This episode is audio from the How Do We Understand the Consumer When We Assess the Prospect of Competition in Fair Use? panel from the Engelberg Center's Conspicuous Consumers Symposium. It was recorded on October 17, 2025.
Announcer 0:00
Welcome to engelberg center live a collection of audio from events held by the engelberg center on innovation Law and Policy at NYU Law. This episode is audio from the how do we understand the consumer when we assess the prospect of competition in Fair Use panel from the engelberg Center's conspicuous consumers symposium. It was recorded on October 17, 2025
Christopher Sprigman 0:29
Okay, welcome to our panel on how do we understand the consumer when we assess the prospect of competition in fair use, and I've got a bunch of people here with me today who will tell us all about that. It's a great group of people for that purpose. I want to introduce them, starting with Eric Anderson, directly to my left, Eric's the former General Counsel at bytedance and was head of the IP group at Microsoft. To Eric's left is cristalia Garcia, the she is the Leo George professor of communications, entertainment and new media, as well as the Anne Fleming research professor and the Associate Dean for Academic Affairs at the Georgetown University School of Law. So to crystallias Left is Katrina Geddes, the assistant professor at Morris College of Law at The Ohio State University. And finally, Guy rube, the Vincent J Morello Professor of Law at the Temple University, Beasley School of Law. So let me just I'm not going to say much because I want to get to the panelists, but I just want to set the scene a little bit. So again, the question is, how do we understand the consumer when we assess the prospect of competition in fair use. So the prospect of competition in fair use? Well, that's been around from the beginning. It was an element of justice story's formulation of fair use. It's, it is the fourth statutory factor in Section 107 of the Copyright Act of 1976 which asks courts to actually directs courts to assess the effect of the use, that is the defendant's use, upon the potential market for or value of the copyrighted work. Now, from the beginning, competition is present in the analysis. There's a fairly straightforward aspect of this, the the effect of the defendant's use on the market for the plaintiff's work itself is assessed, obviously, but also there's a there's a second assessment built into that, which is an assessment of the defendant's use on potential derivative or licensing markets for the plaintiff's work that proves almost immediately in fair use analysis to be a bit of A tricky prospect, in part because there's this prospect of circularity there. You know, if there's a licensing opportunity, more or less automatically, if you just leave it as a tautology, the defendant's use might have an effect on that licensing opportunity if it wasn't licensed. So that's not very satisfying, and from the beginning, courts were looking for ways to think about that a little bit more subtly. Okay, so enter into the in the Campbell case, acre froze versus Campbell the Supreme Court's decision in that case. Enter in Pierre lavals ideas about transformativeness, which he laid out in that 1990 Harvard Law Review article toward a fair use standard transformativeness really changes, in a sense, the way we think about competition, at least for part of the fourth statutory factor a transformative use, courts begin to say following Campbell is not a use that the copyright owner has a putative right to own with respect to licensing. So in the fourth statutory factor analysis, when we're thinking about competition in the market. We basically just exempt competition for licensing markets. If it's a transformative licensing market, we essentially don't look at it. Is that an empirical finding? Not really. It's really a normative finding. It's not a finding that while these kinds of transformative works don't compete, it's really a normative decision that we're going to privilege them. Okay, fast forward again to the Warhol decision of a couple of years ago. Now this question of competition is being kind of promoted. It's no longer just confined to the fourth statutory factor. It's still there, obviously, but it's also been extended, some version of it anyway, has been extended to the consideration of the first statutory factor, and in particular the determination within that the analysis of that factor, whether the defendants use is, in fact, transformative. The court says the formulation, it's very long opinion, and Justice Sotomayor says this like three different ways in the opinion, but I will shorthand it by saying that if a work is transformative, it's less likely to compete with the plaintiffs, a use of the plaintiff's work. Now, you know, turn that on its head, that means that a work that doesn't really compete is more likely to be used by the defendant. That doesn't really compete is more likely to be transformative. I think that's what we can draw from this. Okay, so is that an empirical finding is. That is that a normative finding. I think that's a little bit of a different flavor from the fourth factor. It's a bit more empirical, right? They're asking empirical questions about competition. So, you know, I came to law. I came to IP law through antitrust law. That was my first stop. Remains one of the things I think about. And you know, antitrust law spends a lot of time thinking about the prospect of competition between different products and services, it occasionally thinks about this, even with respect to literary and artistic products and services. So maybe antitrust has something to say about this. We can talk about that a little later. Let me turn this over to Eric, just to get us started.
Erich Andersen 5:35
Okay, thanks a lot Chris, and thanks a lot to the engelberg center and everybody associated with it for the opportunity to be on the panel today, I wrote down a few remarks. I'd say these are more like scene setters, a way to sort of whet the appetite for the more specific topics that Chris mentioned, which I'm happy to engage on. I wrote them down so if you don't mind, I'm going to sort of, you know, read them out, because I think it'll be a little more structured in this way maybe easier to enter into. So as a background, as Chris mentioned, you know, I'm not an academic. I'm a lawyer who spent over 30 years working in house for tech companies, including Microsoft and most recently, bytedance. So that's the perspective that I bring to the conversation. So here's some top of mind thoughts that I have, some of which will will be resonate for people who are just listening to the last session that we had right before lunch, and also the excellent keynote topic that we had as well. So AI, image generation, for me, is clearly a hot topic in this space. I suspect a lot of people in the audience are familiar with Sora or similar tools, Tiktok, which I worked for, is flooded with videos from these tools in just the last few weeks, and the output is getting pretty good, if not always in the best taste, which was something mentioned in the last panel. I've seen pretty realistic looking videos showing Abe Lincoln wrestling, George Washington, for example, and Stephen Hawking engaging in various extremes towards drew Harwell of The Washington Post is calling this AI slop, which seems about right, but it's also, in my opinion, just the beginning of what we're seeing in terms of the kind of content coming out of these tools. So it's a little early to draw, you know, broad conclusions about the value of the kind of content that is coming out of these tools. Clearly, these tools were trained using copyrightable images, and the traditional entertainment industry is quite spooked by it, not to mention actors who are, I think, pretty up in arms. Disney has sued character AI to protect its rights, for example, and we'll almost certainly see new, important cases emerging from this area in the years ahead, perhaps new legislation as well to address the rights of human creators. So I'm going to put a stake in the ground and say it's going to be a challenge for rights holders in this space, in the current legal and industry context, amount effective infringement cases, as the majority noted in the Warhol decision, copyright is replete with escape valves, the idea, expression, distinction, the general rule that facts may not receive protection, requirement of originality, illegal standard for actionable