Subscribe
Share
Share
Embed
AI is the biggest technology shift of our lifetime. This show is about how to profit from it together. Each week I talk with the founders and CEOs closest to AI and Content, the ones figuring this out in real time. I’m also building an AI content business myself and share the lessons I learn along the way.
WHAT WE COVER
THE TITANS: How companies like OpenAI, Anthropic, Meta, and xAI are moving, and why their decisions matter.
THE INCUMBENTS: How content giants like Disney, News Corp, Universal Music Group, and Reddit are responding to AI, and what it means for creators and publishers.
THE PLAYBOOK: Real lessons on AI business models, content strategy, IP licensing, distribution, and getting paid.
ABOUT YOUR HOST: Rob Kelly has interviewed Steve Jobs and Bill Gates, helped pioneer early web content licensing, and built multiple companies with more than $100 million in total sales. His work has appeared on CNBC, CNN, TIME, and Entrepreneur.
Beyond business, every episode explores what AI means for jobs, creativity, families, and the next generation.
If you want clear thinking based on real experience in AI and media, Media and the Machine is your guide
Thanks! -Rob
I'm Rob Kelly, this is Media in the Machine, a show about the biggest technology shift of our lifetime and how to profit from it. Each week, I talk with the founders and CEOs closest to AI and content, the ones figuring this out in real time. I'm also building an AI content business myself and share lessons of what I learned along the way. You know, life's funny. I began my career lucky enough to interview leaders like Steve Jobs and Bill Gates.
Rob Kelly:I went on to be a three time founder and CEO, driving a $100,000,000 plus in revenue and some failures too. And now I'm back at the table, interviewing this new world's current and future leaders. This isn't only a business story, it's a human one. So every episode ends with me asking my guest what AI means for our jobs, our families, and the next generation. We'll figure this out together from the inside.
Rob Kelly:Welcome to Media in the Machine. My guest today is Tommy Petrov, the Ukrainian born co founder and CEO of Copysight AI Tommy and his team are small, but they've built something big. It's called the similarity score. Think of it like a FICO score for copyright risk in the age of AI, whether you're Disney protecting Star Wars or a creator making something new.
Rob Kelly:So when someone uses Midjourney or Gemini to create something that looks like Darth Vader, or images that look like they came straight out of Studio Ghibli, Copysight scores it from one to a 100 on how close it is to the real thing. Tommy explains why under 35% is usually safe, and why 75% plus could be a legal smoking gun. He's interviewed more than 70 general counsels about AI content risk. What makes Tommy different is he's not a lawyer, he's a creator. Before founding Copy Site, he was a creative director at Snap and Meta.
Rob Kelly:Today, he works with legal teams and art directors at major Hollywood studios like Sony and Paramount, and the Russo brothers at AGBO Tommy weighs in on OpenAI's upcoming AI generated film Critters, whether its IP gets flagged, or whether an AI made film could even be copyrighted. And this conversation isn't just a legal debate. We talk about perhaps the biggest content question of all. What happens to art when AI makes creation so easy that few people bother to create anything truly original?
Rob Kelly:And if that happens, what content do these AI models train on next? Please enjoy my conversation with Tommy Petroff. So it seems like with Copysight that one of the top problems you solve is similarity score, if I've got it right, that could be sort of a top problem.
Tommy Petrov:There's two problems on the market right now. Most of the companies we spoke with more than 17 interviews conducted with general counsels, including Metasnap or L'Oreal or big agencies like Omnicom and many, many others. The main is potential risk of IP infringement, intentional or unintentional. And on the other side, there's another problem, that when you generate something in AI, it's a problem to actually copyright it and actually own it. It's not as simple as it was before.
Tommy Petrov:Before, you take your pen, you make a doodle, this is yours. This thing is new. The creator now is AI, not a human being. And according to a law, it's not protected by copyright law, and now you have to prove that you were participating in this process. You actually had a significant human involvement.
Rob Kelly:Maybe we can take sort of an easier extreme example with Darth Vader and Disney and and Midjourney.
Tommy Petrov:Yeah. It was a wild west, and it's still kind of a wild west. But now there's a very strong signal from major studios that they are very serious about their APs being used to train models. When people start to monetize and actually generate content that contain famous IP from studios, the studios becomes nervous. And how studios become nervous?
Tommy Petrov:They're suing you. And in this case, they just start to do research and they generate it in Midjourney and compare the outputs from Midjourney with their original content. So they were willing to generate Dark Vader, they were willing to generate most of the Disney IPs in Midjourney without any specific prompts. They got very close results to the original. And the thing is that they're trying to make it seem like here's original Darvita, here's the AI outputs from Me Journey, and they look the same.
Tommy Petrov:And what Copysight does, we can actually say how the same they are, and this could be like to 99%, 98%, we put a number to resemblance. And it's like credit score, everybody knows how much money they have but the credit score reflects this in the number so we are the credit score for resemblance and IP potential IP infringements.
