an administrator. She really doesn't know how new technology works.
This is why we could've used Andrew Yang as a president back in 2020. A major platform of his was to codify a law to require user consent and users getting a cut off of procured, sold, and resold data into a UBI fund, along with VAT on discretionary, non-essential consumer goods like luxury bags, yachts, and first class flights.
Im not a behavioural professional but when she's answering the question at 1:40 she's shaking her own head while talking. @TheBehaviorPanel please cover this! Would love your informed opinions! @Voidzilla, it seems that OpenAI are taking the Uber rollout approach. Get the technology and infrastructure out and then deal with repercussions later. They are doing this to get funding now which will go toward legal fees.
I'm an artist, and these videos are so funny to me. Its funny because it took AI to suddenly threaten to take RUclipsr's jobs specifically for them to start making the same realizations we did a year ago. The industry standard for AI development has been theft and plagiarism from the start, this is old news. Not saying youtubers don't care about other people's wellbeing, its just interesting how it takes things getting personal before people start paying attention to this stuff.
Thanks for sharing that video. I highly recommend also the Adam Conover interview with Gary Marcus, Emily Bender and Timnit Gebru as additional context on the genAI matter
Not even half way through your video and you have the most accurate understanding of the fair use doctrine and what transformative actually means under the law that I’ve seen outside of a law school lecture. Sora and other A*I is exploiting the copyrighted work of others, there is no change of character or purpose in the end product, the models sample the entirety of the original copyrighted work, and it lowers the market value of the original work, and these works are typically not factual in nature. It doesn’t fit the carveouts such as commentary, journalism, satire, or education in a classroom. Further, these companies are going around like knobs stating it’s fair use like it’s a fact when fair use is an affirmative defense in copyright infringement litigation. It’s not something someone can just declare and it is so, it’s something that is decided in court.
As a tech demo I don't think it should be required to acquire license to develop the framework and fine tune how the model acquire and understand data. Like testing a blender's blade. But the moment it becomes in-use for commercial/public use then absolutely the entirety of the library used should be licensed.
I think it's too much of a leap to trust companies to keep track of everyone they've used content from without permission to retroactively compensate them once they decide to make money off the tech.
I think the problem for me is that whatever they trained sora on, they aren't going to spend money to retrain it. Beyond that, sora is a tech demo. They profited by showing that to investors.
Soon it will be a moot issue really because AI will get so powerful it will just go look for training data itself and if it wanted to could obfuscate it enough even the owners wouldn't know where it got it from.
Here is also another problem, the video might be free when it was used for training but it was copyrighted afterwards. This kind of stuff can also happen. Copyright laws need a lot of research and restructuring for Internet only. For AI its a whole another ballgame
It's not a controversy, It's just a wake up topic that we need to get on top of. Every account we set up has an eula we "sign". The courts just have to sort it all out at taxpayer expense. We abysssin'
Tumblr and Wordpress are planning on selling their data to OpenAI & MidJourney (and probably already did handover a good bit of it based on different sources I've read). As a user of both platforms (and one who's made significant amounts of content for them with the understanding that I own the rights to it) AND an English teacher, I'm not super thrilled with the idea that my writing is being used to train AI's to replace me or help my students cheat.
When you buy anything with a credit card the credit card company knows what you bought, how much you paid, your location and, well, they know pretty much everything about you. The credit card companies then sell that private data and the customer, the one that actual purchased the item, does get even a tiny slice of the money the companies make selling your data. That is, the default position is that your data is their data and you are entitled to precisely nothing of that data's value. So, it was logical for these AI companies to buy or steal your data without telling you or compensating you -- AND, after training the AI to replicate your particular creative ability they can compete against you with your own IP. If the DOJ and SCOTUS drop the ball on this then we are screwed!
This is theft. These artist will be out of a job because of these companies. I can't even put a 30 second clip of a movie for commentary without getting a DMCA claim. Yet when the elights take our work without our consent to literally replace us it okay. We shouldn't stand for this #boycott_openAI.
@@TreeLBollingTreeManThere's just all kinds of technical legal drama when it comes to this stuff no matter how many disclaimers he places unfortunately
She pulled the LA noire meme face. I didn't think that was possible but she did it. Bravo.
Yeah the mocapped footage isnt possible
Lmao LA Noire face is a funny description 😂
My first thought😂
😂😂😂😂
I never woke up after my surgery. I'm still in the coma and my family plays youtube videos for me. This is the only possibility.... I can't believe any other option with how stupid people are....
“Sir, have you stolen any art?”
“It was in a public museum”
Every tourist who has ever taken a picture at an art museum: Uh oh....
@@olivercharles2930 They didn't make any money out of their pictures, and probably didn't use them for their own works.
It's easier to ask for forgiveness than permission.
Especially when you stand to make hundreds of millions of dollars off the backs of other people.
Well put
yeah lol I learned that from drake in high school. for small stuff its usually how I move
Wahey, it's Strat!
Same applies to anal
"Did you steal my bicycle?"
-"It was publicly available."
not the same thing
😂😂😂😂😂😂
a honest comparaison would be "did you look at my bike?", "well it was exposed in a place meant to look at things" -_-
@@bernardthongvanh5613 It would be like if someone took a picture of your bike and put it on a poster they were selling. Which, still isn't a big deal
I think this becomes a problem when they start using personal information and faces and stuff like that.
I think people are afraid of AI (which they should be) but they don't know what to actually worry about or how it works so they complain about superficial stuff like scraping pictures no one would care if anyone looked at in the first place, because they can't articulate what is actually wrong with any of it but have a gut feeling and understanding that it is dangerous, and it is going to take jobs.
@@nickmagrick7702 tech always takes jobs, that is the whole purpose of tech.
that's why there is no ice sellers but fridges now and no horse seller but car sellers, and so on.
All we can do is adapt. I don't think any job is safe up untill I retire, that times was centuries ago.
I’m a digital artist and I am beyond furious that our jobs are threatened by deep pocketed companies stealing our collective work.
I was told a story of a billionaire once who stole patents, grew his empire, and payed the fines on the patent infringement, which were negligible. I thought this was so smart when I was a kid. Now I realize how disgusting it is.
