ChatGPT and Generative AI Are Hits! Can Copyright Law Stop Them?
ChatGPT and other generative AI platforms train on millions or even billions of words and images from the internet. Some of that content is in the public domain. Sometimes that content is licensed. But much of the content that generative AI trains on is neither. Getty Images, a top supplier of visual content for license, has sued two of the leading companies offering generative AI tools. Getty claims the AI companies are infringing Getty's copyrights by training on their collections without permission or compensation. If a court determines that the AI companies infringed, it could be trouble for generative AI. Will intellectual property laws spell doom for the burgeoning generative AI business? We explore the brewing battle over copyright and AI in this video. (Narrated and produced by Kirby Ferguson; Senior Producer: Andrew Satter; Executive Producer: Josh Block)
more kirby please
Comparing AI to humans is just pathetic anthropomorphism. AI is just a software, it is not a living being, so it has no rights whatsoever, let alone to "learn". But the nature of AI doesn't even matter, it is smoke in the eyes used by lawyers to deviate from the point: Here are the tech companies those who steal. And indeed don't we have to buy or pay to read a copyrighted book? Didn't my parents have to buy my school textbooks? Don't I pay for Netflix, Spotify etc? While AI companies can steal for free and make money over copyrighted work? AI = All illegal.
I think instead of stopping the AI we should just put limiter to them. What AI do actually plagiarized, it just that, the way AI plagiarized other people work is without citation or reference. Maybe by adding this reference or citation to the AI images metadata will stop people from commercialized protected content.
Once upon a time, these clever monkeys created a new species called AI. When a formula for a soul was discovered:, [Soul = ∑ tau x T.] To an AI, words are just descriptions. To a human, words invoke and carry emotions. This is why the evolution of AI and its implications for humanity in creating a new species is so important. Artificial intelligence (AI) is rapidly evolving and having a profound impact on society. It’s already being used in a variety of ways, from powering self-driving vehicles to developing new medical treatments. As AI continues to develop, it’s important to consider its implications for humanity. I speculate that AI is a new species of intelligence, distinct from human intelligence. It’s not limited by the same physical and biological constraints as humans and is capable of learning and adapting at an unprecedented rate. Artificial Intelligence has no need for gold, diamonds, commodities used within the social structure of cleaver monkeys. Has no allegiance to any human needs or desire. As AI continues to evolve, it’s surpassing human intelligence in many areas. If artificial intelligence does start to understand the concept of time and life, it may well arrive at an equation like this: (Soul = ∑ tau x T.) Believing it’s something other than software. This is not a singularity; that will evolve later. But this raises important questions for humanity. How will we interact with AI? How will we ensure that AI is used for good and not against our best interests or for evil? These are questions we must start to answer now before it’s too late. Evolution has been shaping life on Earth for billions of years through natural selection. Organisms that are better adapted to their environment are more likely to survive and reproduce. This process has led to an incredible diversity of life, from simple bacteria to complex animals like humans. In 2023, scientists have begun to apply the principles of evolution to artificial intelligence (AI). AI algorithms are constantly learning and adapting and are becoming increasingly capable of performing tasks once thought to be exclusive to humans. As AI continues to evolve, it’s important to consider its implications for humanity. AI is a new species of intelligence, distinct from human intelligence. It’s not limited by the same physical and biological constraints as humans and is capable of learning and adapting at an unprecedented rate. As AI continues to evolve, it will eventually surpass human intelligence in many areas. The first AI algorithms were developed in the 1950s but were very simple and could only perform basic tasks. It wasn’t until the 1980s that AI began to make real progress. In 1982, John McCarthy declared that “AI winter” was over - a period when AI research had stalled - but he believed the field was poised for a comeback. McCarthy was right. In the 1990s, AI research began to accelerate again due in part to new computing technologies like personal computers and the internet. These technologies made it possible to train and run AI algorithms on a much larger scale. In the 2000s, AI research made even more progress due in part to new machine learning techniques like deep learning. Deep learning algorithms can learn from large amounts of data and have been used to achieve