Rob Kelly:So like a FICO score on the similarity.
Tommy Petrov:Yeah. If it's safe or not safe to use. Yeah.
Rob Kelly:In that similarity score, so the copy site similarity score, is there even just a range these days that, you know, is it 50% or higher? Okay. You know, you're in trouble that it came up with Is there something that's becoming acceptable? In other words, is it, hey, if it's only 20%, you're probably safe.
Tommy Petrov:Yes. It's it's not even probably. You are safe to 35%. We call it anthropomorphic similarity. Let's say you have a character.
Rob Kelly:That's a mouthful.
Tommy Petrov:Yes, mouthful. Yes, sorry. So basically, if the character has two hands, two legs, and one head, it's probably a human being, and there's already a similarity from the model perspective. But from the character perspective, doesn't really matter. So 35% is quite safe to say that you're on the safe side.
Tommy Petrov:From 35 to about 75, we call it the orange stage, you go to your IP legal and you say, hey, guys, How risk averse we are? And then you go in 75 plus, this is something we will not copyright for you.
Rob Kelly:So 35% or less, maybe you're in a safe space. Yeah. What was the middle range?
Tommy Petrov:Yeah. 35 to 75 is quite big, and so it's still up to a client, up to the legal teams, and up to their IP legal teams to decide if they're go or no go, but we giving them information that there is some likeness. Or you might not know how Naruto looks like or how Spirited Away characters looks like, but this is them to a certain degree of likeness.
Rob Kelly:And the smoking gun, what percentage would that be? Where you would say almost everyone would agree that that's
Tommy Petrov:Oh, 75 plus. Yeah. 75 plus.
Rob Kelly:Just at 75. Okay.
Tommy Petrov:A lot of things that I show in my demos is 90%. 95, 90%, and that's like you see that this is clearly an infringement. It's have no doubts. Like, it's not it's like, oh, it's looks like Tilda Swinton. No.
Tommy Petrov:No. No. It's Tilda Swinton. So that's
Rob Kelly:And in that case, seems kind of like an obvious extreme example. So if someone wants to ask mid journey to create Darth Vader, then obviously, if something comes up that looks like Darth Vader, Disney, the owner of Star Wars and Darth Vader is gonna be very upset and claim IP infringement.
Tommy Petrov:They're crazy about their IP.
Rob Kelly:They're But it's not always that black and white.
Tommy Petrov:Right? Of course. It would be too simple. It would be too simple.
Rob Kelly:Does it get down to what if someone says, draw me a villain who is a Sith Lord Mhmm. Who is mostly cyber and not all human, but a little bit human.
Tommy Petrov:I'll give you a prompt that I hacked there exactly like you said. Yeah. The prompt sounded like that. Give me a superhero who's revenging his family, wearing all black, and always angry.
Rob Kelly:So a couple things. One is you could argue, hey, Disney wants Midjourney just to not do that. Right? Yeah. And end of story, that's it.
Rob Kelly:No one should be creating it. However, what about if a creator like Disney content owner is okay with it as long as they're not making Darth Vader look out of character?
Tommy Petrov:Ridiculous. Yeah.
Rob Kelly:Or like Yeah. Yeah. Ridiculous in some way. Right. Yeah.
Rob Kelly:As long as they are supporting it, are the Disney's of the world wanting that problem solved so they simply could be paid by the Midjourney users for that?
Tommy Petrov:So this is a great question because actually they are interested in fun arts. They are interested in UGC, user generated content, for those things. They are into that, but they have to have a control on that, a sort of control on that. The brand cannot allow to generate naked Harry Potter's and stuff like that. So this is something they are very concerned, but also they don't want to cut it out.
Tommy Petrov:So whenever a creator does some work that's actually fun and interesting and in brand, and they are not going to monetize it on their scale, that's all right with brands. Even when they monetize their content, but they don't monetize exactly this way of utilizing IP. Let's say you're an influencer, an influencer, you generated some funny videos and one of those videos are with IP, they probably won't prosecute you and they will do it in case by case scenario. And that's why we exist, to be honest. We want to make sure they know about what's happening and have the control on that, and in the same time they will basically choose and make their own decision if they want to pursue or they want to just keep it there because it's fun and interesting.
Tommy Petrov:We had this story with a company called Reface. They were doing like kind of deepfake AI very early on, Really nice, also Ukrainian founders. And sometimes it was like this, they used a deepfake of video with Justin Bieber. And Justin Bieber were like, wow, this is so cool. I love it.
Tommy Petrov:And then the Warner Music was like, yeah, it's so cool. We love it too. And then the company that Warner Music hired, they paid for finding those things and sue people who are infringing. They were just like going after this case because they were hired for that. They were like hunters.