copying, The limited duration of copyright, and yes, the defense of fair use, including all of its factors, some of which Chris mentioned, such as whether the amount taken is reasonable in relationship to the purpose of the use generative AI programs are extremely complicated and becoming more so as the science of AI improves. And there was a comment in the last session about the importance of understanding the engineering piece of this and I think that's this and I think that's absolutely true, and it's moving rapidly. The output of these programs is the product of dozens and sometimes hundreds of separate program nodes working together via an orchestration platform on data models with 1000s of pieces of content. I think it's important to understand the scale of the systems and technologies that are sitting behind tools like image generation, not to mention these large language models like open aI have produced. Often the output of these programs has to go through a kind of trust and safety check to avoid obvious infringement as well, and that's also something that is, I think, growing as a sort of a niche expertise within this frame. In other words, programs from top companies do not simply copy tweak and then output near copies of the original works, like in the Warhol scenario, although I can see that this may be the case with some companies participating in the current AI feeding frenzy. It also may be relatively easy for AI image generation program publishers to change their programs or their data models or safety check mechanisms if accused of infringement, it then becomes a kind of cat and mouse game on infringement and damages. Rather than a case focused on a fixed target. But let's look at the other side for a minute as a practical matter. In some instances, it may be in the best interest of an AI Tool Company, especially large ones, to be pragmatic and enter into licensing agreements to avoid legal and PR risks, even if they have a strong legal argument or can change their program to avoid infringement. And I think this is some of what we're seeing right now. Such companies may wish to avoid potentially greater risks, such as targeted legislation that expands the rights of copyright holders, especially if there's also potential flaws in the way the underlying llms have been constructed. Likewise, in some instances, it may be in the best interest of a user of an AI generation tool to settle an infringement case rather than fight if, for example, a user creates long form entertainment, such as a movie that cannot be easily changed after investments in marketing and distribution, it may present a fixed target more akin to Warhol, rather than something that can be modified. Let me briefly pivot on this topic and just touch on social media, since it's something I worked on for several years, many years ago, many music rights holders came out very strongly against social media companies for uses of music synced to user generated videos. There were a lot of tense licensing conversations, demands for payment, and even scenarios where entire libraries of music were taken down when licensing agreements expired. An interesting thing happened along the way, however, many creators of music, and this goes a little bit some of the economics, many creators of music, started to rebel against the music companies, as if to say, you're not representing my interests. The creators began to recognize the underlying value of the creators of these videos, celebrating their works, essentially in homegrown videos. It was driving fan interest in long form music and concerts. The most prominent of these rebels lives just a few blocks from here. Her name is Taylor Swift over on Cornelia street, she essentially broke from her company Universal Music and made a side deal with Tiktok to allow it to continue to make snippets of her music available for fans on the platform before she released a big album, the balance here was not really one between infringement and fair use. To be clear, the social media platforms, they pay licensing fees. They pay licensing fees to the music companies. So it's not a licensed versus non licensed scenario. But for the it was really about the underlying value of viral video content that drives long form entertainment consumption. So I think that's an interesting part of the consideration in some context, arguably in like the generative AI context, the the the payback, the reward, so to speak, for creators, may be marginal or zero, unless there's some sort of explicit payment that comes to them. But you have to sort of be careful about how broadly use that that brush, because in the case, in the case of social media, for instance, there may be a lot of value from what the platform is offering in terms of generating fan interest in the underlying, long form, monetizable asset. I was talking to my colleague over here before I got up on the stage, and you might think about it similarly to internet search, which I think is, by the way, something people should spend a lot of time. Thinking about internet search as an analogy to generative AI today, where there is a bit of a value exchange going on. Certainly, Google would say this, which is that, you know, I'm, I'm crawling your your copyrightable materials in order to be able to supply my search engine with information that is useful to users, and I'm monetizing that with through advertising. At the same time, I'm making your content searchable and discoverable by users which is valuable to you. So I'm draw I'm driving users to you so that you can monetize in whatever way you choose your underlying content. So I think it's really important to understand those kind of economic dynamics, because I think that it's not this sort of binary, black, white outcome. Sometimes there's a more subtle economic trade off going on. And I think that as we look at generative AI, for example, I think you might start to see a little bit more of this as well. Okay, great, Eric.
Christopher Sprigman 14:23
Thank you. There's a lot I'm going to want to get back to here when we have our discussion amongst ourselves. I should just note that Eric's going to have to leave at two, and we're done at 215 so if you have questions for him, hopefully we can, we can get them early in the question period. Okay?
Kristelia García 14:43
Crystallia, great. Thank you, thanks. It's very nice to be here. I'm going to spend my minutes here talking about the licensing market and some particular concerns I have about exclusive deals within the licensing market, particularly when it comes to AI and content. To tee those up a little bit so that we can pass off to Kat and guy to sort of continue the conversation. I want to start with just a little bit of background setting. As Chris opened with we had the Supreme Court's decision in Warhol arguably shift the primary Fair Use inquiry from transformativeness, where we've spent some time post Campbell to really focusing on whether the use at issue competes with the original use. So sort of first factor to fourth factor. And this reorientation, I think, really brings two antitrust concepts into the realm of fair use, market definition and market substitution. So post Warhol, then fair use analysis arguably hinges on what counts as the market for the work at issue, right? And that requires us to consider who the consumer is and what they're doing with the work, right? So I'm tying this back into consumers, wherever Michael is, so he'll like that, and the answers to these inquiries, I think, really do impact the outcome, right? So most copyrighted works have two markets, the market for the work itself, and then the licensing market for that work. And I'm going to focus on the latter. Unfair competition in the first market is pretty straightforward, right? But unfair competition in the second market, this licensing of the work, can be more complicated. So I want to pose a couple of questions. Does it matter? For example, that the magazine at issue in Warhol wanted to license a Warhol print for the cover, not just an image of prints, right? We would say the same thing if a film wanted to sink a cover of a song for its motion picture, not the original song, right? These are different uses is the magazine the consumer here, right? What are we looking at? And the obvious paradox here, right, is that any substitute market for a given work is the competition for that work, eliminating the substitutes, while perhaps part of the copyright bargain