We shouldn’t let people and companies get away with this kind of thing.
Same here. It’s absolutely infuriating.
@@PrismaticCrafter I am a motion graphics artist. My job is absolutely at risk. Maybe not as in a single LLM is doing everything that I do. But significant portions of my industry tasks are being cut.
Lookdev/style frames, character design, vfx element creation, material/footage development, etc. Lots of parts of our creative workflow are being sourced to “AI” and entire jobs are being lost. There have been massive layoffs in our sector over the last year and a half. I’ve been doing this for over 21 years and I’ve never seen anything like it. I see the tools being applied and I know the artists getting sacked. Entire departments of lookdev artists being let go. And it affects me directly.
I specifically work in title design so things likes the titles for Secret Invasion actively eat away at my specific industry. Lots of my friends are losing work to automated online AI apps that animate explainer videos.
This technology was developed scraping artist’s work and now is replacing them.
@@PrismaticCrafter Complete bs.
If you're actually talented at your work, you can create social media posts on any platform, edit the videos in a satisfying way, and get thousands of people who would want to buy your art. Your job is not being threatened, neither is your creative process. People just want something to complain about, and I don't say that in a patronising way...that is just human nature.
Your art is ass… so why are you even mad when you don’t get clients in the first place
I like how she panics for just a second, thinks about the camera on her, and quickly regains composure. She's definitely not media trained.
she would fit well in politics then
@JG-MV yes, that comment was sarcastic
Shes at least relatable 😅
She’s an AI researcher not so much a person for the cameras
you mean, not trained to be a good lier or keep her mouth shut
Does she not understand that "publicly available" is not the same thing as public domain. Surely it's her job to know this stuff?
Technically it would be their legal team’s job, not the Chief Technology Officer
It is executive policy to comply with all current legislation. They take the responsibility to adopt processes that ensure all legislation is complied with. This means taking responsibility for ensuring that all relevant information is flowed both up and down to the relevant people.
@@OctagonalSquare C-suite level should really know this kind of shit though. In short, your CTO 100% should not be making whincy "I'm not sure" faces when people ask technical questions so close to your product.
They have already stated they don’t care about licensing
It is literally her job to lie about it and pay the smallest fine. Cost of doing business.
I love how she keeps saying publicly available or licensed. Publicly available is still subject to copyright law. And most licensed media still comes with restrictions like not creating derivative works for resale.
I think she us referring to licensed data as licensed for a specific purpose (like their deal with shutterstock), not just 'I paid for a subscription to gettyimages'. Although likely the latter is because getty will sue
If you watch something publicly available, it may provide inspiration for you to make transformstive works, which is allowed under copyright law. Sora videos are certainly unique enough to be transformative.
Though I imagine as was the case with dalle sora may be able to reproduce fragments of the videos it was trained on. That is something they will no doubt work to address, and obviously the value is in creating novel content not redistributing inherently worse versions of original works.
copyright laws only apply to distribution, once a model is trained, it's completely original and unique with no trace of any source materials, it can't even be considered a derivative work
you'd need entirely new laws to make this illegal, as of now it's perfectly legal, they like to keep it vague anyway because of public opinion
there is legal precedent of this, machine learning is decades old, it falls under same legal jurisdiction as statistical analysis
would you write the same thing, if your channel got DMCA?
@@danieljost5881No, because we're not talking about humans, we're talking about a machine that gets fed data and then compresses that data into noise. Whatever you type, it will then try to restore that noise based on your prompt - which is made of keywords associated to the original data. Given the sheer mass amount of data connected to the same keywords, that's what creates randomness, although overfitting, aka almost perfectly replicating the original data does happen. However, this does not happen with purpose. Because it's a machine and does not think. The image generators are not inspired. They don't know what they're doing, or what anything even is. It's an image generating slot machine. Aside from that, as he pointed out in the video, if you take something and use it to replace/outcompete the original work, it cannot be considered fair use. And all AI shills do all day long is maliciously talk about how they're gonna replace the artists, etc. who the data was taken from.
Your honor, his lawn gnome was publicly available.
lmfao pretty much bro. let me just go screenrecrod or stream a movie from max real quick I mean its public
Video creators, welcome to the world that photographers, artists, and writers have been living in for a couple of years now.
You can add programmers to that list too.
I don't really see how they're going to be affected. This'd be like a reaction channel being angry that there are other reaction channels. Whatever someone does that's popular, other's will copy it and that's how trends start. It's how music genres start. All the rappers there are today are inspired by the music they heard from other artists, and they're immitating that style. If they were born in the 1800's they wouldn't have came up with the genre themselves. They'd be copying what's popular in that time.
An AI can only immitate the exact same way humans immitate. If they'd never heard/seen it, they wouldn't be able to immitate it. Like imagine if we did a real life Truman Show and there was no music in that false reality. The real subject(s) wouldn't start inventing all the different music genre's themself. They wouldn't have anything to go off.
Everything is inspired. If any artist has seen further, it's by standing on the shoulders of giants.
@@DlcEnergy you forgot the fact that its stealing from licensed and or copyrighted materials and then GAINING profit from it.
@@genshinF2Play and people inspired by what they see don’t imitate and do the same?
@@genshinF2Play if watching and being inspired by others and wanting to do what they do for a living is stealing, we’d still be in the caves.
"im not sure" "im not going into detail" wew ...... such transparency
Very Open(AI) of them
@@JonathonCwik Elon is trying to sue them because it was supposed to be open source, but then they got rid of all that to profit more.
I am ok with "not going into detail" ... but "I'm not sure" ???? She is the CTO, how can she not know. That is not a small detail buried inside the code.
@@RavenMobileyea hope he wins
@@RavenMobile
Does that mean Elon will also be suing himself over the pickup that was supposed to start at 40k, the full self driving that was supposed to be able to drive coast to coast in 2017, the 30k robo taxi that would generate 30k /yr, the rockets that would lower the cost per tonne to orbit, and so, so much more?
OpenAI are a bunch of knobs, but elon calling them out is the lump of coal calling the kettle black.
Funnily EU just passed law yesterday that compels AI companies to publish summaries on the training data, so from now on they are compelled to tell what data was used if they wish to operate in the EU region.