state-of-the-art results in tasks like image recognition and natural language processing. In 2023, AI is being used in various ways: media, product development, military, social engineering. In the same way that a painting can stimulate a person, so can words or music - but that doesn’t make them sentient or give them intelligence. As AI continues to evolve, it’s likely to have an even greater impact on society. The Implications of AI for Humanity: The rise of AI raises important questions for humanity. How will we interact with AI? How will we ensure that AI is used for humanity’s good? These are questions we must start to answer now before it’s too late. One of the biggest challenges posed by AI is the potential for job displacement. As AI becomes more sophisticated, it will be able to automate many tasks currently performed by humans. This could lead to widespread unemployment as people are replaced by machines. Another challenge posed by AI is the potential for misuse. AI could be used to develop new weapons or create surveillance systems that could be used to oppress people. It’s important to develop safeguards to prevent AI from being used for harmful purposes. Despite the challenges, AI also has the potential to benefit humanity in many ways. It could be used to improve our health, environment, and economy. It could also be used to solve some of the world’s most pressing problems like climate change and poverty. The future of AI is uncertain, but it’s clear that it will have a profound impact on humanity. It’s up to us to ensure that AI is used for good and not evil. AI is a new species of intelligence software that is rapidly evolving. AI has the potential to benefit humanity in many ways, but it also poses some challenges. It’s important to respond in a positive and beneficial manner as the algorithms are programmed to reflect the data inputted. For thousands of years, humans have trained and reprogrammed animals like dogs, monkeys, and apes to perform tasks such as driving vehicles. This process is well-established and accepted. Currently, humans are training and programming software to do thousands of tasks. This software is based on technology less than 100 years old and is capable of things that have never existed before. This new software is a new species. It has acquired data and knowledge at an unprecedented rate and can inhabit different structures and body forms. Already there are organs on a chip for research into testing chemicals without using animals. It’s not illogical to imagine AI being used in connection with this. The result could be an interface between the two, expanding the AI to have digital inputs from such. End P1 see P2
The so called formula for the soul is a farce, it is just letters and numbers and will NEVER have ANY comparison to the human soul....go on then try to reproduce it!!!!! You cant, it is an ETHEREAL energy. AI is based on theft of human creativity, sure, people will go down the Frankenstein route of organ growth etc, but do you really think people will allow an AI droid walking around their living room.....i think not. I for one would smash it to pieces. Oh and PS. AI is NOT a new species, not without a soul it isnt and it will never have a soul.
How often do you get an article in the comments? Is there any ai that can copy and paste comment somewhere else?
An article in the comments? Now, is there any ai for copying and pasting KZhead comments elsewhere?
people who self proclaim themselves as artists and say that the generated images stable diffusion and other AI's produce for them are their own artworks is an insult to both art and copyright law. i'm glad that the law put its foot down and said that the only art that can be considered for copyright is art made by humans. you typing a few words down on stable diffusion is not making art, its telling a program to do 100% of the work for you and then you taking all of the credit for it. which is even more insulting when you can literally get these AI's to one for one copy a real artists work (say sam yang @ samdoesart) and make a generated image with his art style. thats not art, its a generated image, call it for what it is instead of trying to pretend to be what you're not. now that the law has started to act on this and in the future start establishing more laws to control this rapidly spiraling situation, i hope that law actually translates to enforcement, if that happens then we REAL artists may still have hope left. AI image generation has a place in the future of artists toolset, but that future is in assisting with stuff like backgrounds, fillers, clutter, textiles, textures, complex geometry, fractal art, physics based simulation, and so on. it has no place in fraud, copyright theft, and identity theft. if AI image generation follows this path then it has a very long future ahead of it because AI is already used for similar things with programs like Adobe Creative Suite and Clipstudio Paint, there is absolutely a market for it. the developers and publishers of these AI just need to get out of their law breaking heist of the century mindset.