eliminates the competition. And from an antitrust lens, we tend to think that's bad for consumers, right? So with that brief baseline setting, I want to move into the AI space. So what about copyrighted content that's used to train large language models or llms, right? Is the AI company a consumer of the copyrighted work? So it's using them to train? Is the user of an LLM a consumer? Are they both consumers, but they're consuming this work at different stages in the in the the train. So I think that no discussion of licensing markets, which I promised I was going to focus on, would be complete without discussing the recent slew of exclusive licensing deals have been going on between big AI industry incumbents like open AI and media giants like news core, right at first blush, I think we could say that these agreements arguably are laudable, market based solutions that help to promote technological progress because they give AI developers some legally sanctioned access to high quality data which they need to train the algorithms, also ensuring, presumably, right, that creators and content owners are getting paid in some way, are compensated for the use of their work. If we take a more critical examination, though, I'll suggest that this, these deals also reveal some troubling potential for market distortion and competitive foreclosure. Right? Specifically, I worry that these high value, exclusive deals between established, big, well funded llms and major content owners threaten something that I'm going to call a digital enclosure loop. So this is, I'm calling this a self reinforcing cycle that entrenches the more incumbent firms, both the AI space and the content industry, stifling the potential for smaller entrants to come into the market, accelerating market consolidation in both already rather consolidated industries, and importantly, risking creating access bottlenecks that I think threaten copyrights whole constitutional bargain. How well for llms, the quality of the model is inextricably linked to the quality and scale of its training data, right? And in this context, not all data is created equally. As the public internet becomes increasingly populated with lower quality AI generated synthetic content, a phenomenon that researchers are calling model collapse, right? Where you get this degenerative spiral where AI is trained predominantly AI models that are trained predominantly on AI created synthetic outputs gradually lose accuracy and diversity, right? That looms as a significant threat, and this risk renders clean data sets, that is to say, human, curated and generated data sets, exponentially more valuable, and it's precisely these reservoirs of clean, human generated data sets that are now being enclosed in these exclusive licensing deals that we're seeing come out, effectively locking out upstart AI developers and creating this data moat around the AI industry's largest and most powerful incumbents. And I worry that. This dynamic can erect a formidable barrier for emerging AI firms, affecting not only prospective licensors as consumers of training data, but also user consumers of the content outputs, which will be lower quality, lower diversity, et cetera. So to illustrate this concept, imagine that we take a vertically integrated streaming platform like they all are now, right, like Netflix, meaning it's producing its own content and exclusively distributing that content. And then say it partners with an AI developer, or brings one in house, as it has So first it locks up its exclusive, self produced content. Right? The Netflix exclusives, films, TV shows, podcasts at this point, drawing in subscribers and simultaneously insulating itself from competition, right? This is the distribution mode, as I'm calling it. Then the users of Netflix, in this example, they consume that exclusive content. The platform gathers all the proprietary data on their habits and their preferences and their viewing. Those behavioral data sets are incredibly valuable for training recommendation algorithms and generative AI models. Then the platform's AI partner utilizes the platform's content archives and its proprietary user interaction data training even better algorithms they can formulate or even create similar content use behavior data to refine personalized user algorithms. Then the refined AI model can be deployed to enhance the user experience on the platform, give you more accurate content recommendations, offer AI generated subtitles or translations, even create AI generated content, like the content that you view, making the platform more attractive to users and then just reinforcing the dominance of The already dominant platform. Right? So to the extent that I am convincing you this might be problematic, what do we do about it? I think that the conventional legal frameworks that we usually use to balance intellectual property and antitrust or competition concerns are pretty ill suited to address the rapid formation of these data modes around dominant players, we can start with fair use. Right? Arguably, copyright's most critical safety valve. I think it's utility for AI developers, from what I've seen, is severely constrained, right? We can, of course, extrapolate from some opinions, but fair use as nature, is a case by case analysis and application, essentially meaning that if you want legal surety, you really have to vindicate it through litigation, and this is a process that we have seen is incredibly costly and time consuming. Look at Google Books right. It took a decade to get that opinion right, and we are not going to see AI startups be able to wait a decade for some some confirmation that what they're doing is okay, the normative trajectory of fair use in AI also to date, adds profound legal uncertainty. It's true that we have some early district court cases, whatever you make of those, like BARTs versus anthropic, right? That seems like it's saying, okay, maybe training is transformative, but also signaling that depends on how you source the data. I think this adds profound legal uncertainty for startups who are trying to get funding, leaving these new developers in a pretty precarious legal position where the legality of their core business practice hinges on this case by case analysis that is both costly and uncertain, and all of this worries me again, because I think it goes to potentially throttling access that I view as this essential part of copyrights bargain. We can think about antitrust, our conventional tool for combating market concentration, again, I think it's similarly ill equipped, primarily because of its debilitating pacing problem. The speed of technological change and the dynamics in this market far outstripped the speed of litigation, and by the time we see resolution in an antitrust case brought against a tech firm, it largely renders a new remedy moot, right? Like look at the FTCs case against meta. Right? It's challenging acquisitions that were cleared a decade ago, and long after their network effects have already made meta irrevocably the industry leader and all of these things, regardless of the outcome. Moreover, antitrust litigation in these cases is notoriously complex and bogged down in these foundational disputes over market definition, especially for multi sided platforms that serve users and advertisers, or in this case, creators and AI developers. So I think that antitrust really is this ex post remedy, right? That's not going to answer it. So what do we do? Well, that's the question. I don't have a definitive answer, but I feel increasingly certain that we should consider doing something to that end, I'm working on a project that's looking at a bunch of different ways that we might go to
Kristelia García 24:42
offer some anecdotes for these access bottlenecks tax incentives or reversion rights to compulsory Access Licenses, I don't think we're going to have a one size fits all solution to the access bottlenecks that I have described here, but I think that any proposed intervention should focus on per. Protecting three groups, each of them are consumers in their own right. The first would be content owners from uncompensated exploitation. The second would be the public from artificial scarcity that gets created by these exclusive deals. And finally, nascent AI firms and related entities from discriminatory leverage that can be exercised by these entrenched incumbents so that everyone can get access to the data that we really need to have a thriving AI economy.