EU far ahead of the curve when it comes to attempting to control sociopathic company practices.
@@tdx_1138especially after already monopolising usb-c, it's a life changer.
Good.
@@tdx_1138weird how you rely on sociopathic government to combat corporations you call sociopathic.
Still easy for them pull the "sources say" by stating that they use a mix of data from differing views etc.. when they used .5 seconds of one political source and then just make the rest the side you agree with. Technically you did use data from both sides. But you also didn't. Unless they're forcing them to explain how much they used of each thing.
She basically pivots to "well... it was just lying around so..."
The same morality as someone who sees your front door open and just figures "hey, I guess this is a house full of free stuff until someone tells me otherwise".
in the future, if we will be able to generate food at no cost, people will still try to make a profit from it
At this rate they'll start collecting anything that isn't bolted down!
@@AnataBakka - The book Diamond Age actually talks about this. It shows a basically scarcity-free universe because of nanotech that can decompile and recompile matter - shuffling the elements that make it up around into other forms.
Part of that novel shows that there's still a market for artisanal versions of the same things that people can get for free. People talk it up as an ideological thing, or insist that the handmade version is better, and they pay money for it. But the protagonist points out... that's great if you can afford it, but you'd never be able to sustain this business if you had to sell to people who aren't already rich.
I dunno, seems relevant. My guess is that generated food would be, like, the unbranded painkillers that cost a tenth of the price of the branded ones, but which a lot of people won't buy even though they're literally paying for a molecule which can't possibly be better just because it's more expensive. Prestige pricing is weird and I don't really get it but I guess some people like the feeling of spending more money than they have to.
'You will own nothing and you will be happy', just took another step.
LMAO "Hey guys is it okay if I take these RUclips videos you left back here...."
"I think that's a yes"
Scraping is the terminology used to avoid the true word "Stealing".
Man, imagine having to educate a CTO that "publicly available" does not mean "free for the taking"
They know. It's their job to decide if the risk of punishment is worth it. Low risk by the way - companies NEVER get punished.
'publicly available' only means they didn't outright steal them from hard drives. That's not the blanket protection she wants it to be. I hope a class-action lawsuit is on the horizon.
Realistically the big money lies with OpenAI so they will more than likely come out on top.
But these class actions lawsuits never change anything really
I think either they get sued by Alphabet or several of the companies that own copyrights to movie trailers and music videos uploaded on RUclips, or there's nothing that can be gained from legal action. It's unlikely that the legal teams of Disney, Universal and Sony are going to miss something that's obvious to a class action lawsuit lawyer. Going to have to wait for a commercial release of this AI video tool to see what happens though.
RUclips and creators can make the copyright claim.
I taught myself how to fix something via RUclips...
Then I fix it for other people....
Now people ask me how did I learn how to do that and where and did I pay for it or not?
This is a stupid legal case
The connection between data privacy and artists IP ownership seems like it's an important topic right now.
Always has been but only started being questioned like two years ago when Gen ai started to become popular
It really highlights how much we've already given away to massive corporations. For instance, Reddit made a deal with Google but there was never even a possibility that Reddit would compensate its own users for the same data.
Back in the 1820s, the south did bad stuff, but we should worry more about that the north is industrializing. I don't care if big evil corporation is stealing my paycheck. I only want people who have no money to buy my artwork
4
@@Kevin_Street I'm not sure if it's still this way, but iirc a few years back social media such as Instagram and Facebook had clauses in their terms of service stating they reserve the ability to use anything you post as they see fit.
Honestly, what I'd guess is that OpenAI and other such companies are purchasing data sets that happen to include works that users signed away their rights to. This is the unfortunate price of not reading TOS, or not understanding it. But instead of going after the companies repackaging and selling our data, they're going after the companies making legal purchases.
"VoidZilla:Every day we sink further into the Black Abyss."
WHERE IS MY DAMN T-SHIRT Coffee!
It needs to have a top-down view of a big mug of black coffee, as well.
I'd buy that T-shirt
I'm still gunning for "The Void Is Super Petty" T-shirt
In the void I guess.
Void merch would be awesome!
Even if you're not a content creator, stuff like posting selfies online or family pictures can still be scraped to get those very realistic-looking images.
And as someone who, ignorantly, helped with data scraping for a writing AI before they knew how bad things already were, stuff like Wikia pages also got put into data to help teach the AI about fictional characters. Whenever I asked the developers and leaders if they were getting the licensed material legally, like novels, they dodged the question and said it was fine.
That really should've been my sign to stop helping, especially as a creative myself. The hype train is very hypnotizing, and either you see the damage the train is doing and get off or you continue to ignore it for the feeling of contributing to Cool Stuff and The Future.
And then how much clean freshwater those datacenters need that gets trashed with toxic chemicals and isn't available for people to drink or use in agriculture
Yes, The Future is important. You’re going to stop technological advancements because they got information from the internet, oh noes. That’s insane. You’re insane. Thank goodness the majority of people working with AI are making them better and better and not abandoning them. I’d fear for the future of the world if everyone felt like this. We’d be back in the stone ages.
@@bakedandbeaded - Important yet bleaker than you can ever imagine.
They're untouchable at this point. They know it. You said it exactly. They're growing fast while capturing all the market share, all of the users and it's the same thing drug companies do. Your product may cause harm, oh well, the government just has you pay (not even) pennies on the dollars that you've earned. It's a tried and true model at this point. There is no punishment for going fast and loose when you're that size.
They saw it worked with Uber. Uber went around knowingly breaking laws, knowing that if they spend enough on lobbing or get enough market share people will just change the laws for them. They were even caught saying that with Australia.
And... it worked.
“I’m not sure about that…”
Ah… so… your lawyers trained you on legal plausible deniability…
More like deniable plausibility. :P
In classic tech startup fashion open, AI has decided that it's better to ask for forgiveness than to ask for permission. Although Google's lawyers might have something to say about them using their websites, videos for training data
@@klaykid117 what’s wrong with using free, publicly, available videos? As I mentioned before, I’m allowed to train myself on how to make videos by watching videos on RUclips, especially if what I come up with, is my own creation from it.
