AI COPYRIGHT

So Sam Altman and OpenAI have essentially said something that is a pretty big statement. They basically said that the AI race might be over if this law isn't passed. And I'll dive into exactly what they're talking about. But this is something that is super interesting. This entire thing is about copyright and how AI companies are struggling to face their lawsuits where they are facing copyright infringements. Now, if you don't know why this is such a big issue, it's because AI models are trained on trillions and trillions of tokens. But you have to understand that the problem is that these AI companies needed to go ahead and get that data from somewhere. And these companies actually scraped the Internet. Now, the Internet contains a lot of public data, but a lot of times you can't just scrape that data and use it for wherever you want, especially when you go on to make it into a product. You have to collect the data ethically. And because a lot of these AI companies did not collect the data ethically or get people to create the data themselves, they, they are running into an issue now where they essentially trained these large language models on copyrighted material. And this ranges from image generation models to text generation models. And it's all now coming home to roost where they've trained these models on vast amounts of copyrighted data. And now they're like, okay, we have the models, they are here, but if we don't have a situation where these lawsuits go away, they then we're gonna be in a really big problem. That's why they're saying, look, if this doesn't go away, then the AI race might actually be over. So you can see right here it says that OpenAI urges the United States to allow AI models to train on copyrighted material. You can see it says the tech giant behind ChatGPT urged the Trump administration to let go of unnecessary, burdensome regulations on AI. And essentially this is something that they do need because the lawsuits are on copyright are mounting. So basically why they want to talk about this is because they are framing it in the sense that they need to sort of maintain America's lead. So you can see right here that it says what they want to do is strengthen America's lead. And this is where they're saying that, look, if we don't allow companies to do this, then America won't stay ahead. And I think it's a smart marketing tactic because America wants to stay ahead by far. And of course they're going to do anything they can in order to do that. So of course, the proposal is a part of a wider plan that the company submitted to the US Government as part of Donald Trump's coming AI Action Plan. And the AI Action Plan is a policy initiative aimed at enhancing America's leadership in AI while ensuring national security and promoting competitiveness. And this is actually being developed under President Trump's executive order, which seeks to sustain and enhance USAI dominance. Now, you can see right here that it says the administration solicited input from interested parties across the private sector, the government and academia, framing the future policy as a shift that would prevent unnecessary burdensome requirements from hindering private sector innovation. Basically stating that, look, if we don't do this, then this is truly going to hinder the innovation that we are working on. And you can see that the real thing that OpenAI is citing here is fair use. And this they kind of do have a point, because these models are generative, because these models don't actually produce one for one what they've read. So they kind of do have a point here. And they're saying that OpenAI's models are trained to not replicate works for consumption by the public. Instead, they learn from the works and extract patterns, linguistic structures and contextual insights. And this means that our AI model training aligns with the core objectives of copyright and and the fair use doctrine, using existing works to create something that is wholly new and different without eroding the commercial value of those existing works. Now, I will say it's easy for OpenAI to state fair use since they are the company that is benefiting from this. But I will argue those who are working in certain industries where opportunities, jobs and many other things are being eroded due to the new generative AI are certainly on the opposite end of the spectrum. Imagine you are an AI art creator and you're essentially getting nothing in return when you realize that these companies have downloaded all of your images, trained AIs on your styles, and people can use a simple text prompt to generate what would take you hours. This is something that I would argue many other individuals would say, nope, this is not fair use at all. Now, I completely understand the use case here, and I'm not going to dive into all of the specific details, but I will say that in some specific cases, certain individuals and certain works who have specific styles definitely need to be compensated. There are literally certain prompts that are mimicking certain creators art styles, and it definitely takes away from the initial piece of created art. Now, writing is another category that I think is just too vague and I think there are so many tokens that it is pretty difficult to realize where the copyright is there. But the legality surrounding this is pretty great at the moment with several lawsuits still progressing. Now what they are suggesting is that OpenAI basically are stating that the US should end these court fights by shifting its copyright strategy to promote the AI industry's freedom to learn. Otherwise, the People's Republic of China will likely continue accessing copyrighted data that US companies cannot access. And, and this is going to give China a leg up while gaining little in the way for protections of original IP creators. And so this is where OpenAI have said something that I kind of agree with. They're basically stating that look, let's say you ban us from accessing anything copyright and we have to do the data creation ourselves, which some of them kind of do. And companies like Adobe have actually managed to ethically source all of their data while still maintaining good models. I do think that this is a situation where they're staying. Even if US companies decide not to actually do this, other companies from other countries are going to be able to access all of that copyrighted data and their governments are going to do nothing to stop them, which will allow them to catch up to the United States. So that is actually a good point because we know that there are other countries out there that are just going to scrape the entire Internet and do with whatever they want honestly with that data. So their argument is that if we don't, China will. And I do think that with the AI race, companies and countries are definitely going to do anything they can to maintain a competitive advantage. Now you can see here that it is actually a difficult choice. It says that the federal government can't both secure Americans freedom to learn from AI and avoid forfeiting our AI lead to the PRC by preserving American AI models ability to learn from copyright material. In their policy recommendations, OpenAI made it clear that it thinks funneling as much data as possible to AI companies, regardless of right holders concerns is the only path to global AI leadership. Basically taking a look, this is the only path that we can take to global AI leadership. If we do anything else, we're going to be in a serious situation where other companies and countries are allowed to train on those copyrighted materials. So they're stating that, look you, if you do this, you're probably not going to maintain leadership. Now you can see right here that if other companies and countries have unfettered access to data and American companies are left without fair use access, then the race for AI is effectively open. Now maybe they're being a bit dramatic and I will explain that later on, but they're stating that this is the key thing that will determine whether or not they're successful. They say that if America loses, as does the success of Democrat AI, ultimately access to more data from the widest possible range of sources will ensure more access to more powerful innovations that deliver even more knowledge. And I kind of do have a point here. Data is one of the fewer sources for AI. But of course I do think that is changing. So let me know what you guys think about this. But I have a question to ask all of you that I didn't really see most people talking about. Will this even matter in 12 to 18 months? And hear me out when I say this. So of course, I've done many different videos on AI, and one of the videos I did a video on was Ilya Sutskeva's video. So Ilya Sutskeva is the brain genius behind OpenAI's recent innovation, which is test time compute. Now, remember he did this video where he spoke about many different things and things that were coming in the future. And one of the things he spoke about, which was a very, very big thing at the time, was that pre training, as we know it, will end. And this is where I want you guys to pay attention. He said, what is growing is compute. We got better hardware, better algorithms, and larger clusters. But what is not growing is data. We have one Internet, which is the fossil fuel of AI. And basically in this video, he said that we've exhausted all possible data. And this is why I'm stating that if AI companies have already exhausted all possible data, will this even matter in 12 to 18 months, when data is not going to be the primary source for what's driving AI innovation? Previously it was, but in the future, it's quite likely that data won't be. The reason why I say that is because look at this. Remember, GPT5, GPT5 and other large language models like the second iteration are actually falling short of expectations. This article states that OpenAI's efforts to develop its next major model, GPT5, are running behind schedule with results that don't yet justify the enormous costs, according to the new report in the Wall Street Journal. And you can see right here, GPT5 was not a giant leap forward as previous models. And basically what I'm saying here, guys, is that, look, previously we used data to build these AI models and that was the key to accessing higher levels of intelligence. But now the paradigm has entirely shifted. We're not growing this data anymore. Of course you can go out and get really selective data sets, but you're probably going to be creating those yourselves anyways. So the point is, is that since data is not growing, and this is something that companies have already probably exhausted all of, and we know that that is not, of course, what is going to drive innovation, isn't this new paradigm what is to pay attention to. Because of course the innovation here has actually come from test, time, compute. And if you aren't familiar with this paradigm, this is basically where, if you allow the models to think for longer, this is basically where the models get smarter. And the reason I say this as well, I may actually include a clip from Yann Lecun, but he talks about this problem as well. And this is why I say this may not matter in 12 months. If we've reached a point in AI where we've exhausted all of our data and put it into AI models, companies are now going to start looking at different ways to innovate. And one of the ways they did that was of course with test, time, compute. And all of these new innovations do not rely on gathering more data. In fact, some would argue that gathering more data is actually driving us down the wrong street. Because humans, when they learn, they don't need 5 million examples in order to become proficient at something. Now of course this might be something like Moravec's paradox, but I do think that this is an interesting point. And I will include the clip of Jan Lecken.