Christopher Sprigman 25:29
Chris Delia, thank you, Katrina,
Katrina Geddes 25:34
thank you so so. Thank you to the engelberg center for organizing this impressive and important event. Thank you to Chris for moderating. Chris Delia, I'm just going to say that the whole time he was speaking, I was thinking that data moat is a great title for a law review paper. Please call it that when this eventually comes out. I'm going to begin also with their Warhol decision and how the court treated the concept of consumers within its own reasoning. So the majority treated the relevant consumer as a magazine looking for images of prints to accompany a magazine article about prints. And on that basis, the majority found that Warhol's use shared substantially the same purpose as Goldsmith's use, because orange print served as a commercial substitute for Goldsmith's photograph. But as Justice Kagan observed in her dissent, this analysis effectively transplanted factor four, which, as we've discussed, deals with market harm, into the court's analysis of factor one, which that decision was ostensibly limited to, and the majority, at least, according to Kagan, completely misapprehended how the relevant consumer would actually have perceived the two images, which was not remotely as substitutes for one another. So Justice Kagan gives the example that if you were an assistant holding up both images for a creative director, the director would, in kakins words, appreciate the Gulf in esthetics and meaning between the two works, between Goldsmith's photo, realistic image of Prince's humanity and Warhol's surreal, dreamlike, lurid, disembodied, iconic image of a celebrity that had been dehumanized by the publicity machine. As Justice Kagan observed to a seasoned magazine editor, these images were not remotely substitute for one another, but that's how the majority understood the consumer. So against this specific backdrop of what does substitution actually look like and how do courts interpret it, I want to discuss very briefly the treatment of consumers by recent judicial opinions on the status of unauthorized AI training. So in two judgments handed down this summer, two courts took very different approaches to the market substitution effects caused by generative AI models trained on copyrighted training data. So in cadre versus meta, Judge chabria argued that even if an AI model does not regurgitate the plaintiff's copyrighted works or generate substantially similar output, it can nevertheless produce works that are quote, unquote, similar enough that they will compete with the originals and indirectly substitute for them. He speculated at length about the different ways in which AI generated works might crowd out the market for human authored works. Do you think that we can converge on a definition of what that is? The specific examples he gave were biographies, magazine articles and romance novels. He argued that consumers would choose to consume AI generated romance novels rather than human authored romance novels, and that this form of market dilution or indirect substitution would be enough to defeat matters Fair Use defense, if only the plaintiffs had presented evidence to this effect, it didn't matter that Lama was incapable of generating more than 50 words from any of the plaintiff's novels. If it generated works that were similar in subject matter and genre, such that they competed with the plaintiff's works, that would be enough. So there's a lot to say about this, like novel theory of harm. I want to say specifically that Judge strawberry has clearly never read a romance novel. I don't personally know anyone who would think that an AI generated romance novel was remotely substitutable for any bardus Ripper by Julia Quinn and judge strauer did not present any empirical evidence for his speculations about possible future market dilution in Bart's V anthropic judge also took the opposite approach. He said that arguments that AI generated works would flood the market for human authored works were like arguing that training school children to write well will result in an explosion of competing works. And he explicitly said that this type of competition is not something the Copyright Act was designed to shield authors from empirically because this is an empirical question, do AI generated creative works crowd out the market for human, authored, expressive works? There is some early empirical evidence to suggest that this is the case. So there was a study out of Stanford by Sam Goldberg and Ty lamb that showed that at least in the market for stock. Images, the entry of AI generated images did lead to the exit of some human creators. But then, in general, the volume, quality and variety of images available on the market increased to the benefit of consumers, and the increase in image quality was driven not just by generative AI, but also by increases in the quality of non Gen I images produced by incumbents that remained on the market, indicating that low quality artists were exiting. Obviously, these empirical results are early. These are not generalizable to all markets, but they provide some interesting anecdotal evidence. Importantly, and I want to stress this, both decisions out of the Northern District of California did converge on some issues related to AI training. Both decisions held that the use of copyrighted works train AI models was highly transformative under the first factor, and also that the market for licensing copyrighted works as training data was not a market that copyright owners were automatically entitled to. So take what comfort from their conversions you will okay. What other evidence is there that consumers actually care about whether creative works were AI generated? So there are a handful of anecdotal examples that I think are interesting. Wizards of the Coast received a lot of backlash when fans discovered that illustrations created for a Dungeons and Dragons source book had been generated using AI. Wizards of the Coast subsequently committed that its artists would not use AI to create art for Dungeons and Dragons, and they commissioned new art to replace the AI generated works. Some creators have explicitly denounced the use of AI in their creative process as a way of both alleviating consumer concerns and maintaining market share. So cosmetic brand dove has publicly committed to never using AI generated content to represent real women in their ads. This is part of their commitment to, quote, unquote, real beauty. Disney scrapped plans to use AI in two of its upcoming films, the live action Moana and Tron Aries, because of concerns about negative publicity. Once the use of aI had been disclosed, there was also controversy around the use of an AI voice editing tool for two movies that won Academy Awards, the brutalist and Amelia Perez. And then finally, Natasha Leone faced a lot of backlash for her use of ethical AI, which she defines as copyright clear data in her upcoming film on county Valley. And I say all of this to say. And Jacob mentioned this yesterday, that I think attribution is important to a non trivial fraction of consumers, and for this reason, and when I say attribution, I don't mean just general transparency around whether and to what degree AI was used in the creative process, but the influence of specific training data on generated outputs. I will say that tracing the influence of specific training data on generated outputs is a very active area of research within the machine unlearning community, but a lot of technical work that's currently directed towards TDA is being funded by powerful rights holders who are very invested in being able to trace generated outputs back to specific training data in order to establish a per use proportional compensation model with their words. So Universal Music Group, for example, has partnered with an AI startup called pro rata AI, which has promised using their I think they the words were proprietary algorithmic technology to provide fractional attribution for generated output, so that musicians and other artists can get paid for every use of their work in AI generated outputs. So the question of attribution, if it's important and valuable, how can we, how can we legally compel it? Existing law does not require it, setting aside, you know, the provisions under the visual artists Rights Act and 1202 Congress could enact an AI specific attribution, right? And I raise this sort of as a question because I'm still normatively figuring out my position on this. The EU has done some interesting things in terms of mandating the disclosure of training data for AI models. So I guess I will conclude by saying, watch this space.