@@codycastyou can use it for personal use but if you want to make money reselling that info you cannot take it for free.
@@codycast exactly, youtubers are just mad they are going to be useless - along with 99% of the population - when they though they were untouchable.
Publicly available content does not mean there is no requirement for permission because the US has implied copyright when it is unique to the creator.
would someone write the same thing, if they got dmca?
@@menjolnoyou’ve said this exact same thing twice now and it means nothing. Yes copyright law is fucked, but that doesn’t mean you should be able to just take peoples work and then monetize it. Even if the ai makes a new image or video or whatever none of it would be possible without the people who made the data.
Also what constitutes publicly available? Could be a show on ABC? That's available to the public the same way social media posts are.
US laws are fuzzy due to fair use rather than an explicit list of what is allowed but generally speaking any jurisdiction with without fair use laws will have laws stating Data Mining and training models on any public data has always been allowed. Common Crawl dataset is from 2008 and google has been training models on all internet data since inception to power search.
Copyright as the name implies only really gives exclusive right to create copies but doesn't grant other control over the work. ie. If I wanted to train a model that analyzed how amazon prices items or accuracy of their descriptions, they/sellers have copyright over their webpages but they can't stop me from doing anything I want with their data as long as I am not distributing copies.
Perhaps laws should change for generative AI but I am pretty sure they have strong precedent on their side legally with current laws for training on public data.
@@pushpindersingh785 this is a good point and I think that the most frustrating part of the explosive popularity of AI and the visibility into data collection is the lack of pre-existing laws that adequately define what is and is not allowed. People feel taken advantage of for corporate profit. Companies are arguing that they aren't doing anything wrong because there hasn't been a high number of similar legal battles to establish precedence. I'm don't know about you but I'm American and the absolute biggest flaw in this fight between people and LLMs is that Americans think of the Internet as if the US sets the rules and/or that other countries citizens can't break them. Even if our government cracked down on scraping and data mining to the extreme, there is a massive world of Internet that isn't US content. What prevents a US company from building an LLM from any sources outside the US. There aren't world-wide laws and punishments enforced by global police. Just the same, a company outside the US could create an LLM from US hosted/created content and there is nothing we could do. Hell, a US based company could even buy that LLM data set and exploit a loophole. It's a really out of control problem that naive people think can be controlled by a single country or small group. It's unfixable.
Don't make them pay - make them make the model open source since it was trained on forced open source data. That'll ruin them.
or fine them, and forde them to remove that data from the set. as it can't really be done, without regenerate everything, which is expensive too, and needs time. and if they don't comply, double the fine
@@schwingedeshaehersfining never works. Next
unironically there's definitely some GPL/MIT in openai's code, most major advances in machine learning are academic and opensource, and even openai themselves started with open source (their name used to mean something)
@@InMyBunker with exponential fines? and it can work, look at apple and EU (and some other comanys and the EU)
@@InMyBunker also, did you read my message? it wasn't said, that the fine is the only punishment
The point you make in the last two minutes is so crucial. Content creators/users should have a say as to whether or not their data gets used to train AI models.
I personally don’t feel comfortable with that, but sites/companies/entities that might have my data may not care, especially if they’re given a fat check or other incentives from OpenAI.
Daily reminder to anyone who needs to hear this: _publicly available_ and _in the public domain_ are not the same thing. Just because it's on the internet, doesn't mean you can copy and redistribute it without permission.
They don't copy and distribute it though. Humans can watch those videos to "train" why can't computers?
@@JGnLAU8OAWF6they’re directly profiting from the work done by RUclipsrs, using it for a public platform, no credit, no compensation. You can’t just reupload a video and monetize it. Legally it’s close. It’s also scummy as fuck morally. They should ask for the data. Data IS the commodity of the future. It costs money.
@@beekydogg but i can learn from RUclips for free without credit or compensation. Re-uploading is different, they do create NEW videos that didn't exist anywhere. Just like every other person in the world taking inspiration from other content that was before theirs.
They are not redistributing it. They are using the fundamental videos, without even looking at most of it, to train the models. As an IT person who uses Python, I don't care if they are using my videos for this.
@beekydogg nobody People learn from RUclips for free everyday, Why should an AI model have to pay?
It's deeply concerning to think about how user generated content could be being used without explicit consent for training AI. The lack of transparency from AI companies definitely adds to the unease. Seeking clarity on this issue should be a priority for all content creators and users alike!
Transformative use of copyrighted material is allowed without permission under the fair use exceptions.
@codex5080 i read actual legal opinions on the matter, not just some random youtubers’ and angry artists’ opinions.
These generative machine learning models do not save the material itself they learn on, they extract mathematical relationships and they save those. That is transformative, because it makes something very different from the video/image/text.
If you use an AI tool to “recreate” something it was trained on… well unless you can prove it that it regenerated the “original” by itself and it wasn’t you tweaking it hard… well that is the same as recreating a copyrighted work in any drawing program. The end user who creates the artwork is responsible.
An additional fun revelation to add.
So a couple months back i applied for an AI data training position. The application process included multiple video and written content tasks that involved using ones voice, face and reasoning per se. And i mean alot of it.
I was abit thrown off but since the company claimed to be GDPR compliant I didn't think much of it.
An entire month after the application and automated interview rounds no response. I reach out to the company and they claim they are still going through the applications. Another month goes by and still nothing. I reach out again and ask them to at least inform me if i was not selected, they rephrase their previous response which was very shady. So i logically conclude these guys don't want to end the application process since the application agreement states my personal information, biometric data and additional provided info is to be used by the company for the entire application process.
I respond to them asking that my application be withdrawn and to have my entire data redacted as per GDPR regulation. They respond and they then admit that someone else was selected already and the position is no longer available but they express that keeping my profile with them would mean I would be shortlisted for the next opening.
I respond and clearly state that i no longer want to have my personal data kept for any upcoming openings and i request that my data be deleted/redacted and to recieve an email confirmation informing me of said compliance.
They go radio silent basically refusing to comply to my valid GDPR redaction request.
Obviously i could take this to court and hold them accountable. But i am not financially nor mentally prepared for said process as I'm focusing on my job search atm which i am sure they know is the same case for every person applying for said position.