Tommy Petrov:They were going through through the for the whole process. So that's how things sometimes happens.
Rob Kelly:How's it work today? Let's say pre AI. So for instance, my son's a huge Star Wars fan. He's right here. He watches a couple YouTube channels called Vader's Order and Fantasy Folklore.
Rob Kelly:It's what ifs, like, if Darth Vader cloned Padme and what if Luke time traveled back to the Jedi Order? You know, crazy stuff.
Tommy Petrov:Yeah. Yeah. The whole series in Disney, they're called What If?
Rob Kelly:They figured out that that's okay because it grows the culture of Star Wars and it's not naked Harry Potter?
Tommy Petrov:That's all their Marvel Universe or kind of what ifs. Yeah. And it's not like super public or official data, but what we learned that they're very concerned of recent IP and slightly less concerned of old IP because they already can get the whole money. So the most budget goes to the most recent IP because they are are where they're profiting the most right now. There's a certain prioritization on what they pursue or not.
Tommy Petrov:Overall, most media companies have different risk aversity for those, allowing derivative art, fan arts, and stuff like that. So they're allowing certain UGC and certain playfulness with IP, but each company decides on what level of this. And again, it comes to that if the check up, if it's too close, if it's too ridiculous, or it breaks their rules of like universe bible, that universes like Marvel and DC has those bibles, the rules, who cannot ever fight with anyone. So when I was building my own agency, we were happened to work with Disney for a commercial with Avengers. So we had all the Avengers and we wrote about 45 scripts in one month because, oh, Hulk and Spider Man can never fight or something like that.
Tommy Petrov:So there's certain rules of the universe that they they have not to they have to comply.
Rob Kelly:Is the general counsel your your ideal customer profile?
Tommy Petrov:It's always two people. It's like you're either a creative director, art manager, and general counsel. Yeah. They come together. Fascinating.
Tommy Petrov:It's an entertainment studios and gaming studios are, like, a hugest market for that.
Rob Kelly:So interesting question here. OpenAI just they've got this movie that they're committed to doing, AI generated called called Critters.
Tommy Petrov:Yeah. I I already checked the characters.
Rob Kelly:And what'd you find?
Tommy Petrov:So so there's an IP, Critters. There's a very, like, sharp teeth and ugly looking character. It's like really really scary one. I checked it, so Copysight really knows how they look like, but those new OpenAI characters are according to Copysight are clean.
Rob Kelly:And by clean, what would the percentage be?
Tommy Petrov:No. There's like there are like no matches found. 0%. It's basically less than 35 because this is how we arrange that. So less than 35%, so then they're the safe zone there.
Tommy Petrov:In our system, it's just green lidded thing. And I tried it from different angles, six characters. I did it yesterday, actually, for one of our clients. They asked me, hey, can you check it? The main reason for that, because the derivative from critters is so different.
Tommy Petrov:The horns are different, the shape of the teeth are different. And it's only fine, one of the characters slightly reminds for like 70 something percent, remind this absolutely amazing movie on Netflix with Benedict Cumberbatch. Mhmm. So he has like imaginary character. He was working on the TV and there was like imaginary character.
Tommy Petrov:So there was also a monster, and those two monsters were slightly allied. But I think their legal team would say, yeah, it's right. It's like there's no direct comparison there, but this is the only thing I found. And the logo on one of the cap was some brand, so I found something, but there's no problem with original creator's IP. Yeah.
Tommy Petrov:So we checked it. It's funny that you ask it.
Rob Kelly:Yeah. OpenAI, you know, they then let's say they released the Critters film. They haven't infringed on any copyright. Let's just go with that and assume Copy Site did did its normal great job. Yeah.
Rob Kelly:And that said though, if they use their own tools, if they ate their own dog food, drank their own champagne Yeah. In terms of using OpenAI's own tools to create that, and their own tools were trained on unknown datasets
Tommy Petrov:Yeah.
Rob Kelly:With folks maybe still suing them over that Yeah. Then it would sound like they would have a hard time trademarking or copywriting the output of their film.
Tommy Petrov:Yes. It's still the problem is not about like how close they are to the model, the problem is that it's generated by the model and model is not a So human they need to say that their creative director Chad or like Nick who is one of their founders of the company behind creators, Those guys have to say that I actually had a pencil sketch of the characters. I have this the brief that I described how they should look like, and I have photoshopped some of the first sketches, and I was impacting creative decisions on top of it to make this scene legit and characters to be copyrighted, and the video copyrighted after that, because they have the trademark on the characters. So the OpenAI is quite they're huge, and they have all unlimited resources. So, even when they released their first experiments with Sora, before they released Sora publicly, they had this fluffy character.
Tommy Petrov:Next day they released ironic tests, and it was saying, it's a character, the purple monster owned by OpenAI. So they they already had this in their dataset, at least, that this character is already in OpenAI dataset. So it's kind of interesting. They were so that showed me that they're very serious about that. So they will not release characters that look like someone, even when they play around and showing capabilities of the model.