Christopher Sprigman 33:59
Okay, thank you, guy. All right,
Guy Rub 34:05
thanks for having me always happy to be here. And I want to talk again about licensing arm and fourth factor and the circularity problem that Chris alluded to also. You know, in abstract, I think everybody agrees, almost everybody agrees that some licensing arm should be part of the analysis of the fourth factor, fair use. And there are actually two types of licensing arm you can think about or licensing competition. One is when the plaintiff and the defendant are trying to license their creativity to someone else. That's the example in wall in the Supreme Court. So the idea there both a plaintiff and the defendant trying to sell their work, maybe allegedly in competing with each other over magazine. There is also the situation in which the licensing that we are talking about is actually the defendant work itself. The defendant actually themselves are the type we are trying to license it. Scrip code didn't really talk about that, but the Second Circuit did that. Actually was a big part of the Second Circuit opinion in walls, and this is another type of licensing arm, not the arm to a magazine, but the arm in fact, that most of those creativity, most of those wall prints were not authorized. The problem, of course, is if you take it to the logical extreme, you basically almost eliminate the fair use. Maybe you live. Maybe you leave a small island of criticism that cannot be licensed for various reasons, but for the most part, you're just eliminated almost everything. And it's circular, because if fair use both is one of the mechanisms to set the boundary of how far the plaintiff right extend, and also need that boundary to figure out fair use, then it's circular. Now many courts, not all of them, and I'll talk a few new decisions that seems to sort of run through the circularity problem without really noticing it. But many court and many commentators noticed that problem, and I want to talk about some of the approaches that those courts have taken to address that. But I'll start with the spoiler alert. I don't think that we have a good solution to this problem, and I think what courts are doing and maybe not telling us that that's what they are doing once they're using their intuition to a very large degree, think this is right or this is wrong, and using it almost like a conclusion, right? I mean, if I think, for whatever reason, that this is fair use, then I'm going to align, to say that, then the plaintiff does not have ownership right over this licensing market, and vice versa. If I think that is not fair use, that, I will say, well, there is clearly a license in art, because the plaintiff is the defendant is invading the plaintiff space. Some have said the transformative use itself as the same problem. So that thing now, I think there is another line here, and I think what other thing that I think play role in some of those decision is, let's not rock the boat. This is sort of a very sort of maybe conservative. You could call it a traditionalist view, saying, if that is what we used to do, then let's continue doing what we are used to. And I'll give an example in a few minutes. So what do they actually say that they are doing one thing that they some of those courts saying, well, let's see what market the plaintiff is using now. And if the plaintiff have concrete plans to use another market. Then, by all mean, let's use it a then debt count that the old decision said that over time, we have this formula that coming from the textical opinion that use that says we need to look whether that market that the plaintiff is arguing that they are entitled to a to use while the defendant is sort of invading his traditional reasonable or likely to develop. So, I mean, the good part about that, at least it's recognized the circularity, and at least it tries to give us some rule to deal with that. The problem is, of course, that this rule is extremely vague, and if it's taken to the extreme, it's also extremely, extremely broad, and what is traditional, what is reasonable and likely to develop, is really the kicker in which you can expand this to encompass almost everything. So a subset of that test is looking at what market can exist. This is, I'm thinking about it as a Wendy Gordon type argument, transaction cost. Let's liquid this action cost. See if this market can even can we even figure out that market? So we can think about how many licenses do we have, and do they have a collective action, collective mechanism that can negotiate together and reduce transaction costs? Do we have an automatic system that can do that. And in cases, do you see cases, some cases do that? And can we distribute the fee properly? And actually the machine learning cases, going back to at least Google Books, and certainly the new AI cases actually talk about that. And in the briefing those mechanisms, those topics are discussed at length. For example, in the meta litigation, there was a long, a lot of discussion there whether meta tried to get a license, but those attempts failed, and so the parties spend a lot of real estate in their brief trying to say, Okay, why did it fail? With meta? Arguing, well, it can succeed. It can succeed, because we obviously cannot negotiate with individual offer and the publishers do not have the right to license it for us. So we don't have those institute that can reduce our transaction costs. Now, of course, you can argue it on the other side, and we have those dealers, as Christiania mentioned, so these markets are sort of start to shape up. So if that's your inquiry, then you get to a different situation, and you get those barrier to entry problem that Christel point out. You know, Jacob has a paper about that. Now, in other case, we see other approaches. Chris alluded to that in like in Bill Graham, and we see the court saying, Well, if he trusts.
Erich Andersen 45:00
Like today's set of problems, where the reality is, you know, these things can can change in an instant if you just change the underlying program that generates the output. Except in the case that I'm that I mentioned, you know, interestingly and enough, if you create long form, you know, work, but these, let's say it's not copyrightable because it was generated by an artificial intelligence program that where the law might conclude that output is not protectable by copyright. But then you have to, like, look at the broader economics of the situation where, if you have a company that's invested in the distribution and the marketing around that work, even if not copyrightable. There is an economic stake, so to speak, around that, around that work. And then now you've got a right controversy, because if someone sues for infringement, you know, you may be sitting there thinking, if you're the person who who published the output, you know, you have to deal with it on that basis. Like now I've got, I've got a problem I got to deal with, even if I think that the person asserting the claim has got no basis for this claim. It's a risk for me, and I've got to sort of deal with the issue as it comes to me.
Christopher Sprigman 46:07
Okay, so crystally, I want to get back to this question of competition. And you essentially say, and I think it's right. So I wrote a short article on this, basically saying, so the Supreme Court, I think they have to understand this. They have stepped into antitrust comfort zone, right? They have promoted competition to the first factor analysis to determine whether some use by a defendant is transformative. If the use is transformative, the rest of the analysis is going to go quite differently. It doesn't, doesn't mean necessarily that it comes out in favor of the defendant, but it's just much more likely to Okay. So competition plays that role, and there's a whole apparatus that antitrust has developed, as you started to sketch out for assessing the prospect of competition between goods and services in a relevant market. What do you think is going to happen? Are courts following Warhol going to kind of dutifully access and apply that apparatus? Are they? Are they going to, are they going to kind of wing it?
Kristelia García 47:12
I think my honest answer instinctually would be, they're going to wing it like just a cat talked about the Trav decision, right? Which I think is exactly that judge saying, like, wing it, and then like, going, like, continuing to fly right? With the whole, like, you know, if it just is in the space of someone who's looking for a romance novel, then you know, it would be potentially, what did you call it, substitutively similar, right? And I think I share the sentiment of this panel that, like no substitutive similarity is not substantial similarity. If it is, then we now have this new area of love, this sort of copyright antitrust hybrid beast that is, you know, introducing an entirely new market which is just perspective similarity.
Christopher Sprigman 48:02
Would that be better or worse for copyright? I mean, as someone who's worked in both, I think of antitrust as being just much deeper in terms of, you know, its achievements in kind of bringing about social welfare than copyright.
Kristelia García 48:16
I was going to say I wouldn't dismiss it right out of hand as necessarily bad. You know, we see some contribution in some ways, as you said, like antitrust is a bit more let's call it a more disciplined legal area, smarter law,
Christopher Sprigman 48:32
no more disciplined, I'm being provocative, but
Kristelia García 48:36
more disciplined area of the law. And so maybe bringing some of those black and white lines to copyright, which is infamous for its gray lines, could be nice. How well that will work outside of, like, you know, consent decrees, which, you know they've written about, I don't know, we haven't seen lots of like, Great importation of antitrust into copyright yet, outside of, like, the very limited context of consent decrees. But I'm not against it.
Christopher Sprigman 49:04
Okay, so we should return to this. We can. I just wanted to talk a bit about this market dilution theory because I simple question first, like, where does it come from? So does it have a basis in the statute? So the Copyright Office seemed to think so, although didn't really explain in any detail why. But where does it come from? Does it have a history in copyright, or does it have a history anywhere else? And what do you think of its overall prospects going forward?