Moral of the story is that companies are now posting fake jobs to simply harvest data using the perpetually open application legal loophole.
It's amazing how the corporate mindset finds a way to capitalize on our most vulnerable actions....
.
The Hollywood writers strike seemed to end favourably.. hopefully if shit goes wonky for creators, they can be afforded similar legal protections.
@@ZiggyozI feel for you mate
for the public data, companies should be able to use it as making a functional AI is already insanely hard, requires lots of capital, and making profitable is hard.
but that should translate to the AI being open source, the company being non profit, or have the AI being public. Open AI was all that, but not forever. they turned into for profit, wont make it open source. the AI will prob not stay public when they burn mircrosofts cash.
the lawsuit is justified, we're letting so much unregulated shit going on. but they should be able to use public data if the company makes a compromise. I think even non share holders should value an AI product being properly made. that will give society so much value.
The entire interview was right out of LA Noire. Even after she said "I'm not sure about that" 1:25 you can audibly hear the interviewer "pressing X to doubt!" 😄
AI cannot possibly comprehend the majesty that is the plain, black background. You're safe, Coffee.
I love that the coffeezilla A.I. doesn't lie about having permission,
It straight up just lies and says " I am the real coffeezilla" lol
Real talk - I'm not sure how useful something like Sora will be for anything but boilerplate corporate videos. The problem I see with gen AI art is that it's a total black box, you only have very abstract control over the process in the form of your prompts so you're kinda stuck making generic slop or inferior copies of actual artists.
@@Colddirector Find me an old buzzfeed article where the images couldn't have been replaced by AI. I'm not an artist or a photographer but I bet most of the money come from people who need "generic slop".
@@Colddirector All the AI models today are too small and aren't even using all of the optimal tech, because it comes out so fast. Claude 3 Opus is by far the best, but it still is small, wasn't trained on 3D environments, and does not have Gemini 1.5s context window. When models have all of these with the kinks worked out and 100 trillion parameters, AI will be more than useful
@@tapwater424 that’s kind of the point - buzzfeed was a slop site
You can't steal from the void. You can only scream into it.
Thank you so much for covering this, Coffeezilla. We need more voices like you out there! I stand on the side of integrity. And there is nothing about generative AI in it's current state that sells the idea of conducting business with integrity intact.
OpenAI - opensource btw - oh wait no, they turned an open sourced company into a private company and now they charge money. Money for the publicly available data.
Forgetting to rebrand themselves as ClosedAI 😂
OpenAI believes they have the most valuable technology mankind has ever seen. Of course they are going to hold on to that for power and wealth. They probably think they will become gods among men 😅
Ask the Chief Technology Officer the most basic question about the technology that's been created: "Eeuuuurrrrggghh I'm not sure about that."
Reminds me of when Coffee interviewed SBF.
What is technogolical about copyright laws and licenses?
@@darekmistrz4364 The copyright and licenses relating to the technology she is responsible for as chief technology officer. She's not some standard 9-to-5 sysadmin who makes things work in the server room and takes orders from her boss. She is the boss. She's supposed to be aware of not just the technology itself, but everything surrounding the development, licensing, and logistics of said technology.
@@tatherva7387 I'm not on their side but the way I view it is that she doesn't have to be aware of anything. She is responsible for technology on paper and that is it. If people under her supervision are doing 100% of needed work, there is no need for her to know anything. That is what a good manager does: delegate. Also if they have some "data gathering team" then it's responsibility of that team to respect the law, licensing and copyright. And being in such role in the past, CTO doesn't know anything about laws and you need to go to lawyers for consultation about laws and licensing.
Also the interviewer didn't ask the correct question because "Videos on RUclips?" is a huge set of data where there are videos that are in public domain, there are videos where they are published on RUclips license, there are reuploads that are illegal. Same video that is uploaded as public domain on RUclips might be uploaded to Facebok under different license. Some videos might be available outside RUclips on public domain license but on different license on RUclips.
You shouldn't invite CTO to a meeting about licensing and law interpretation discuss. She could have been prepared for technical interview where they would talk about technical pieces that are building blocks of OpenAI. Same as not inviting CFO to talk about technology or CTO to talk about finance.
"I'm sorry I cannot answer that"
She made the face of the guy who said "That's all that was took!" in LA Noir
they was workin on the tires!
LMAO
“You f young boys Mira?” 😂
Publicly available IS NOT public domain. They are totally hiding the fact that they are using copyrighted material.
Using copyrighted material is generally legal. Copyright law is a bit crazy and covers lots of things that have nothing to do with copying. But using something is entirely different than copying it. I read copyrighted books, and watch copyrighted videos. I also create things using what i've learned from that. This too, is generally legal. Accuse them of copying if you like, but using is not copying.
i think coffee is a 100% right on what he's saying "recklessly train the AI on every facit of data then take the slap on the wrist then profit"
what the punishment should be instead of a financial fine is the deletion of all data revolving around the copyrighted works & if it can't be figured out what is and what isn't copyrighted then the full AI should be deleted
LOL! That face! Just like my daughter when i literally caught her with her hand in the cookie jar! She ain't got no defence for that.
You mean those cookies weren’t publicly available?? 😅
@@Nick-cs4oc and claims to be licensed... 🤣
Hahaha thanks for the laugh
@@guusgeluk3693 glad to be of service 😊👍
Let your kids eat what they want though.
Uh, lady, intellectual property might be "publically available" but that doesn't mean they're not still protected property under copyright law.
But no one knows if this is a violation of IP. Data mining and analysis qualifies as fair use, which is the same reason that you can't use Google for indexing your side and showing it in search results. It's totally plausible that openai can use the same argument.
@@cyropox8235citation or calling BS. Actually calling BS. Many websites have warnings at the bottom about illegal electronic copying (scraping). And if you go to ChatGPT right now and say ‘create a python routine to scrape data off xyz site’ it will but it will warn you to check to see if it’s legal to use.
Copyright prohibits reproducing a work. It doesn’t prohibit analyzing a work.
So, what I'm hearing, is the new EU laws mean they have to release that information to the public. Sooner or later. Then.... oh wait, copyright law is catching up to them.
