BRIAN KENNY: Welcome to Cold Call, the podcast where we dive deep into the groundbreaking ideas behind Harvard Business School case studies. Artificial intelligence is no longer just helping people create. It’s beginning to replicate some of the most valuable forms of human expertise. As AI systems become capable of reproducing our voices, faces, writing, and even professional judgment, a new set of questions comes into focus. Who owns the digital version of you? If your experience and talent can be transformed into data, who should control it, profit from it, or decide how it’s used? Today’s conversation explores the intersection of technology, economics, labor markets, and intellectual property through the lens of one company working at the forefront of AI-generated media. We’ll examine why Hollywood may offer an unexpected blueprint for the future of work and what that could mean for every profession.
Today on Cold Call, we’ll discuss the case, “Metaphysic AI: Rethinking the Value of Human Expertise,” with Professor Zoe Cullen. I’m your host, Brian Kenny, and you’re listening to Cold Call on the HBR Podcast Network. Zoe Cullen studies the design of labor markets and the choices of employers and labor platforms that affect matters of public interest. That sounds like a perfect range of things for this conversation. Zoe, welcome.
ZOE CULLEN: Thank you for having me.
BRIAN KENNY: You’re a first-time visitor to Cold Call, but given the kind of work that you’re doing at the School, I suspect we’ll try to get you back on the podcast at sometime soon. So I think this will be a super interesting conversation because most people saw… The video that it opens up with, that the case opens up with, was really the first taste that a lot of people had about AI image generation and just how realistic it could be. And I think it was both thrilling, but also a little scary for people to see that because the implications are big and this case touches on so many of those. What drew you to the story? I’m wondering why you decided to write a case about Metaphysic and what they’re doing.
ZOE CULLEN: Well, I’m so happy you started with that clip because that clip tells the whole story. That clip shows how AI deepfakes don’t really require the consent of the creator, especially for someone like Tom Hanks, who has a huge amount of public footage out there. So Metaphysic is this AI deepfake platform that operates within the norms and contracts of Hollywood and created a movie, “Here,” which did actually also use Tom Hanks, but used it with his full consent and that came at a very high price. So I wanted to understand why is a company doing this? How is it good for business and how does it reflect on the contracts that Hollywood has put in place over the last 100 years?
BRIAN KENNY: Okay. But what’s your cold call when you start the discussion? I’m curious.
ZOE CULLEN: The class actually starts with Suno AI, the music industry.
BRIAN KENNY: Oh, yeah.
ZOE CULLEN: So students will produce a clip, usually about one minute long of their original music creation having explored Suno, the website that produces AI generated music.
BRIAN KENNY: It’s amazing actually. I’ve spent some time on Suno. It’s really cool.
ZOE CULLEN: So the class starts where students play their clips and they play their clips and everyone in the class has a chance to appreciate just what an amazing musician they’ve become in the last six months. And the reason why we start with the music industry and their clips is because they can see just how easy it was to expropriate the talent of many, many musicians in a 30-second clip. And not only can they expropriate it, but they have a way of commercializing it. And it’s very hard for them to see how or even creatively come up with a way for the original musicians in that piece to be paid for that work.
BRIAN KENNY: Yes.
ZOE CULLEN: And they also, in the context of Suno, they try hard, many of them initially, to replicate the voice of a particular musician that they appreciate. So they’ll go in and try to do in the style of Taylor Swift, write me a song about my own dog. And that process highlights for them that Taylor Swift is not consenting to this process. And the app will absolutely not have the intellectual property rights to create something in Taylor Swift’s song. Now that’s in sharp contrast to what Metaphysic is able to do, which is to bring in the very top talent and attract them towards creating high production films. And it’s the difference between the contracts these two industries have in place that leads to a situation where you have access to the very best talent-
BRIAN KENNY: Yes.
ZOE CULLEN: And a way of paying the very best talent.