Katrina Geddes 49:32
So I think it's so. I think it's a scary development. It doesn't appear to me to have any rational Genesis. It sort of feels like looking at market substitution, which, you know, copyright jurors have always looked at, but on steroids, because, as crystallia mentioned, if something can compete with an original copyrighted work by virtue of sharing subject matter and genre, and that's enough to defeat a fair use defense. I. Mean that just blows everything that we, you know, have developed over the last, however, many years, out of the water, because everything becomes substitutable. And I was thinking, as you were discussing this with crystalliam, maybe then we just have, like, very intense fights over, like, market definition, right? And like, how narrowly or broadly do we define the market so that we can actually think about substitutes? But yeah, to me, just answer. Question directly, it doesn't appear to have any rational Genesis. I think, I think, honestly, you know, there is a lot of ethical concern about what is happening with unauthorized, you know, scraping and training, and it feels intuitively morally wrong, because we have all of these, like, very deep seated notions about copyright law should reflect plagiarism norms, etc. But it, you know, it's, this is not what copyright was designed to protect. It's not designed to protect creators with this, like enormous shield against any anti competitive forces. So it, I mean, I hope it doesn't go anywhere. It's, you know, right now, it's just dicta, but we'll see. It's, it's, it is interesting that the copyright office and this decision had the same approach.
Christopher Sprigman 51:04
So can I say I just, I just wanted to get in a little bit on this, because I think there's an intuition, and like a lot of intuitions, I'm not sure it's right that people have, but it's even if it's not right, it, I think it has some power, which is something like the following. So think about the camera's introduction, right in the 1850s and if you read back, you see painters saying, This is bad for painting, right? This is going to destroy painting. This is illegitimate competition from people who have no skill, right? This is a machine creating what looks like art, but isn't really art, okay? So we don't really think of cameras that way anymore. We think of them as a tool that artists use, and so that that that has been settled, but, but, you know, when I when I retail that historical episode, what I get, the reaction I get a lot of times is, yeah, but it's different, like a camera is being controlled by a human, and the camera is being used by human to create art, and AI is not being used in that way. AI is different. And so there's an empirical question, is that actually right? And then there's a deeper question, which is, assume that's right for the moment. What difference does it make to this argument? Anyone have any thoughts about that?
Guy Rub 52:20
I'll jump in about that. I mean, I I went back and forth about those intuitive because when I read that opinion says, and the copyright office before it said, This is not what fair use has ever been. And if you ask, what is its rooted in? The answer is nothing. It's new. I'll give credit to the judge, he sort of admits that he sort of said nothing that we have done is the same that being said, I think. And there is the empirical question, let's, for the sake of argument, assume that empirically the judge is right, which I think is tricky. But for the sake of argument, that copyright that AI does significantly affect the abilities of human creator to make a living, presumably. Let's assume that it's right, and then I think you go into a lot of intuition that feeds into what the judge is saying, which I think is sort of saying the quiet part out loud. I think many people are concerned about that and concerned about arm that is not what we think about copyright. So we are using copyright to achieve labor policy, which is usually not what we think, but we use copyright in the past to achieve privacy and other policy, and usually it doesn't work very well, but it's not the first time that we are trying to do that. When I when I talk about that with my students, for example, I thought we talked about the camera. We talked about recording music is going to kill the livelihood of performing artists. We talked about how the all those arguments have been made, the TV is going to kill the movie industry, and the Internet, of course, is going to kill everything. And time and time again, it's proven to be wrong. The counter argument that I get, and I think that's the argument, I don't know. I don't know if it's true or not, that this is singularity. AI is different because of quantity. The argument is not quality. The argument is not that because the human action is slightly different. Is because the reduction in prices is so dramatic and the reduction in the increasing quantity is so dramatic that that makes all my me trying to convince my student that we should be careful and giving this Sony space for new technology before we judge new technology. I get that pushback from my student, and I think that what I think in the back of the mind will will affect many judges. I think many judges, and in an area that is so much of so much room for intuitive judgment. Yeah, use. I don't know if you can stop that intuition from getting into the opinion.
Christopher Sprigman 54:54
So I have to say, so I'm teaching trademark this semester. And, you know, I think about trademark. We. Maybe less often than I think about copyright, but still quite a bit. And one thing that really strikes me about trademark is it doesn't have the same kind of encompassing fair use doctrine attached to it, although it has specific ones, but it does have an encompassing penetration by competition principles, the penetration is imperfect. Sometimes trademark pays attention to competition concerns. Sometimes trademark does not crystallia. You said, Look, fair use not a solution. Antitrust is not a solution. Might a solution be something like in every decision we make about copyright, from now on, we are importing competition principles. So if, if a fair use decision implicates the ability, for example, of the big AI in Commons, to construct and defend data modes, then that is going to shape the fair use analysis. After all, Congress said, this is a this is an equitable rule of reasoning. It's not a four factor test, which I keep telling my students, anything and everything that you can make a plausible argument is relevant, can be included in that analysis. So is that the route
Kristelia García 56:03
I would, I would be in favor of that, and I actually think that would put us in a much better place than we are now, with the sort of, you know, feeling like we're locked into these four factors. I tell my students the same thing. They're a guideline, perhaps, but, you know, you could shove anything you want into these, right, this, this bucket, and I think that it would be both logical and laudable to include these competitive principles in those analyzes, both in the context of, you know, AI in these data modes that I discussed, but even more broadly, I think there, you know, other areas where we could, we could use these questions. Again, my only worry there being that, you know, are they are the courts going to get it right when they start importing market substitutability? Are they going to understand what, when and whether something is a market substitute? And I think I agree with Kat like, maybe we need to play with market and how narrow and how broad which trademark does a lot,
Christopher Sprigman 56:56
so how narrow and how broad is part of what I want to end on, because I want to get plenty of time for audience questions here. But let me just pull back for a second. So the Warhol court presents us with competition as this now criterion for transformativeness. So it's it really has risen in importance. But I mean, let's resist a little bit. You know, the Supreme Court doesn't have a very high approval rating, so I think we have license to resist. Is this the best way to be thinking about transformativeness? So, you know, several of you have made the point that these are very different artistic works, right? There's, you know, you guys talked about this at length, that there's a difference between this naturalistic photograph of Prince and this highly constructed, right, iconic image, is competition really what we should be talking about here? Or is there? Is there some other nettle we should be grasping?