This is so interesting to watch in real time.
Correct. Transparency is required.
thats the irony thou, theres no such thing as transparency in AI tecnology other than replacing work regardless of ethics or laws.@@byrnemeister2008
Yes. OpenAI is going to see a mountain of lawsuits.
The EU seems to be taking the lead whenever modern, digital ownership issues pop up.
that is because the justice system isnt dominated by senile old farts and isnt made for only two political parties.@@Chas-OTE
The entire training dataset should be discoverable in any lawsuit against them. Of course, they probably "didn't keep records" of what was used, specifically so it couldn't be discovered, which is both an admission of guilt and a deliberate attempt to frustrate the court's ability to make determinations about copyright. The courts should force them to restart training from scratch, with explicit limits on what can and can't be used and a requirement to document everything.
You should find someone who has access to Sora and prompt it with, "Show me a video of Coffeezilla, or another random RUclipsr, doing jazz hands." If it can reproduce any of them, then Sora committed an insane level of copyright infringement. Not a good look for the Albanian-born CTO raised in a culture of stubborn corruption..
When you get the 'I just bit into a lime' face, that was the moment she realized... they fu**ed up.
Interviewer asking the real questions for once and we got this banger 😄 AI companies abandoned all ethics to race for as much money as possible.
Being fairly compensated for my original work is my love language
Sounds like communism to me
how dare you actually get paid for your work you should just give it away for free to the big conglomerates.
Why should you get paid for fair use?
User data is the new gold of the 21st century
“Data” is doing a lot of the work of deflecting from “a creative work” posted for consumption, but not for training an AI model to reproduce it for commercial purposes. I mean, the Facebook or Instagram or RUclips question is important when you think about what access to someone’s social media profiles would allow these AI models to do … make a digital replica of you saying or doing ANYTHING and without your permission. Radio personalities are already terrified of this, people who do ad reads now and have control over what they endorse could now be forced to endorse anything.
These AI companies went from text, to images, to sound, to video so f’ing fast the law hasn’t had time to catch up with the fact that these companies know they are likely to be found breaking the law. And the ramifications are becoming pretty terrifying. Deepfaked porn is arguable just the start.
These AI companies getting an object lesson that "publicly available" is not interchangeable with open source, public domain, or copyright free.
Great business model these companies have. Steal a bunch of content to train and then make products that will compete. It's like a grocery store that is stocked by products stolen from other grocery stores.
Just another way that companies rip off the public.
Fair use. You can use copyrighted materials without permission under certain circumstances. Transformative use is one such thing, and an ai model is transformative.
@@realGBx64creating derivatives of copyrighted work is different.
@@GeometricPidgeon derivatives are specifically allowed under specific conditions
@@realGBx64Coffee covered the fair use argument in this very video
They were so preoccupied with whether or not they could that they didn't think about whether or not they should
Jurassic park 👍
Technology 👍🏾
They already knew they could. No companies making LLMs have faced consequences proportional to the money and user base they can generate from the stolen data
Dinner 🩸👍🏿
100% should.
Every day we sink further and further into the VOID
Void goblins ASSEMMBBLLLEEEEE
"I'm not confident about that..." 🤣
I'm going to have to use that one now!
Gotta love one of the other arguments that popped up...the "well if we didn't do it then someone else would". Imagine the bank robbers in court: "well your honour, if we didn't do it then someone else would" 😂
"I'm not gonna talk about that" sounds WAY worse than "I'm not sure"
I disagree, the CTO not knowing how they built the product sounds way worse to me
As far as I’m aware. That’s proprietary information. I’m not at liberty to discuss. Those are classified trade secrets.
Cop: "Is that you on the CCTV with the bloody knife in your hands?"
Perp: "Weeeeell, I'm not going to go into that right now. But if it was, I would only use a knife someone else left lying around."
Or in the very near future: "No sir, and you cannot prove to me that this is real CCTV footage and not AI generated."
"I can assure you that the knife I was holding was publicly available."
Here's the main issue of Open AI, no one would care if it was still under the umbrella of being open source and being a free boon for everyone. But the fact that it's a company with a profit motive now means that whatever they do will be fought and hit at with all their might. Mean while smaller AI companies will swoop in and just be more sneaky.
What Im saying is that OpenAI will most likely collapse and be reformatted with there being better options like Claude out there. But they will be the shield that other companies hide behind
I'd like to agree with you but in reality the times would also sue them if they were still open source, because they would still be threatening their business using their own data
Apparently there are programs that can completely brick AI’s if they scan an image with a certain filter on it, and I really hope this will be applied to all videos and images in the future to prevent theft.
She's lying. Straight up
So if south did 1860s bad stuff, the real problem is the north industrializing?
She's not even good at lying. She screams diversity hire
She's not lying... publicly available can still be copyrighted.
@@apocalypseap Except, their is a thing call fair use.
I mean she could just say yes and citing fair use law.
"did you use stolen items?"
"only if they were publicly available to steal"
"I only stole the bikes that weren't locked"
There is no theft here. Analysis is not the same as reproduction.
@@alfred9805 what if I had a tool to scan the bike and make a new one instantly? would that be theft? cause its not the same bike, The bike is the Data I required to scan to get a new bike for myself. Did I need to ask the owner of the bike for permission to scan? Does it matter that I scanned without permission? What if the bike owner was actually a company that sold those bikes, Would it be theft? Im not physically steal or even touching the bike, My tool just scans it and recreates a bike out of thin air
@@G0DKILLER_ I can't tell if this is a joke or not lol but the answer is: Yes, its theft from the company that owns the design and the rights to reproduce the design. It would not be theft from the person with the bike however because they only own that instance.
While we take for granted that any videos uploaded to youtube belong to it's creator, I wonder if youtube doesn't have some ambiguous clause deep in their ToS that they can use to partially 'own' your videos and lend them to these AI companies or to google's own AI.
AI companies won't ask Google, RUclips or anyone. These scumbags will just scrape the site.
I was thinking the same. To what extend is the video you host on yt still "yours"
I would be very surprised if YT did not.