Tommy Petrov:They're very serious about that. Not like other models. So then I found a lot of cool stuff. I sent a few videos to founders of models. They were releasing public available.
Tommy Petrov:It was with celebrities, with, like, you know, famous scenes from famous movies with characters. It was like, guys, this is not even ninety percent. It's 95 or 98% of similarity. You're kind of literally literally using the character. So yeah.
Rob Kelly:So can you copyright AI content?
Tommy Petrov:According to US corporate office, in case by case scenario, you can prove significant human involvement and copyright If you just hit generate and prompt it, it's not enough. You cannot copyright. Like, prompting and generating first image, whatever is getting there, it's not copyrightable.
Rob Kelly:And will copyright law change because of AI?
Tommy Petrov:I do think it will change. And right now, US corporate was pretty strong on we're not going to change the law. We're going to see case by case, and we can go from there. But I think they will just buying time to create something smart and thoughtful, and I hope it is, to make their new update for that. What I think the closest change will be is about AI generated style.
Tommy Petrov:So, style, when you're copying someone's style, is not copyrightable because you're creating a new asset, and you're an artist that trains your LLM yourself. I mean, your brain and your experience. In this case, when you train on someone's IP and you're utilizing someone's style as AI and AI generation, it means you're utilizing IP and that's why I think AI generated style kind of usage will be copyrightable. That's that's my take.
Rob Kelly:If you were a betting man, would so if I'm paraphrasing
Tommy Petrov:Mhmm.
Rob Kelly:Copywriting the output from the top big models right now is not doable.
Tommy Petrov:Mhmm.
Rob Kelly:Copyright law is likely to change due to AI just because there's so many new things coming up. Mhmm. And if I'm hearing you right, maybe if it does change, could change during this Yeah. Current administration because it is very receptive and supportive of Mhmm. Eliminating AI obstacles to AI business for The United States.
Rob Kelly:Mhmm. Interesting.
Tommy Petrov:It will be like fair use and style can be kind of a thing that they can consider. It's a different story. It's like because fair use is more like releasing less bottlenecks and style is because it's just a logical thing to do. That was not a thing before AI.
Rob Kelly:So on the one hand, you were you could say that Copysight your startup, is a defender of creators. Right? Whether it's a Disney or Warner Brothers and so forth, and a defender of the IP Mhmm. But also then a monetization
Tommy Petrov:Potential on
Rob Kelly:the IP as well. Yes. Is there a case of a creator or a content company who sort of is less of the protective, you know, the general counsels maybe don't wield a bunch of power and it's more the content creator side that Justin Bieber's, you know, saying, hey, this is cool. Go nuts, AI. Like, we want more people doing every deep fake of Justin Bieber or or some other character, and maybe they wanna get paid too if it goes into signed commercial.
Rob Kelly:Are there content companies out there like that?
Tommy Petrov:So first of there are companies like that. I think Roblox is kind of that company. They are interested.
Rob Kelly:The video game company? Roblox?
Tommy Petrov:Yeah. Yeah. Because they allow a lot of, like, kids creating games there. It's a very creative platform, and I really want to support that. I really think that this is how IP should live.
Tommy Petrov:People should create a derivative from the IP they love. And that's how fan art happens, but then it goes further and further. It could be monetized. And I believe in this fan art, and I believe that derivative art is something very, very important. And everything AI generated is technically a derivative art.
Tommy Petrov:But the reason why we're doing that is to, like you said, to protect creators. Because how you will maintain the creator seat in the game if everything is so easy to create? Who will train the next models if the amount of content will reduce? Less artists will be producing content. And you think that I'm crazy, but that's exactly what's happened with Stack Overflow, the website that contains all the solutions where people discussing code.
Tommy Petrov:Whenever ChatGPT three released the option for coding, new content on Stack Overflow declined for 70%. It's crazy. It's just people stopped asking each other. They start to generate codes and without asking other people. That means that the new content for training new models in
Rob Kelly:the future will be less, and we need more content. We need more data to maintain new models and actually keep it going. What about the case where an artist is creating something new? And let's say the artist is starting the next Disney. They're a new film company.
Rob Kelly:And they wanna use AI rightfully so because they could do things cheaper, more efficiently, and create new kinds of things that we haven't thought of before. Mhmm. Right now with the big frontier models, let's just say we're talking about OpenAI, Anthropic, and Google, they all have enterprise customers. They all sell to the enterprise. I have to imagine some of those customers are creators, entertainment companies.
Rob Kelly:Right? Content creators, like Mhmm. Not just text content, but film, media, So right now, is none of that copyrightable?