Erich Andersen 57:55
I'm not. I'm not a fan of abusing competition law in this context. I'll just say for two reasons. One, you know is, is that it strikes me, I, you know, as being you know, are we? Are we judging the output in competition with the protectable element of the unprotectable element? We put that great question that crystallized for me, but the way I was thinking about it, which is that I think it's a false positive, or could take you down a very strange road. The second is, and this is, this is me coming from, you know, from industry, so to speak, which is, I think you've taken a and this is not a reason not to do it. I'm just saying I think you've taken a case now, which is, is cost x, and you've turned it into a case that cost 10x because that's the difference between antitrust case and a copyright case. And you talk about the Moats. Talk about a moat in terms of, like, you know, disincentives for litigating things, you know, turn it into an antitrust case, yeah, and they've got a big incentive not to, not to pursue it instead, you know, you know, the game Go, go talk a regulator into pursuing it, because it's impossible. They have a private antitrust case. So anyway, I'm not sure that that's that helpful. But I think the touchstone, I worry about that as being like, you know, I think that's the realm of damages, not the realm of liability. And I, you know, in the in these kinds of cases. And so I worry about that,
Katrina Geddes 59:19
can I make a quick turn on this? So the idea of relying like, sort of focusing predominantly on competition, also, I agree that strikes me as strange. Because if the counterfactual is like comparing an AI generated work to a human authored work, like, if the counterfactual is everything is made by humans. That's not the world that we exist in anymore. Like we haven't existed in that world for a very long time. Like almost every creator relies on some form of technology. Like every 18 year old who's like creating videos on Tiktok is using AI. All of those videos are AI assisted creative so because I don't to the extent that we could even reach a consensus on what human author. Considered, like, how we define that, I don't think that's the thing that we're competing against. I think everything that is human authored will be in some way technologically assisted, or AI assisted.
Christopher Sprigman 1:00:09
Yeah, the more I think about this, and let's get audience questions, the more I think about this, the more I think underneath all this is is a very deep resistance to dehumanization in the creation of artistic and literary works that this isn't just about AI, it's also about us. So we think of our capacities as being, you know, what makes us unique as essentially tied to this creative ability that we have? AI is not being creative, but it's create. It is making producing plausible artifacts of creativity. And we're confused about how to react to that. I think it's, it's deeply unsettling, and copyright law may be, may end up as the framework through which we talk about this, which is, I think, the saddest thing about this, because the copyright law, again, I'm going to say it is not particularly deep, and it, it is not a good vessel to talk about these things. All right, so I'd love to invite some audience questions.
Jacob Noti-Victor 1:01:02
Hi everyone. Great, great panel. I want to pick up on what guy said about the saying The quiet part out loud, and a little bit on what Chris and Chris Delia were talking about, about incorporating more competition discussion into the fair use inquiry. You know, something I always, I used to like about fair use, is that it was an opportunity for judges to sort of say the quiet of say the quiet part out loud and actually talk about transformativeness, and do this sort of direct balancing in a way that we don't often see in outside of, you know, equity analyzes and other areas of law. But like the chabria opinion, you know, makes me think that this is maybe like a bug actually, that it gives too much license to judges to reinvent the wheel. I mean, I liked it when Judge Lavalle reinvented the wheel, but I don't know if I love it when Judge tavria is reinventing the wheel. So is a possibility here to retreat, maybe away from fair use, back towards doctrines and copyright that function more as kind of bright line rules. So we actually saw avenues for this in Warhol. We could have resolved Warhol as substantial similarity. The two images are not substantially similar in the air, in the fair, in the AI context, you know, or in bracha has suggested that this type of use is just inherently outside the scope of the reproduction, right, much more of a narrow way of define it, you know, a way of essentially making a policy judgment, but tethering it to to a narrow, scoped entitlement, rather than through the whole sort of antitrust style, capacious rule of reason, style, policy balancing that we're seeing in fair use. So is this? Is this a better way?
Guy Rub 1:02:32
Perhaps, I'm not sure. I think it's different. I mean, I have the same intuition as you when, I mean, he's saying the fight part out loud, but it's also something that it's almost separation of power issue at that point, right? I mean, we are letting the judge with various opinion completely rewrite labor policy, and how do we treat with I mean, that the phrasing of that dictum is extremely broad, yes, is realigning the entire area of the field. Usually, I'm not even sure the court should do that, and court under fair use. Yeah, it felt uncomfortable, I think, for many of us to see, to see that kind of legislative fair use power. But I'm not, I'm not sure what is the other solution? I think those solutions can work in something. I mean, the substance similarity, maybe could have worked in war, or maybe, I think that the Supreme Court made a big mistake of accepting this case just as the first factor a third petition, the meaning, which would be the entire Fair Use inquiry, maybe with substantial similarity. And I think it would help, would have helped the judges see the full picture that I think many of us have issues with. The one thing I would say to, maybe to for world, I think the test that we use a lot before World, which is new message and meaning, was extremely vague, also fair enough.
Speaker 2 1:04:16
Thank you. Thanks so much. This is a fascinating panel, and leaves me with many more questions than answers, as any good panel should I'm speaking a bit from the outside, as an art historian without any legal training, but I'm involved in these questions, and I wonder whether there is another problematic intuition that underlies the Warhol decision that may relate also to our unease around AI and questions of fair use. And that would be, I think, a deep seated to use a biblical illusion to give a sense of the deep seatedness of this intuition is the prohibition against boiling a baby goat a kid. Its mother's milk, something about that, just like you can have goat and you can boil it in milk, but in its mother's milk, it just doesn't seem right. And when photography is introduced, it completely wipes out. Miniaturist painters like that. They all become photographers because nobody's going to pay for a miniature painting any longer. So they're wiped out. They basically cease to exist. Some become photographers. Others do whatever they do, but they're not. The key thing is, is that photography didn't use miniaturist painting as the basis for the new technology, and eventually the studios concede and put feature films on television. It takes a little while before the good films Come on, but they're all licensed. And it's not that TV built itself on the back of existent, a major Hollywood investment in film. I think where Warhol, the Warhol decision, and where AI the same, where there may have a similar intuition is here is a famous artist who got this photograph, paid for it, then uses that photograph to compete in the exact same market that that photographer competed in. Now, as an art historian, I know that the photograph and that the silk screen are not the same. I don't you know, there's no question, but they're competing in the exact same market. And it's, it's very similar to, I think, what we're seeing now with with producers, with artists of various kinds, who are seeing their own work being used to compete with them. And that's different than the steady march of technological progress. That's, it's, it's something icky and just that feels morally repugnant in a certain way. Which is why I think that the Warhol case never should have gone to the Supreme Court. It was. It's a terrible decision, a terrible case in many ways, and not really relevant for the vast majority of copyright decisions. But interestingly, I think, surprisingly relevant impression, even for our AI predicament when the training on artists' own materials is then used to compete against them in the very same market. And that ick factor, that ick factor, I think, is going to be something we have to address one in one way or another.