How do you think Shutterstock got permission to sell user uploaded stock video to OpenAI?
I didn't agree to that
@@Julez60This is actually even more dangerous, especially since Alphabet has its own competing models. Remember how reddit and Twitter shut down their APIs and effectively killed all third-party clients? Given their war on ad-blockers the last thing we need is for OpenAI to give RUclips a reason.
No thought is truly independent. We read books, watch movies, gain experience at work. We then process this and create new patterns of thought, which can be unique to us but not without inspiration/seed material.
I think if the material shouldn’t be used in training data, it shouldn’t be shared at all.
Of course, the problem is that we now have a system which industrializes this on a scale and pace never before seen.
I'm loving the voidzilla channel
it's great
Better than the main channel
@@sdavis7096it’s different than the main channel, not better. It’s better for casual content, but for in depth investigations, this channel is worse, because it doesn’t do that (nothing wrong with that)
Both are great in their own way. I like the deep dives, but more frequent coffezilla commentary is nice, too.@@dahleno2014
If Sora generated a jurnalist covering scams and con artists, and it generated Coffee, I would not be too shocked.
Who else could it generate? None other than "the one and only"
This whole AI thing feels like crypto all over again
It really does. It kind of does things sort of. But those things.... aren't really popular. Go look up "ai video" and look at the view count. So many books written by ai, but... They suck. They're also not accurate (so hard to use them for things that need precision as they're only 80% accurate) So..... how does this "create value". Only thing it seems really good at is grifting.
@@Saliferousno it doesn’t. I’m probably a 5x programmer with chat gpt than without it. That’s value. My company uses it to summarize massive amounts of data that would take our customers way too long to read. That’s value. We use it to filter text related things that are presented to us in multiple languages. It’s actually incredibly valuable but there’s also a lot of bullshit
@@nhanon67asif you're a 5x programmer with ChatGPT how good were you to begin with? Let's forget about the amount of bad code it generates but also the amount of actual vulnerabilities some of the code it spits out produces.
Are you a junior developer? Because if you are, a 5x productivity boost makes sense but if you've been doing this for some time? That kind of productivity boost is... YIKES
@@Saliferousthat's not a very accurate take, ai isn't useful to a singular entity without an interest in what ai excels at. I know because I am a trench digging jarhead, I got no use for it, and to expect ai to somehow shovel for me just because some grifter said it would and when it doesn't I call it out as non valuable is just not a realistic way to look at ai, do you get my point if it even made sense?
@@nhanon67asThat just means you suck at programming
“the ingestion of copyrighted works to create large language models or other AI training databases generally is a fair use.” - Library Copyright Alliance (LCA)
I tend do agree - for the moment. First of all, ML algorithms do not copy pictures, they "learn", they are "transformative". If we do not want to limit artists from learning by analyzing copyrighted materials, we cannot outright stop machines from doing so.
Secondly, the products presented by OpenAI are definitely innovative, they contribute to our understanding of machine "intelligence", so they can be considered "research", which is traditional exempted from copyright rules.
That said, this technology is going to stay, and become more and more commercial, and less about pushing the barriers of human knowledge. So we still can and should develop mechanisms for rewarding authors of the source materials.
In other words: OpenAI is currently an all too easy target for grudges, instead of focusing on the pioneers we need systemic improvements (in law and technology) to reflect the new realities
This will be reeled in when the first copyright lawsuit is a 200 million dollar payout
This is my thesis for my masters right now. Its scary out there. So many interests are invested here. the worst part is once the data is used to train the model you can't just remove it out. It will be so integrated into everything that it would be impossible to track a particular copyrighted content if the original owner wants his data to be removed.
There should be laws making models that doesn't have it's latent space coordinates annotated with licensing/attribution data illegal to use commercially.
it is possible.
it is named rm -rf and then rebuild without that data. and it is expensive
maybe stop and delete... Humanity stopped and deleted a lot of bad stuff in the past
I don't understand what you guys are talking about but lawmakers are discussing all this right now and are trying to understand the scope of potential risk before making a comprehensive law
@@ramarajuKannaa essentially everything a generative AI produces consists of different data points in it's latent space being sampled and remixed. If it was annotated and the algorithm collected a list of the influences all resulting works could at the very least be attributable the resulting work correctly or ideally pay a portion of the API fee to the author of the influences.
Most of not all AI models are the result of wholesale acquisition of data that has not been suitably licensed.
Current implementations do not do this sort of annotation partially to optimize performance, mostly to avoid liability.
I called this out at a company I used to work for, where we were using data but never asked the licensing question. We started to ask the lawyers, turned into a huge "pay us or no" and the project nearly got canned. That was several years ago before chatgpt, and when that came out, I shuddered. This whole thing is trending towards a disaster.
I just re-watched Terminator 2 last night for the first time since before all this ai sh** started going into turbo mode……… the parallels between Skynet and open ai is straight up unsettling……
That's because you have no idea what AI is. It's just programming. Basically it's not than different than math. If you look up how math evolves, there are questions that are fundamental that take centuries to solve. Same will happen with computer science and AI.
The answer is a simple as: We read all “common creative license” video on RUclips or any other platform that contains media on that license. Period.
When are people going to learn: interviewers are like lawyers. They usually do not ask questions they either A. don't know the answer to or B. know you can only answer in a way they want you to answer. Either way, just say "no" to the interview lol
there's a lot of activity in the void it seems
its thriving
The void saves battery life because of the vanta black background
She’s ready to be a politician judging by that little clip 😂
Even if Shutterstock did license the images and videos to openAI, that doesn’t mean that the original copyright owner gave permission or benefitted from the deal. Given that these models will completely and permanently destroy the market for real images and videos, making those original images essentially worthless, those copyright holders should have expected vast payoffs for the use of their material.
Nonprofits AI outfits might have an argument for Fair Use of content in training LLMs (research purposes), but when they're developing AI for commercial use I think the letter of the law and jurisprudence are pretty clear that licensing fees need to be paid to the copyright holder(s). It's probably an issue of creators not knowing about their work being infringed (which lots of these AI companies try to avoid through opacity and NDAs), but that's neither a responsible nor a sustainable business model.