BRIAN KENNY: And I think that becomes super relevant. So in the case of Suno, most of us are like, “Oh, write a song for my partner or for my child or whatever.” You’re going to use it that way. It’s going to be a very personal use. But when you actually take that and try to commercialize it, it becomes a whole different set of issues.
ZOE CULLEN: That’s right. That’s right. That’s right. But cleverly, Suno has been able to allow people who are casual users of the app who pay a subscription fee-
BRIAN KENNY: Yes.
ZOE CULLEN: To try to commercialize that. And the next lesson that students learn is they think of ways that they can actually bring in money for that new song that they made. And they come up with all sorts of platforms that they can host it on. They can host it on Spotify. They can host it on Instagram. They can share it with their communities. And then someone eventually in the classroom will recognize that if they were to be a hit, someone would copy them just the way they copied others.
BRIAN KENNY: Yes. Yeah.
ZOE CULLEN: And thus, the whole pyramid structure falls.
BRIAN KENNY: It opens up such a big set of issues. So one of the things that you do well in the case, I think most people, when they think about AI, the first thing they think about is, oh, it’s a technology or it’s a technology platform of some sort. But you’ve actually framed it up more as a story about the value of human expertise. And this is exactly what you’re talking about in the case of Suno. Why is this distinction so important between the technology and the value of human expertise?
ZOE CULLEN: In this case, many agree that the quality and the value of the technology hinges on the quality of the human data used in the training of that technology. And so the link between the human expertise and the success of the AI innovation is very tight behind the scenes. But many people are, at this moment in time, still unaware of the connection between, say, the surveillance that they have in their job and how that data stream is so pivotal for the ability to replicate some of what they’re doing.
BRIAN KENNY: It’s interesting that Hollywood becomes kind of ground zero for the argument that you’re making in the case. Why does Hollywood make such an effective platform to talk about these kinds of issues?
ZOE CULLEN: It all starts in the 1930s.
BRIAN KENNY: Ah.
ZOE CULLEN: So Hollywood in particular, let me focus on the Screen Writer’s Guild for start. The screenwriters have always had their labor developed in the form of a physical asset in the sense that a good screenplay or even a mediocre screenplay takes the form of something that can be copied, modified, scaled. And so early on, they realized that if they were to pass off their screenplay to say the executive producer, they’ve lost control in some important ways. And there were lots of stories of people higher up in the chain for the production to just rename who wrote that script, make a few modifications. Sometimes they would give credit to their wives.
And through this process, screenwriters grappled with some of the issues that AI has extended as an issue to many more professions, which is what do you do when your labor can be a physical form? And that means that you could potentially be, your very expertise, your ideas, can be disembodied. And that’s why Hollywood is such a nice place to turn to because it’s a profession that grappled with that issue 100 years ago. And I can go into lots of detail about what I think they did in those early days that led to a company like Metaphysic operating so ethically within the norms and confines of this industry.
BRIAN KENNY: Yeah. So let’s go back to Metaphysic for a second. What are they doing? How are they approaching this differently than some of the other AI technology platforms that are out there? The case talks about the fact that they’re really building their model around consent, ownership, and compensation. Are they sort of breaking new ground with this? Are they kind of setting a new bar?
ZOE CULLEN: Yes. Imagine an AI deepfake platform that’s trying to itself go and copyright and get protection for those deepfakes, for the individuals who are creating those deepfakes. And so sorry, I’m sort of abusing the word deepfake. So let me use a different phrase here. So this is a company whose technology could be used to warp reality in many different ways.
And therefore, it’s very easy from the perspective of the technology to imagine obscuring the creators behind the film. And that is what many other AI companies have taken advantage of, that you can combine many different small pieces of human expertise to create something for which finding the origin is challenging and therefore probably many little micropayments are not either necessary or going to happen. So Metaphysic went ahead in the other direction, which was to say we really want to make sure that the creators behind the film have their intellectual property, both protected, preserved, and they are going to provide kind of the legal support for that themselves.
BRIAN KENNY: Okay.