Tommy Petrov:It's possible manually with a very extensive and weird process. And there's a lot of like famous cases. The first case was a Kashtanawa case when an illustrator created a comic book and she generated images around that, and she came to US Corporate office to register that, and corporate office, yeah, we can't you you have not enough control on their output, so it's not copyrightable. So they worked really hard, and with the recent updates, the US Copyright Office were actually copyrighted first work from Khashtanova, so it was like illustration. And there's a few cases right now, and you have to basically bring a very clear framework of how it was created, proving significant human involvement.
Tommy Petrov:I was impacting this work so much by my creative decisions, my choices, my retouching after that, my choosing of effects and choosing of colors and choosing of composition, and everything is grounded in certain instances. Like hundred years ago, photography was not copyrightable, exactly like AI was not copyrightable. And it took a while, like ten or more years, to came across to the place where photography is copyrightable. If you make a lot of decisions like that as a creative involvement, that makes it copyrightable. And we successfully copyrighted two very complicated cases in the I generated content that officially now copyrighted by US corporate office.
Rob Kelly:So if if I wanted to create a new Pixar of AI right now
Tommy Petrov:Yeah.
Rob Kelly:And if I were doing that, I would use AI. If I were to create new content, I would use it in some form and figure the rest out. So I'm creating a new character, would in the very beginning days and let's just say I pick a young wizard Yeah. As my protagonist. Right?
Rob Kelly:There's gonna be stories about wizards for forever. Yeah. So would you be able to use CopyCites similarity score to help me? So I'm creating the new, you know, let's just call it Willie the wizard. Yeah.
Rob Kelly:And are you able to then look at Willie the wizard and the content IP I've created and and basically say, okay. This is only point 1% overlap with Harry Potter
Tommy Petrov:Yeah.
Rob Kelly:And other wizards in Harry Potter. So, you know, then maybe I'm okay.
Tommy Petrov:Yeah.
Rob Kelly:Is that the kind of use case that will be coming up?
Tommy Petrov:But it's it's even more complex. So you work on the movie, and you create multiple characters, designs, castles, Batmobiles, armors. Mhmm. And all of those things are can be copyrightable elements from the game or from like something that you missed, like manga or anime if you're not familiar with those universes. So when you generate something, sometimes you think, okay, I don't want it to look like Harry Potter because it's some young wizard.
Tommy Petrov:But it can be also remind some other characters that you're not familiar with it and less popular, doesn't make it less dangerous. I generate like a wizard, and wizard was alright, but the logo on his cape was from Harry Potter, was a crest from from Hogwarts, one of the schools.
Rob Kelly:So going even deeper, I know a documentary film owner, someone who owns documentary films Yeah. They're licensing it to AI models. So let's say my my son's a huge fan of Scooby Doo. I love Scooby Doo too. I'd watch it any night he wants to.
Tommy Petrov:I ran I ran recently the Scooby Doo check for Warner Brothers. Warner Brothers sued Midjourney, and there's another big event happening right now. And so Scooby Doo was one of those IP infringed assets from Warner Brothers. Wow. Amazing.
Tommy Petrov:So protecting Scooby as well.
Rob Kelly:Absolutely. But let's say and Scooby Doo is a Great Dane. So my understanding is AI is being trained right now. One of the things it's being trained on is also just film of dogs. Right?
Rob Kelly:So you have Great Danes running, trotting, right, sitting still, eating, and that's in an AI, and I wanna create a new Great Dane character. Right? Looks nothing like Scooby Doo, doesn't act like it, not even a detective, totally different. But I go and use ChatGPT to create pictures or video of it or one of OpenAI's products. And those documentaries are part of the training set.
Rob Kelly:Yes. And some documentary owner out there has a two hour documentary on Great Danes and how they you know, the beauty of Great Danes. And let's say, ChatGPT gives me out the video or the pictures of a new Great Dane based on this, in part, documentary. Yeah. And then I potentially am at risk.
Rob Kelly:Right?
Tommy Petrov:You are, but, like, this is like a more complex story in here. There's a big push to claim that training sets are fair use. All the models will say it's a fair use. We're not recreating those IP. We're trained on that.
Tommy Petrov:That's kind of public knowledge and something that
Rob Kelly:And I know that case you're talking about. Isn't that fair use for the AI models? Yes. Not for the opposite. Yes.
Rob Kelly:Transformative. Yeah. And I do see, you know, that looks like a direction things are heading. It's transformative, so it's okay. But not necessarily for the client, for me in that case, to create a new Great Dane character Yes.
Rob Kelly:Exactly. I could be in trouble. Right? More trouble.
Tommy Petrov:You are you are in trouble. Yes. The other thing was if you're not creating the same like you use the style of whatever artist or you have like a certain similarities on the style thing, style technically is not copyrightable. That's why I think there's a strong case about this Studio Ghibli thing is that, oh, but style is not copyrightable. So what do you worry what the problem is?