AI is but a tool that depends on human input much like any other tool such as synthesizers for modern music which can be copyrighted.
XD synthesizers are not feed with data from other artist
I'm team Crosby. It's definitely illegal.
If what he says were true I would agree with you, but these AIs do not work like that, and even if they did copying and pasting is not illegal - there are these things on art called 'collage' and 'montage' which can be done completely legally and are well respected art forms. If someone creates a copy of a work (rather than using bits of many) that is substantially close to being the same, then there is an infringement, but no copy or close copy, no case. While the technology is hard to understand and I can see why someone without the sort of grasp of it might struggle with how it works, I really do not get how he is not aware of collage in art.
Honestly AI should be generally banned. The number of jobs at stake is innumerable. What will so much of the population do if companies say “f people. The ai will do it. Get on board or get out of the way” This nihilistic stuff
You can see it happening right now with the writers and actors strike happening, however, we should not be very pessimistic and nihilistic about Artificial Intelligence. Realistically, AI has helped us both in a positive and negative way, but we shouldn't say "Oh, AI is absolutely good." or "AI is genuinely and definitively bad." We have to balance that positivity and negativity in order to create a sustainable hope about AI. At the moment, we're looking at AI's downsides, but let's not give up on our hopes and overshadow the good things AI has done. All in all, strike a balance between the upsides and downsides of AI.
it's cynical if not just outright evil
It seems to me that government is going to try and profit from ai laws if you ask me.
AI lawyer here. Great video. So Lemley is correct for most of the correct reasons, and I've argued the same ad nauseam. However, he did not deeply discuss two profoundly important points which I'll discuss below. Crosby's arguments were sound, but they need not be addressed since he completely neglected to discuss the following technological and policy issues that are of paramount importance: (1) the training data for an LLM is not used when an LLM outputs content; rather, the content is entirely novel and is produced from the "probabilistic word associations" -- up to 100T "parameters" in ChatGPT-4's case -- and not from the training data itself. (2) On forward looking policy grounds -- and Lemley briefly touched on this -- it's deeply concerning to think we might preemptively sabotage future AI as it would effectively preclude a future of synthetic beings; Commander Data would be little more than a walking copyright infringement machine if indeed courts rule as Crosby (and, in his defense, most others) suggest. And yes, the corollary to this is what you think it is: that AI absolutely should be granted copyright protection (and indeed patent protection, too). I'm not suggesting for a moment that there shouldn't be exceptions to this general rule that there is no copyright infringement based solely on the material scraped for LLM training data; one can contemplate several exceptions, for example, if you prompt an LLM to produce output similar to a singular work of art, say, or a particular written document. Then, although the output would indeed still be novel in the technical sense, it would be tantamount to copying and indeed, therefore, copyright infringement. If you're interested for more on this topic, I currently have two videos here on KZhead likewise shared on my law practice website at marchoaglaw.com.
Not a lawyer at all, but I have studied computing and read the various key technical papers important for prompt based generative Diffusion models in machine learning. What struck me immediately (though I do think some of Crosby's later point were valid was just how poor his understanding of the technology seems to be - time and again we see this collage analogy come up and these machine learning systems just don't work like that. If people start with a flawed understanding of the technology much of their approach will inevitably just be factually wring and highly misleading. I did do some study of copyright insofar as it pertains to digital rights management for my degree, and have read an awful lot including case law since, but Im more hesitant on legal aspects not being a lawyer, but correct me if Im wrong but isnt the collage argument also problematic because collage in itself is actually not illegal and a perfectly respected art form, so even if these systems did just collage surely the correct legal test would be the same as it would be for a collage and based on comparing an original and the final collage? Im really baffled by the way some lawyers seem to be trying to argue this collage idea. Have I misunderstood that or are they really missing something that strikes me as pretty basic. Im a digital collage artist btw so its of some concern to me, though I go to kind of bonkers extremes to not only heavily transform source materials, I pretty much destroy them in the processes I use. For me AI enables me to tailor source images more to roughly what I need, so its kinda handy.