Kristelia García 1:07:23
Yeah, I think that's a really insightful intuition. I've spent a bit of time late, of late, thinking of that. Like, I recall when they the image generator, OpenAI put it out, and, like, part of the splash was Studio Ghibli. You could say, you know, you wanted it, and then it gave it to you. And like, how did it do it? Because they trained on all of it. And so I think, you know, to put it a little more colloquially, like to add insult to injury, like, not only are you taking something from a from an artist who said, I don't want you to and then splashing it and saying, Look, we can do you. We can do your exact style, right? It does have the ick factor to it. And, you know, ties into all the stuff that all of us work on about, you know, the human element and the copyright office's ploy that, like, Well, what we really want is, like, you know, human is the ultimate output imbued with human creation. You know, how big of a role did a human play? And I think, to go to Chris's point to tie it to photography like it's so clearly parallel, right? When the camera came out, the machine fixed it. You just pointed it at an apple, and so you gave it the idea of the apple, like the prompt, right? But what if you do 6000 prompts, and what if you do other things? And you know, when does it become the product of human creation? I don't think we have a good answer. But I definitely think that that, that that insight is, is, you know, paramount in the AI discussion, and played a huge role in Warhol, which was, you know, bad on the facts, for all of the reasons you said, Yeah, you guys want to, I
Christopher Sprigman 1:08:53
would just say,
Guy Rub 1:08:54
I think it's a wonderful insight. I think also the idea is that the new one is making more money, one that the world is making more money than than Goldstein and AI companies making billions of dollars. That clearly is one of the driving force
Christopher Sprigman 1:09:08
of the ickiness. Ken, yeah,
Katrina Geddes 1:09:11
so I actually was gonna say that I, although I'm very sympathetic to artists, I don't, I don't think the ick is, like, particularly original in the sense that, like, you know, I could spend all my free time, if I was inclined in this direction, like listening to, like Taylor Swift songs, so that I could write music that sounds exactly like hers, and it would compete in the market for her, for her music. You know, again, if the super theoretically, super, theoretical experiment. And, you know, there are, there are great creators who've trained assistants to replace them. So I guess I'm saying that, like, just because you learn from the way that someone creates, and then you create things that compete with what they originally created, is not, I don't think it's inherently wrong. I think there are different circumstances in which people would be okay with that, depending on, like, Did you credit me? Did you give me some form of compensation? Like, I don't think I would cast it all generally as like, icky, because I think it's really nuanced. And very quickly in response to this question you asked at the start, like, I think a lot of the intuitions we have that like aI generated art is not truly art, or it's not truly human, authored can sometimes approach like very ableist arguments. I think what people are doing with AI, in some cases, is really extraordinary and really compelling and powerful, and like the camera, it can be an incredibly empowering tool. That's not to say that, like aI slot doesn't exist. It does, but there are different categories, right? So I'm not going to, I'm not going to consume an AI generated romance novel, because I know at least now it will be bad. But at my entire Pinterest field, my entire Pinterest feed is AI modified, and I love it like it's amazing. So I think, yeah, I guess I would, I would encourage nuance relative to ick, because I don't think ick is is as universal as as it I want
Christopher Sprigman 1:11:05
to get more questions in, but I also know we got to talk later about competition, because an antitrust lawyer would look at Warhol and the print and the photograph and say, Well, if the photographs, price gets raised by 10% would people substitute the print or the other way around? I mean, and so it's not even so coke and milk? Do they compete? No, right, in a sense, they do. But this is always a market definition. It's all so it goes from like, your intuition, the Supreme Court's pointing in a way that I think is based on that intuition, but their reaction to it is to invoke that something that's like hyper technical and not driven by intuition at all. So it's a very strange move that they make
Chris Cotropia 1:11:47
just to piggyback on, I think the enlightening kind of conversation, I mean, we have this intermediary kind of copying machine, but in the end, this could just be a substantial similarity argument, right? So I'm the person who's got the camera. I'm looking at the miniatures painting and saying, you know, I'd really like to create something like this. Take a picture of Chris Brigman, and then the way we would typically play that one out as well as their sub was there copying, yes, I had access. Is there substantial similarity or not? And this gets back the whole question whether we have this exceptionalism. You know, we kind of played this out into the internet with Netcom, and said, Look, there's this intermediate copying. This is just the way the tool works. No, no, no, no. And so maybe I'm pushing even further on, like, what is the differential? Because I have a little bit of, I kind of agree with cat and that, you know, what is the difference? I mean, other than the fact I've got this tool that's going to help me do it really well, right? And maybe we should just be looking at the outputs and substantial similarity, and this intermediate copying stuff is just net com all over again. So I just just comment. But interesting to hear any reactions, but thank you.
Christopher Sprigman 1:13:01
I mean, you guys, I think we've crossed a frontier. I mean, in the sense that cat's right, like all around us are people making art who, in a sense, are free riding on the earlier efforts of others without compensation. So people create art within a recognized genre. There were genre originators who originated that genre, and their debt goes unpaid, right? And that's kind of the way it is. Occasionally people complain about that. They say, you know, my particular style should belong to me. And most of the time, not all the time, because nothing's true all of the time in Anglo American law. But you know, most of the time courts are like, No, thanks. Right to that claim. I think the difference here is that this is the essence of the machine. The machine is doing something in terms of, you know, creating statistical representations of reality that makes in a weird way, because people always say, AI is so mysterious, mysterious compared to what like inside our brains, is complete, completely mysterious. When we experience reality, we may be creating something analogous to a statistical model that allows us to replicate, if we have the skills, the ableist argument that cat alludes to, I think, is actually part of this, right? Some of us are able to reproduce that in the form of art. Many of us aren't. And this is the complaint that artists had about the camera, right? It was like enabling people, and that was a threat, a competitive threat, but you know, part of the problem, I think, is like, we don't know what's going on for us when we reproduce. I have a feeling we're doing something like we're integrating all this information and creating variations on it. That's what AI is doing. But we're, you know, we're creeped out when a machine does it.
Guy Rub 1:14:38
I think it's quantity. Yeah, quantity, quantity, Kathy's right, that I can learn from other and do it myself, but there's a huge friction there that slows me down. And without that friction, that I'm not slowed down, intuitively, it seems different, and I'm not, doctrinally, it's probably not.
Katrina Geddes 1:14:57
So, yeah, Can I say one quick thing? Least, yeah, so I agree with guy that I think the this, the sheer scale and ease of competition, is really striking and makes this novel. Interestingly, I think it is the ick that's slowing down the competition, right? Because some people really do care about whether or not something is AI generated, and they will not consume something that's if they know at least it is AI generated. And I think, like, serious markets will develop around, hey, was this human created? However you define that, in the same way that, like, mass manufacturing has created, like, very robust markets for things that are handmade or artisanal. So I think all of this is going to develop just organically over time, if
Christopher Sprigman 1:15:39
Copperhead let us get in the way, so are we out of time? We are sorry. Please join me in thanking the panelists.
Speaker 1 1:15:51
The engelberg center live podcast is a production of the engelberg center on innovation Law and Policy at NYU Law, and is released under a Creative Commons Attribution, 4.0 International license. Our theme music is by Jessica batke and is licensed under a Creative Commons. Attribution, 4.0 International license. You.