Rob Kelly:Can you share this Studio Ghibli story and what's going on AI, IP wise?
Tommy Petrov:Yeah. So OpenAI openly, admittedly, can use all their IP in the training set. So we let's start even step back. Every VLM and large model that you know scrape the whole Internet. Everything on closed, not public available data, doesn't matter.
Tommy Petrov:They just did it all. Like, all everything online is in the training set. But they will never admit it. They will never show you what they trained it on. And there will be very long conversations about that with regulators.
Tommy Petrov:That's the ground zero. So now when you go and you want to train today your model, you will be probably in much harder position to get the same kind of level of data access. Because when they were doing it, nobody were even thinking about it. They were just doing it in the shadows before AI become a thing. So they do it for like ten years.
Rob Kelly:So right now, you couldn't copy the Internet in the same way that people Yeah. But there's people like
Tommy Petrov:DeepSeek, they basically copy what what they that what they train and all. So they're basically copying from
Rob Kelly:Well, also they're in China where kind of anything goes for their own copyrights on. They
Tommy Petrov:just don't care. And they just distill data from large language models that are already pre trained. So that's how they save money. But in the same time, it's an interesting position to be right now because if you want to do it legitimately, it's it's much harder than it was in the beginning. But even though you are so OpenAI trained on the IPs of Studio Ghibli and Hayao Miyazaki, beloved anime cartoons and movies, it's amazing.
Tommy Petrov:He does it. He spends eighty years in perfectioning his style for like, it's amazing work. It's
Rob Kelly:I watched the documentary.
Tommy Petrov:It takes years to create this, and suddenly every person in the world can generate a funny meme or, like, a lazy photo in the same style.
Rob Kelly:Yeah. Didn't Sam Altman create, you know, Studio Ghibli of himself? I'm trying to remember. I thought like a bunch of them I mean, he he was proud.
Tommy Petrov:They literally named this filter. They said it's a Studio Ghibli style. So you can do the Studio Ghibli style. They haven't paid them. They haven't done anything about that.
Tommy Petrov:They just just yeah. Use it. Amazing. Do do whatever you want. And I think the lawyer says, like, style is not copyrightable, so we are safe.
Tommy Petrov:But while they were using it, the valuation of the OpenAI actually raised their number of subscription rates, so they were basically using it very commercially, and they we know that they train on that because I ran it with Copysight I ran this famous meme, Girl in the House on Fire, that done in Studio Ghibli style, I put it in the Copysight and it says 95% of Studio Ghibli IP, and specifically trained on spirited away characters, specifically recognized certain elements of style from spirited away. That was like, yeah, I think it's kind of a proof. So it's kind of like you can reverse engineer in this case and say, hey guys, you're definitely trained on that. I have a number of resemblance. It's clearly you know this IP, meaning that so how you will settle?
Tommy Petrov:So we talked to several companies, found partners, we signed with partners, Kitsugi Global and some other IP legal firm from New York. And board of directors of Studio Ghibli actually really loved the idea of protecting legacy of Japan, because Japan loses $15,000,000,000 per year on rip offs of manga and anime.
Rob Kelly:And when you and I last chatted, you said you were in a meeting with a couple 100 folks at Amazon. Can you share just
Tommy Petrov:Yeah.
Rob Kelly:What was going on there? What's their take on this world, and what'd you learn from that that meeting?
Tommy Petrov:So it was all their creators and AI people from Amazon, about 200 people. It's their, like, a monthly kind of update for what happens in AI. There was a couple demos, and one of the demos was Copyside. And we talked about it, about how to protect the assets.
Rob Kelly:What's Amazon's sort of play in all this, you think? Zooming out from that meeting even, where are they in this world of AI, IP, and creative?
Tommy Petrov:They're very seriously taking this thing. So the only AI they they have a very selective models. They are considered to be clean or claim to be clean. They are using some other things in the pipeline, and they will only onboard it with Copysight in this case because otherwise they they are not sure about those results. They have to be very confident in clean models like Bria or Firefly.
Rob Kelly:In Amazon's case, do you happen to know the name of the model that they are claiming are is clean?
Tommy Petrov:Right now, there's a very few clean models, and it's always a choice between Firefly and Bria. Bria?
Rob Kelly:Yeah. Are some large content companies, Picture of Disney, Warner Brothers Mhmm. And so forth, are some of them creating their own AI models that then their creators can use? In other words, they feel safe because their model is filled with how they Disney drew Mickey Mouse to how Pixar made Toy Story to how George Lucas created all the Star Wars.
Tommy Petrov:I know that Omnicom builds their own model, like some advertising Yeah. Some others are doing like Getty Images is a good example. They build you their own safe model. But the core of the safe model is some of their models we named already. So it's never like a ground up LLM.