ZOE CULLEN: They show up in Washington and try to copyright the likeness of individuals who are using their technology.
BRIAN KENNY: So when you say the people, the creators behind the film, how does that play out in the case of Tom Hanks, which you mentioned before? Does Tom Hanks in any way benefit from this or is it just the folks at Metaphysic who are creating that content?
ZOE CULLEN: So Metaphysic could have made a movie about Tom Hanks without Tom Hanks, technically speaking. And they could have made a movie about Tom Hanks with just Tom Hanks and no one else. But instead what they did was they really involved the creators at many levels. So it was not just Tom Hanks who ended up getting paid and fully onboarded to this process of creation, but also the studios that were involved in the early footage of Tom Hanks.
BRIAN KENNY: Ah.
ZOE CULLEN: So Gracie Studios who created Big also was at the negotiation table when it came time to make the most contemporary film Here using Tom Hanks’ likeness to track him between ages 18 and 80.
BRIAN KENNY: Right. So they basically created a film that built on the likeness of him in other films that he had been in throughout his long, decades long career, right?
ZOE CULLEN: Exactly. Exactly. But the technology technically could take old footage or even public footage and project it onto a face that just had similarities to Tom Hanks’ facial structure.
BRIAN KENNY: Yeah.
ZOE CULLEN: And even of course, like Tom Hanks’ face is particular to a time and place. And so the footage that they’re projecting to de-age him and all these things is already manipulating exactly what he looks like.
BRIAN KENNY: It’s incredibly complicated, right? It sounds like it could get very messy very fast.
ZOE CULLEN: Oh, you mean the technology of doing something like this?
BRIAN KENNY: Well, the technology, but also what Metaphysic is trying to do in terms of bringing all of the interested parties together so that people are properly compensated for whatever role they might have played over an extended period of time. It just sounds like almost impossible.
ZOE CULLEN: Okay. But this gets back to what happens in the 1930s.
BRIAN KENNY: Yeah.
ZOE CULLEN: So the very first thing that writers pushed for and other guilds within Hollywood also understood to be critical is credit attribution, meaning that there’d be a documentation of who contributed what that could be preserved over time.
BRIAN KENNY: Okay.
ZOE CULLEN: And you could always go back into what, it could open up the black box of a project and see who contributed what.
BRIAN KENNY: Is this the concept of separate rights? Separated rights. I’m sorry. That’s one of the concepts that comes up in the case. Is this what you’re kind of getting at?
ZOE CULLEN: So separated rights is built on top of a credit attribution system.
BRIAN KENNY: Okay.
ZOE CULLEN: So imagine you have a dozen different versions of the script that went into making The Hulk. Now there’s a peer review process which was negotiated by the Guild. So when the Guild formed, they took over the process of allocating credit. Through that peer review process, the person who came up with the word “Hulk” has assigned to them their share of the production process. And all of those data go into an archive. There’s actually an amazing archive that the Hollywood Guild still preserves. In that archive, someone who wanted to reuse material from The Hulk because a new technology came out or they have a new audience that they have in mind, they are going to be able to find the original creators and understand who contributed what. Now I said separated rights sits on top of that credit attribution system.
BRIAN KENNY: Yes.
ZOE CULLEN: So let me just be really clear about that. So when the original creators of the screenplay for The Hulk were paid for the movie or the initial version of The Hulk, they were contracting over something very specific. It was well specified what would be the product and they could have in their mind an expectation about the revenue it would produce.
BRIAN KENNY: Okay.
ZOE CULLEN: Now in this interaction, that well-defined labor exchange allows for all other things to be rights that are returned back to the original creator. So any other use outside of that direct contract is a set of, it’s elegant in how abstract it is. All other uses of the material, those rights go back to the original creator. That means that suppose down the road somebody like Metaphysic want to make a deepfake of the voice actor from The Hulk, they will have now a new contract for new uses of the same data.
BRIAN KENNY: Yeah.
ZOE CULLEN: And those are rights that have returned back to the original voice actor.