Its fun to see Coffee Stepping out of his usual crypto/scammers field to cover Tech news.
i am so Used to his style of investigative journalism now.
I'm glad he's chilling and experimenting.
this was so fun - I love this void channel - you posting whatever tf you want is such great content.
CTO: "beats me, I have no idea how the tech works lol".
She's just like me fr
If you've worked directly with a lot of CTOs you'd know that's the case most of the time.
She's a manager. She might have worked as a computer programmer or software engineer at some point but for years she's been an administrator. She really doesn't know how new technology works.
@@Lashiec9
They're managers. They haven't worked on tech for a long time.
This is why we could've used Andrew Yang as a president back in 2020. A major platform of his was to codify a law to require user consent and users getting a cut off of procured, sold, and resold data into a UBI fund, along with VAT on discretionary, non-essential consumer goods like luxury bags, yachts, and first class flights.
Im not a behavioural professional but when she's answering the question at 1:40 she's shaking her own head while talking. @TheBehaviorPanel please cover this! Would love your informed opinions! @Voidzilla, it seems that OpenAI are taking the Uber rollout approach. Get the technology and infrastructure out and then deal with repercussions later. They are doing this to get funding now which will go toward legal fees.
I'm an artist, and these videos are so funny to me.
Its funny because it took AI to suddenly threaten to take RUclipsr's jobs specifically for them to start making the same realizations we did a year ago. The industry standard for AI development has been theft and plagiarism from the start, this is old news.
Not saying youtubers don't care about other people's wellbeing, its just interesting how it takes things getting personal before people start paying attention to this stuff.
Thanks for sharing that video. I highly recommend also the Adam Conover interview with Gary Marcus, Emily Bender and Timnit Gebru as additional context on the genAI matter
I just like hearing coffee talk and give his opinions. I will watch The Void and Coffeezilla as long as I have the means to do so.
Not even half way through your video and you have the most accurate understanding of the fair use doctrine and what transformative actually means under the law that I’ve seen outside of a law school lecture. Sora and other A*I is exploiting the copyrighted work of others, there is no change of character or purpose in the end product, the models sample the entirety of the original copyrighted work, and it lowers the market value of the original work, and these works are typically not factual in nature. It doesn’t fit the carveouts such as commentary, journalism, satire, or education in a classroom. Further, these companies are going around like knobs stating it’s fair use like it’s a fact when fair use is an affirmative defense in copyright infringement litigation. It’s not something someone can just declare and it is so, it’s something that is decided in court.
This SHOULD have been resolved ten years ago with Web 2.0. All these companies have BEEN screwing over users for over 20 years at this point.
Excellent take, so true
when the in court they have to tell truth no matter how much they avoid to tell the truth
When they get sued they will have to turn over their records. Will be easy to see who they illegally scraped. They’re going to get rekt.
She's an AI 😂
Oh no the void has its hooks in deep 😧🕳
As an artist, they cant have my data even if they pay me
If I thow shit at a wall, can I call myself an artist?
@@G0DKILLER_ yes
A bike on the street is also publicly available. Doesnt mean you should take it without permission. Public domain is not publicly available.
As a tech demo I don't think it should be required to acquire license to develop the framework and fine tune how the model acquire and understand data. Like testing a blender's blade. But the moment it becomes in-use for commercial/public use then absolutely the entirety of the library used should be licensed.
I think this goes for all generative AI models being exploited commercially.
No worse than Claude and ChatGPT looting open source code.
I think it's too much of a leap to trust companies to keep track of everyone they've used content from without permission to retroactively compensate them once they decide to make money off the tech.
They use those tech demo's to get investments, so imo they should be licensing the data from day 1.
I think the problem for me is that whatever they trained sora on, they aren't going to spend money to retrain it. Beyond that, sora is a tech demo. They profited by showing that to investors.
Soon it will be a moot issue really because AI will get so powerful it will just go look for training data itself and if it wanted to could obfuscate it enough even the owners wouldn't know where it got it from.
Data? What data? 😅
Here is also another problem, the video might be free when it was used for training but it was copyrighted afterwards. This kind of stuff can also happen. Copyright laws need a lot of research and restructuring for Internet only. For AI its a whole another ballgame
what do you mean with free?
and if it hade an "open source" license, these aren't revocable
@@schwingedeshaehersit might not have been licensed yet
It's not a controversy, It's just a wake up topic that we need to get on top of. Every account we set up has an eula we "sign". The courts just have to sort it all out at taxpayer expense. We abysssin'
Tumblr and Wordpress are planning on selling their data to OpenAI & MidJourney (and probably already did handover a good bit of it based on different sources I've read). As a user of both platforms (and one who's made significant amounts of content for them with the understanding that I own the rights to it) AND an English teacher, I'm not super thrilled with the idea that my writing is being used to train AI's to replace me or help my students cheat.
I wonder if the AI has a youtube premium account?
I´m using your data for free, but will never give public access to my work, so that humanity can benefit from it. What a great company
Chicken nuggets.
Chicken nuggets.
Chicken nuggets
Chicken nuggets
Chimken nuggies
Publically available doesn't make this better.
It does make it infinitely better lol
When you buy anything with a credit card the credit card company knows what you bought, how much you paid, your location and, well, they know pretty much everything about you. The credit card companies then sell that private data and the customer, the one that actual purchased the item, does get even a tiny slice of the money the companies make selling your data. That is, the default position is that your data is their data and you are entitled to precisely nothing of that data's value.
So, it was logical for these AI companies to buy or steal your data without telling you or compensating you -- AND, after training the AI to replicate your particular creative ability they can compete against you with your own IP. If the DOJ and SCOTUS drop the ball on this then we are screwed!
This is theft. These artist will be out of a job because of these companies. I can't even put a 30 second clip of a movie for commentary without getting a DMCA claim. Yet when the elights take our work without our consent to literally replace us it okay. We shouldn't stand for this #boycott_openAI.
Gotta watch these quick before zilla takes them down
Yeah, I was wanting to rewatch the "Boeing" video and it could not be found. Why does he take down?
@@TreeLBollingTreeManThere's just all kinds of technical legal drama when it comes to this stuff no matter how many disclaimers he places unfortunately