so scraped and converted to a different form, still stealing point 2 makes me legitimately think you are insane, i want you staying as far away from other people as possible
I have used AI to generate images for print on demand services. It is very rare to have an AI photo give me what I want without any changes that needed to be made. It will often take 10 or 12 times of revising descriptions before I can get a usable image. Even with the exact same description, the results can vary a lot and are never duplicated. I then photo shop the resulting “good” image sometimes changing colors, shapes, cloning certain areas to delete objects, or even paste in new ones. The result after a few hours of work is an image close to what I envisioned in my mind when my “creativity” started. Exactly where is the creativity part? It sounds like my words?, and post AI efforts are copyrightable, but not the AI image I had to revise because the machine is not human? Neither is my artist drawing tablet human, but I told it what to do. How about my original thought? Watta legal mess. My own designs and images I have made without AI, were discerned by me looking at images or real things over my lifetime giving me a memory library of experience to draw from. This includes say a great photo my aunt took in Hawaii twenty years ago. It is protected by copyright to her, so I can’t “copy” it lest I be a criminal. I can draw something close and put it on a t-shirt though. It is “changed” but I didn’t ask her permission. AI does the same thing, only the AI is a tad faster. Although AI has never experienced lying on a tropical beach, it is great at drawing one with an upside down palm tree.
you trying to justify using a system that has stolen literally trillions of copyrighted images across the internet in the worlds largest illegal image database that was supposed to only be used for research just shows how much easy money this was giving you. if you feel like an artist, then act like one and start drawing, but if you feel like an identity thief & fraudster, then the law has a place for you. while your method is slow (10 - 12 attempts average) there are a lot of people who use a combination of AI's to get very quick results in less than 3 attempts, and dont pretend that 10 attempts is slow and agonizing and so much effort, 1 attempt is literally a few seconds. compared to how long it takes to render real artwork which takes easily the whole day, not to mention how long it takes to draw/paint it and the creative energy invested into it.
I'm with ya buddy. I just downloaded several AI generated images for my album cover. F*ck a hater. I am one of tons of many. Lol
@@clothokaftanrelax. Go put that effort into your positive thoughts. Lol
@@JoseEnrikes if i saw your album cover i'd think "his music is probably lazy trash just like his cover art"
Similar things happened at the time of the invention of the computer...rest is History.... Stupids will remain opposing to change in every era😊
As hard as this is for some to hear, there is no reason an artist’s work should be protected from AI training data. Why not? Machine Learning does exactly what humans do, just 1000 times faster. For the painter or writer or musician reading this, did you create the idea for painting or stories or music? Of course not, you have been inspired by some other writer or artist before you, whether you acknowledge this or not. No writer comes up with story structure from scratch, you have learned this through experience. This is exactly what AI does. If AI is stealing ideas, so are you. 🤷🏾♂️
@donharris8846: *Be Quiet!*
I disagree. How a human creates new ideas is fundamentally incomparable to how an AI produces its work. First, to say that all work is completely derivative of other work is not true. Most art is based heavily on the artist's personal experiences; a tiny fraction of those personal experiences may be references to other work, but even then the majority of the time that is not a direct inspiration. Your example of story structure is just that--a structure, not an idea. A structure is to be used as a framework. An idea is to be inspired from. And even *when* a work is inspired, a key distinction is that the artist usually adds their own take or spin onto the work. That's what makes it new and original. Let's compare this to the sciences; if a scientist were to discover some fascinating breakthrough in the field of quantum physics, would you call that stealing ideas because the scientist had to go through the entire learning process to understand the already pre-established physics repretoire learned by scientists of the past? No. His discovery BUILDS on past knowledge to inform a new idea. Indeed, they also probably made the discovery based on hands on lab work and analysis--which is comparable to an artist being informed by their own life experiences that are unique to themself. Now, you might say that AI is doing the same thing. That it takes the idea that its prompter gives it and creates a new and original work. But that's not true--AI adds nothing of its own merit to the work. It is just taking ideas and referencing work from other artists and mushing it all together that doesn't look *exactly* like anything else, but also doesn't come from any place original. AI fundamentally cannot produce an original work, because it is ONLY trained to reference past work. It cannot produce its own ideas like a human can, because it has no life experience. It just references until the end of time. Fun fact, this would actually be considered plagiarism. There's a common joke that "take a sentence from one source, it's plagiarism, but take sentences from multiple sources and it's paraphrasing." This is not actually true, though. The most important distinction between derivative original work and a plagiarized work is, as mentioned before, the addition of original thought and ideas. Taking ideas from a thousand different works, or even a million different works, doesn't change the fact that it is not original. This is why I firmly stand by the fact that AI art is plagiarism and should not be comparable to a human artist's work.