Tommy Petrov:It's like we're taking LLM that is not trained on anyone's IP but has a certain core data knowledge of licensed data. So imagine you have the core model and you add to this everything GetImages has in there, hundreds and millions of those images and videos, and they train it on that. And it's still not good enough, unfortunately, in terms of quality of outputs. But they're very proud of saying, oh, our model doesn't know who is Taylor Swift, or our model doesn't know what Coca Cola is, and and things like that. And it's kind of interesting for certain use cases.
Tommy Petrov:Warner Brothers would consider taking like Bria or like some clean model and train it on Harry Potter and use it for certain instances, but it won't be all the use cases. So it's a use case just to keep it clean for certain areas or certain IPs. But when you really want to create something brilliant, the data will be critical for the quality. And quality will be critical for the mass production or if you want to show the output publicly. So then we come back to this discussion of creators don't want to use clean models.
Tommy Petrov:They want to use best models, the best outputs.
Rob Kelly:Has any LLM has any model run their model through Copysight to get it? In other words, to see if they're you know, if you and I were creating a new model, wouldn't you wanna
Tommy Petrov:Yeah. But they know run the
Rob Kelly:data through Copysight and just see
Tommy Petrov:But they know they know they know what they're training on. They know, like, the big model big big big models, like, let's say, like, OpenAI
Rob Kelly:They can't handle the truth?
Tommy Petrov:They legally cannot admit it. They know the truth. They cannot legally admit it. By hiring corporate side, they're basically saying, yeah, we train and all that. It's think their legal teams are saying, don't even talk to them because just talking to them shows that you are in this kind of, like, in that risk category of a company that steal or take someone's IP without permission and licensing.
Rob Kelly:Tell me about what's going on with Copysight and Agbo, the Russo brothers.
Tommy Petrov:They're one of their first clients. And Agbo were very receptive and very like ahead of the game than everybody else. They knew that AI can potentially be a thing and they wanted to adopt it in the way will not hurt creators.
Rob Kelly:Can you share what's sort of famous about Agbo and
Tommy Petrov:what Oh, they The Russo brothers created Avengers, Endgame, Captain America, Everything Everywhere All at Once, amazing movie that won Oscar, and almost every blockbuster from Netflix, to be honest, including The Electric State and Gray Man and many, many others. And the main reason to work with us is just because they really, really serious about IPs, and they know that it's inevitable for Hollywood to go there. Hollywood is desperate about AI because it will reduce their costs, and they need to have reduction of cost. And AI is part of it. And the safe use of AI without disturbing unions, without making people lose their jobs, is absolutely critical.
Tommy Petrov:So they cannot survive without AI, but also they cannot survive another strike. And I think that AGBO and other clients of ours are trying to make it really, really legit and soft without, like, taking someone's job and things like that.
Rob Kelly:And it seemed like if you look at the last wave that entertainment companies, content companies in this space that we're talking about sort of didn't get the Internet and the impact. You can give different examples, just streaming and, you know, music, Netflix, and, you know, they they didn't figure it out for a while. Right? Do you feel like with the Disney's and the Warner Brothers and the Sony Ventures and the Agbos of the world that you're working with, do you feel like they get AI? Is it now lesson learned?
Rob Kelly:They get it? This is urgent?
Tommy Petrov:They get it, but they're rigid organizations. They have their their processes, and they're talking about, oh, we need to move fast.
Rob Kelly:So if I'm hearing you right, it sounds like they've got the sense of urgency about AI, but there's still Yeah.
Tommy Petrov:There's still a huge company.
Rob Kelly:Move slow Yeah. Can be rigid and so forth. Just looking back, I mean, Spotify and Netflix both created new brands, massive brands. I'm sure any one of the big entertainment companies would have loved to have created that. Yeah.
Rob Kelly:Who's gonna be the next Spotify, Netflix in the AI space?
Tommy Petrov:Two things about Spotify and Netflix is both are very big exceptions. So nobody believed in what they kind of in their approach, and that's why they missed the point when they become really big. Right now producing 300 movies per year, it's like the closest competitor is 70. So they're ahead of all the industry, they are much more efficient, and they are like almost unreachable from the competitors.
Rob Kelly:But an AI movie company could crush that. They could
Tommy Petrov:can try.
Rob Kelly:10,000 in a year.
Tommy Petrov:Yeah. And then Netflix has already used the whole sequence of generating the AI and published after they released the show, they basically said, you know, there's fifteen minutes of train crashing into the building, it's all AI. And nobody noticed, nobody's paid attention, and they said, I know how much we saved, dollars 15,000,000, so let's go for it. And so they will adopt it, but you're right, the AI companies are coming for them, and there's some time for that to happen. So I do believe that companies like Promise or others can actually create something interesting, but I think we have some time before it will happen.