BRIAN KENNY: Okay. Okay. So I’m starting to see why Hollywood and Metaphysic made this such a perfect case for you to write, to explain these pretty complicated topics. And it does go back to the ’30s. So that’s interesting. Hollywood had built a system that in some ways lends itself to what Metaphysic is trying to do. Are there other industries that you think could be facing similar sets of challenges and issues that maybe they’re not as well-equipped? Maybe they don’t have the same foundation that Hollywood has put in place.
ZOE CULLEN: Well, so the music industry is what we turn to in the context of class, but I actually want to broaden the scope because with the technologies that we have today, many new forms of tacit expertise are going to be in this physical form. And that means that say a doctor who has a particular way of diagnosing cardiac arrest, that way that they put together the little pieces of what the patient describes or presents, that pattern that allows them to diagnose, it can be now a physical asset that’s been mimicked by say AI technology. And now they have a contract potentially with a hospital to be paid for this work that they’re producing. Now suppose the hospital want to take that data, the disembodied form, and sell it more broadly, repackage it and sell it to say an umbrella system. So that hospital now didn’t actually include anything about this in the initial contract they drafted with the doctor-
BRIAN KENNY: Right.
ZOE CULLEN: Presumably, because the technology didn’t exist to do it. And the medical system doesn’t have in place the separated rights that Hollywood did. But you could imagine moving towards a world where we do have separated rights so that the physician who entered the contract not knowing how their data were going to be used can be included in the next iteration of negotiations for the use, the new uses of their data. Does that make sense?
BRIAN KENNY: It does. And so now I’m thinking this could actually apply to almost any field.
ZOE CULLEN: Exactly.
BRIAN KENNY: Right. And I’m thinking about your field, academia. I mean, part of what you do is create knowledge. And that knowledge belongs to you. You’ve created it, but it also belongs in some ways to Harvard Business School, right? Because you’ve got a compact with the school, but nothing’s been negotiated beyond that.
ZOE CULLEN: That’s right. And also the kind of expertise that I do, much of it comes in the form of something codified. So in the context of a peer-reviewed article. And previously those articles might have been challenging to apply to say a consulting setting. So like a setting where maybe, in my case, an HR expert wants to understand how to roll out pay transparency. But now for a variety of reasons, the tools might be able to take my articles and my way of thinking about the world and commercialize it in new ways and say, serve HR professionals through using my research on the topic.
BRIAN KENNY: Yes.
ZOE CULLEN: So in some ways that’s my broader market for my expertise. And you mentioned Harvard Business School. Yes, absolutely. I upload all the videos of my classes. Potentially Harvard Business School can recombine my videos with Clay Christensen, make something even better than what I could do by my own… And sell it in new varieties of ways that I couldn’t possibly have foreseen when I first signed my contract with HBS.
BRIAN KENNY: One of the things that you talk about in the case is the shadow workforce. And I think this is the idea of just average ordinary people being on the internet and generating data from their activity on the internet that could potentially be used in some sort of a commercial way without their knowledge or consent. Can you talk a little about that?
ZOE CULLEN: Yes. My collaborator, Danielle Lee, on the research that we’re doing together, the two of us actually have started to refer to this shadow workforce as the scab farms. Because of course in any context, in any workplace context, it might be possible to, once you see the link between your data stream and how it’s used to replicate what you do, you can play an active role in mediating how much that data contains, how it’s transferred. You could imagine that in the context of any firm, there’s some agency. However, if companies were to just be able to turn to the global talent market and hire per hour to data label something similar to what you’re doing, this means that there’s a huge pool of substitutes for either your employer or a third party to hire in order to replicate some of the things you’re doing.
BRIAN KENNY: Yeah.
ZOE CULLEN: And that of course means that your ability to control your asset, the data about your labor and to sell it for something reasonably lucrative is mediated by how many substitutes there are.