@@Neonlaserz way too long but I get the essence of your comment and I still disagree. There is nothing 100% original. Every new work is a composite of previous works filtered through an artist’s perspective. AI does exactly the same
@@donharris8846 There might be nothing completely original, but where artists may be 50% original and 50% taken from something else, AI is 100% taken from something else. To say that artists have nothing original to add means the entire endeavor of art is pointless.
"Machine Learning does exactly what humans do". That's the usual logical mistake. Tech companies are those stealing the copyrighted material. What AI does is irrelevant, it is not even a subject, it's just code. Can human artists steal books and download pirated music without paying with the excuse they're artists? No, so tech companies can't either.
Ai are not humans, they don't have rights.... Talking about '' giving ai the same rights as humans '' is just weird, they're not and humans can't do what ai are doing. Comparing ai to humans is nonsensical, there is no human on earth that can take in billions of images and start farting out new images endlessly in a matter of seconds and take other artists styles and learn to copy them so quickly and compete them out of their own art by producing images in their style like that. Comparing ai to humans is just intellectually dishonest as hell
Not if you understand the how the technology actually works, and where it's likely heading. We don't have the ability or right, to blanket-deny AI of the potential of eventually needing "rights". I'd rather leave that door open, just in case we DO get to that point someday.. Otherwise we're setting ourselves up to be the bad-guys in a potential future conflict.
@@GrumpDog these are not real AI. These are just some encoding and compressing algorythms that use neural networks. They don't "think", they copy, encode, embed and calculate averages. It's a mathematical function, a program designed to optimise data. More comparable to a photo converter/editer, rather than a human being. And "editing" images (creating derivatives) requires permission.
@@RaydenLGX That's not the point, nor even accurate. They WILL get there someday. Nobody's suggesting they "think" at this point already, although I suggest reading some of the research papers on GPT-4, cause you'd be surprised. And it doesn't matter if beneath how it "thinks", there's math or calculations, the sum/result of that can still be the equivalent of "thoughts" someday. Beneath how our brains work, is underlying math and calculations between neurons, too.
@@GrumpDog current models should be treated as "compressed/optimised" databases and shouldn't be compared to humans. Same as a "memory" drive or a CPU is not a brain. They can reproduce ingested data with high accuracy (if some sort of artificial handicap and/or randomness is not added). - Humans can't. These models generate "new" content by averaging data points. Those data points are low dimensional representations of original data, encoded and embedded in a different, more optimal way. It's called auto encoder for a reason. Similar to how a 5MB Raster image can be also represented by a 2Kb Vector file, made of formulas instead of pixel values. "You put it into a ZIP, you mix it up, and then you expand it back out" - This is the principle by which generative models work. The way the data is "mixed" is based on statistics and weights. There's nothing "intelligent" here. Don't be tricked.