Tommy Petrov:So what I'm trying to build, I start to build my relationship with studios immediately from the day zero. So that's why we are talking to Sony, we are talking with Warner Brothers and working with them. They should not be scared of us, should not consider us as a competitor, they should consider us as the way in into this AI space. Why obsessed with this copyright ability of the AI? Because I really want to create a IP marketplace.
Tommy Petrov:The place where you can put your IP and monetize it. And this is the way for royalty distribution through those IP whenever a big model, big studio, big agency using it accidentally or intentionally, they're using your IP. And whenever scoring is high, you get paid automatically through smart contract, through the other means and tools. You just get this payment directly to you as an IP owner or the author that makes sense. It can sustain creators' life.
Tommy Petrov:It can sustain their well-being during the time when probably some of them will not have a job.
Rob Kelly:I like to end on the human humanitarian sort of topics. Just a couple things to get your take on. One is, will creativity or art, we can call it, will art, in your opinion, you're an artist, will it improve or decline due to AI?
Tommy Petrov:The value will grow. The value of a person, handmade art, crafts, and a certain thing that are really cannot be generated, cannot be done. Like I was giving this example, so the whole value, the biggest value influencer with 20,000,000 followers can give is a personal meeting. So influencers will become event agencies basically because they will organize events because people will want to have a human decent connection. Because we cannot sympathize to AI generated content when we know it.
Tommy Petrov:When we don't know it, it still will be a big discussion when you don't know if it's AI generated or not and you still sympathize to whatever is human being on the image, on the video. With that said, I think everything that craft warm and then real will make more sense and more expensive. And I love this joke in TikTok, like the person in 2035 going to movie and he says, do you want to AI generate movie or real one? Oh, I want AI generated, $10. Do you want oh, I actually want the real 1, $300.
Rob Kelly:What advice do you have for kids? You know, those in elementary school, middle school, high school, maybe even college these days, but young kids about using AI? My advice will be don't train model.
Tommy Petrov:Train your brain. Train your LLM, your brain, and use it for your authentic curiosity. So I believe that every child, every young person has authentic curiosity and it's different than just curiosity. It's this thing that drives our knowledge and drives our passion. You learn about dinosaurs, you know, like all every boy wants to learn about dinosaurs or about pirates, or about whatever, and then you ask, oh, why they were killed?
Tommy Petrov:What's happened with the dinosaurs? Or why were they illegal, the pirates were illegal? And you just start to kind of ask questions, and this is like a Wikipedia of knowledge that we have access to, but we are not using it that way. We're not using it to feed our authentic curiosity, we use it to solve problems. And that's I think is the biggest problem because we will become dumber, that's for sure.
Tommy Petrov:And now I really want us to build guardrails in education for kids to use AI as the food for authentic curiosity rather than a thing that solve their homework.
Rob Kelly:Have you created an avatar of yourself yet using AI?
Tommy Petrov:I I have some in meta and some experiential ones. I have quite kind of avatar at at meta ecosystem. Not for now, not for, like, anything. I still think that the value of me being me, not perfect, not ideal, is quite significant and I sell this value for human connection.
Rob Kelly:Do you think you'll create one for when you're gone? I I Yeah. For family members, friends?
Tommy Petrov:For sure. And even even before I gone, I will create it to produce more educational content. That's for sure.
Rob Kelly:What didn't I ask you that I should have?
Tommy Petrov:I think the question we all have to ask so that the changes that AI brings, it's either you're on, either you're out, and now you have to find yourself learning the language of how to speak with AI. Because still, the best prompt engineer for legal AI will be legal. The best prompt engineer for design AI will be designer. So become this person who knows a language. How to ask the right question, how to get the right outputs.
Tommy Petrov:So become a better human and just learn the language of LLMs and VLMs.
Rob Kelly:I really enjoyed this, Tommy.
Tommy Petrov:Thank you. Thank you. Thank you for inviting me.
Rob Kelly:Well, this is Media and the Machine. A few things about you and me. If you wanna hear about the next new episode, make sure you hit follow on the show on your app. If you wanna go a little deeper, head to mediaandthemachine.com and subscribe. When you share your email with me, can see handcrafted transcripts, read the essays in my newsletter, and be the first to hear about who the guest is on the next show.
Rob Kelly:You can also email me directly from there. Maybe you wanna recommend a guest. I'll give you a shout out if you do. I love paying it forward. From time to time, I also open up office hours and host small meetups for subscribers, just to meet, talk, and build things together.
Rob Kelly:And if you're creating something of your own or thinking about it, I'd love to help. Maybe you've got a podcast in you. Finally, I don't have a marketing budget for this show. So if it's finding you and others, it's because someone like you passed it along. I'm genuinely grateful.
Rob Kelly:And if you have a moment, an honest rating helps me make this better for you. You just go to the show page and click one of the stars. And I'd rather you give me a low rating than no rating at all. I mean it. It pushes me to get better.
Rob Kelly:Thanks again, and see you next time.