BRIAN KENNY: Yes. Okay. And I guess that extends into the workforce, too because we’re spending a lot of time trying to get our whole staff here upskilled on how to use Claude and ChatGPT and other things. They’re creating custom chats that might make their job easier by automating certain functions of their work. I would imagine that’s also susceptible to being commercialized in some way without their knowledge.
ZOE CULLEN: The nice part about teaching this particular context is that it really proves that collectively there are reasonable and feasible actions to take so that the people who are contributing to the AI innovation are rewarded for it.
BRIAN KENNY: Yes.
ZOE CULLEN: I know that it feels far from the professional context for many today, but in the 1930s, the idea of peer reviewed credit attribution for screenplays must have felt wildly fanciful.
BRIAN KENNY: Yes.
ZOE CULLEN: And then now it’s the basis for how the whole industry operates. And so I think this opens up the imagination for what kind of institutions make sense now that much of labor can be codified.
BRIAN KENNY: Yeah. So if you were advising a CEO, not in the entertainment industry, but just any other sector, what would you tell them that they need to be thinking about with regard to this?
ZOE CULLEN: So in the context of knowledge work in industry, I think CEOs have a huge opportunity in learning from this case because you can see how someone, a superstar like Tom Hanks, is proactively getting involved in AI footage. And the reason for that is because incentives are aligned. And I think many employers that I’ve worked with, they kind of just at the point where they’re trying to encourage employees to train their co-workers and to be really participatory in AI innovation, they’re finding progress is stalled. And one explanation and an explanation that could really lead to fruitful innovation is the contracts are not in place to the incentive alignments that have been created in Hollywood into say the context of a consulting firm is to say, “Okay, I understand that I’m asking from my workforce for them to transfer their core expertise into a physical asset that then I, the firm might own.” And that’s very exciting, lots of potential use cases. Now in order for an employee to be really excited by that and motivated by that, how can I give them a stake in that physical asset rather than leave them just with the fear that that physical asset then reduces headcount or reduces their leverage in the firm?
And so I’ve experimented with a number of different contracts that I think are really executable inside the workplace. One is just merely realizing that the co-worker is like an extension of a person so that their promotion and their career incentives are tied to how that co-worker can help others in the firm. And also finding a way for that worker to invest in the technology, knowing that they can take some part of it with them, develop their own IP as part of the process so that they want to do an excellent job in the context of one firm, knowing that even if they get fired in that role, they’re going to move to another firm and have something really valuable to offer.
BRIAN KENNY: That’s fascinating. And what you’re describing sounds like it could fundamentally change the nature of the relationship between employees and employers.
ZOE CULLEN: Ah, that’s the hope, right? It’s the hope that all your employees are now part of your R&D team.
BRIAN KENNY: Yeah.
ZOE CULLEN: And we think about the complicated IP relationship around that appropriately and everyone’s incentivized to really contribute to the innovative part of the company.
BRIAN KENNY: Yeah. Well, Zoe, my mind is kind of blown. This has been a really amazing conversation. And when I read the case, I didn’t have all this. I’ve gone off script on some of these questions because I was sort of hearing things differently as we spoke about it. So thank you for that. Let me ask you just one last question, which is if you want our listeners to remember one thing about the Metaphysics case, what would it be?
ZOE CULLEN: As an AI company or the user of an AI tool, you have access to the very top talent as long as you respect what property rights are necessary for the very top talent to be fully invested.
BRIAN KENNY: ZOE CULLEN, thank you for joining me on Cold Call.
ZOE CULLEN: Thank you very much, Brian. It’s been a pleasure.
BRIAN KENNY: If you enjoy Cold Call, you might like our other podcasts: Climate Rising, Coaching Real Leaders, IdeaCast, Managing the Future of Work, Skydeck, and Think Big, Buy Small. Find them wherever you get your podcasts.
If you have any suggestions or just want to say hello, we want to hear from you. Email us at coldcall@hbs.edu. Thanks again for joining us. I’m your host BRIAN KENNY, and you’ve been listening to Cold Call, an official podcast of Harvard Business School and part of the HBR Podcast Network.