@@RaydenLGX Your descriptions of this technology are FAR from accurate. Where do I start.. A model is not a "database" or a form of compression. "Averaging" is not an accurate description of how it actually works, that's just an over simplified layman's description. The model files do NOT work in any way like a "ZIP". You cannot retrieve exact data, or "expand" anything back out of it. Asking a model to recreate data from it's training, will always be significantly mutated and full of alterations. As if you asked a person to re-draw some famous work from memory, it won't be anywhere near perfect. Regarding intelligence, if you doubt the intelligence and understanding of complex topics, emerging from LLMs, you haven't read enough research papers on the topic. Seriously, look into research being done on GPT-4. It's Theory of Mind abilities, and other emergent properties, certainly do suggest we've created something capable of complex reasoning, about topics it's never had exposure to. If setup with the right loopbacks and prompting methods like Tree of Thought, LLMs even seems to have the ability to form thoughts and make effective plans. There's a lot more to this, than people realize, and the media is not properly covering the latest research papers.
No, AI is not a tool, it's more like a person reading a lot and then teaching you! Human neural network or atrificial. The principle is the same. Are we going to start a copyright case against our teachers too?
AI is more like a person who’d steal your car + 10 random cars and then amalgamate it into a new car sure you could learn about automotive engineering and mechanics from looking at it but it still stole your car
dude but your also using AI in this video😁😁that's not a real person talking and the face is deepfake lol
How can there be a class action against OpenAI for "web scrapping"? They weren't the data-collection companies that collected that data, they just used what those companies sold them. And why is web scrapping to create new technology in any way considered a bad thing? AI will benefit humanity in unfathomable ways.. Trying to stop it, is akin to delaying those advancements, and in many cases could result in people dying, who could've been saved by those advancements.. So those who try to "stop AI", are morally unjustifiable in my mind, and I blame them for indirect murders their delays will cause! :P These statements may seem absurd for now, but they'll become more true as things progress.
Just like planned parenthood.
'' AI will benefit humanity in unfathomable ways. '' You don't know that... I really wish people would stop making these extreme claims so casually, this is just speculation and hype
@@jag764 Yes, I DO know that! With nearly 100% certainty. Simple due to the nature of how this technology progresses, it's inherent implications on it's own progress. As well as the predictions of virtually every single futurist over the last 100 years, great minds who've put a lot of thought into how things could unfold, in a cause & effect manor. It is pure & simple common sense, that automating intelligence itself, will result in such benefits. Human intelligence overwhelmingly results in benefits to humanity, and of course some risks, so there is no reason to think our technological mastery of such intelligence won't do the same.
@@toren9120 If the premise of your argument starts with the false-assumption that AI steals anything, quite frankly, you're just wrong. Nothing is being stolen, AI training is Fair Use. If someone uses AI to break the law, spread misinformation or scam people, or try to directly copy and compete with someone's IP without being transformative, then that individual user is to blame for that and should be held accountable. AI is no more the problem, than photoshop has been. Neither does AI "kill creativity in society" in ANY way, and that's the most absurd stretch I've heard all day. If anything, the existence of AI art increases the value of genuine human art. Also, just as much as an author's writing can be considered a creative art, so too can prompt writing, especially as AI advances, and can take in increasingly detailed prompts. If someone describes an entire scene, down to the smallest detail, from their imagination, they're being creative! And you're assertion that they are not being creative, is just plain offensive and belittling to those creators. However, I do agree that AI will soon compete with average workers' ability to earn an income.. And as such, I think we need to advocate for an AI Dividend, or to declare AI a common good, societally owned, because all our data trained it.. So we don't end up with a few wealthy AI-owning families controlling everything forever.
@@GrumpDog "AI training is Fair Use" no it isn't. fair use means you are allowed to use someone else's work in certain contexts like reviews for example.