Now we're missing something really big because, you know, never mind trying to reproduce human intelligence, we can't even reproduce cat intelligence or rats intelligence, let alone dog intelligence. They can do amazing feats. They understand the physical world. Any house cat can plan very highly complex actions and they have causal models of the world. Some of them know how to open doors and taps and things of that type. And in humans, a 10 year old can clear up the dinner table and fill up the dishwasher without learning zero shot. The first time you ask a 10 year old to do it, yes, she will do it. Any 17 year old can learn to drive a car in 20 hours of practice. But we still don't have robots that can act like a cat. We don't have domestic robots that can clear up the dinner table, and we don't have level 5 self driving cars, despite the fact that we have hundreds of thousands, if not millions of hours of supervised training data. Okay, so that tells you we're missing something really big.

But the point I'm trying to make here, guys, is that I do think that whilst yes, of course, maybe they should be an exception for these AI companies and probably should kind of have some rebates to those who have helped develop those models, maybe if they had stolen certain art styles, I do think some artists should definitely be compensated. But the point I'm trying to make here guys, is that in 18 months I think there will be new innovations that are not based on large data collection like before, such as pre training, which is exactly what Ilya Sutskova said. And I do think that more innovations will simply be as a result of better algorithms and more innovative ways of using that data, rather than exhausting more sources. Of course, that's just my opinion. Let me know what you guys think. And it's going to be super interesting to see how these lawsuits play out, because recently in a recent lawsuit, one of the individuals who sued one of these large companies, they actually won that lawsuit. And I do know that these companies are actually now starting to sign deals. For example, OpenAI, they've signed a deal with Reddit to access their data websites like the New York Times. And it will be interesting to see how this issue progresses as time goes on. Like I said, my opinion is that this doesn't really matter in 12 to 18 months with some potentially new innovation that allows AI to learn like humans. But at the end of the day, we have no idea. With that being said, if you guys enjoyed this video, I'd love to see you guys in the next one.