Pixar Co-Founder Edwin Catmull on AI and Storytelling

Edwin Catmull is a computer scientist -- and a force of creativity. He helped bring to life beloved, generation-defining movies including Toy Story, Finding Nemo, and Ratatouille. Motley Fool host Ricky Mulvey caught up with Catmull to discuss:

• Being in the "business of exponential change."
• Potential AI upheaval in the animation industry.
• How technology and story advance each other.

To catch full episodes of all The Motley Fool's free podcasts, check out our podcast center. To get started investing, check out our quick-start guide to investing in stocks. A full transcript follows the video.

This video was recorded on January 21, 2024.

Edwin Catmull: We wanted to do research, I don't mean research in the technical sense. Research where you go out into the world and find out something you don't know. The reason is you actually need stakes to get away from the stereotypes. If your expertise is something you learned from movies, then you tend to copy what's in movies, which makes everything derivative. For us, it was to say, well, we need to go out in the world and discover that which is not obvious.

Mary Long: I'm Mary Long, and that's Ed Catmull, co-founder of Pixar, and a pioneer in both computer animation and in storytelling. Ricky Mulvey caught up with Catmull to discuss AI's impact on the animation industry, stakes in storytelling, and why imagining oneself riding bareback on a herd of wild horses can help in dealing with unpredictable talents.

Ricky Mulvey: Ed, I want to start at a jumping-off place, which is Pixar characters are often admired for their curiosity. I know you have to be a curious person as you have developed software in stories. So right now, what are some things you're curious about?

Edwin Catmull: Well, there are several things that I'm curious about. There are actually too many things now, but one of them is just looking back at the implications of the rate of change. Over my entire history, it's been fairly clear that the underlying change in the speed and cost of computing was at an exponential rate. I think as a practical matter like 30% per year, other improvement. I think the actual numbers, it isn't forget transistor count. It's really the effectiveness of the chips and the cost is probably more like 20-25% annual increase over 50 years.

Now having said that, the companies along the way like the workstation companies and the other companies were in the business of building on this exponential change. Yet they didn't see the implications of it. This for me was mind blowing, what in the world was going on? This is separate from the cultural questions, how you solve the technical problems. It's just looking at the industry as a whole that I was in the midst of thinking, why is it that people weren't seeing this? Now I've got more time. It's to talk with some of the people involved, and to find that and actually a lot of the engineers didn't know this, but the leaders and managers of the company, in thinking about management and growth and the pressures on them, we're not listing to their own people, and their engineers.

One of the results is none of the workstation companies have survived until today. Now for me, this is curious. Now I'm going back and looking at it and working with a friend at MIT Sloan is to think, what is it that's going on? Why are people missing something which is so clear and instead reacting when something actually becomes so big, it hits them in the face where they could have prepared ahead of time.

Ricky Mulvey: In the cost of computer power declining in the rate of change in the amount of computing power that people get going up, what are the implications that you and the engineers were worried about that the leaders were not?

Edwin Catmull: Well, if you can actually clearly see that something is going to change by more than an order of magnitude or some major component of your business is going to change, it's going to have an impact. That impact, it means that you have to think out 5, 10, 15 years, you really have to put some attention on it. Most industries didn't even have anybody inside who could see that. If they did, they typically ignored them. It was an ongoing problem, I think in the case of what we're seeing now with an even faster exponential growth rate with machine learning, the annual growth rate of processing power and cost for the GPUs is even greater than it was for CPUs, and has been for several years. We're now seeing the consequences of that, and those consequences regarding machine learning, they're now very obvious, but we don't typically have people in companies that can even think about the implications of the change.

Ricky Mulvey: I'm sure you also have thoughts on the effects of AI on the animation business. I hope you will forgive me for reading a take from Jeffrey Katzenberg about this. This has made some news, he said, "At the Bloomberg New Economy Summit in the Good Old days, when I made an animated movie, it took 500 artists five years to make a world-class animated movie. I don't think it will take 10% of that. Literally, I don't think it will take 10% of that three years from now."

Edwin Catmull: I think in terms of let's say, making an animated film, the number of people who knew how to use the tools for doing 3D animation. It's very small to begin with. As Pixar started, we had to train people, and we wanted people who were trained as artists at observational skills. We didn't actually care whether or not they knew how to use the tools, because we trained them to need the tools, they want the schools or the skills that we needed. Now as we grew, we tried to be very careful, and this is starting back at Lucasfilm is to understand that this was not about technology, it was about telling stories. If we had a technological intuitive force, then we would have failed. But if we got the story right, we would succeed.

There is that issue of what is the optimal group for creating something which is thoughtful and able to make a good story, but use the technology in the process of doing it. Now that the number of people that we would take to make films is, I'm going to say roughly on the order of 300 people dedicated to it. Near the tail end, it's a small group to begin with. Then it's only when you figure out what it is that you add a lot of skilled people to do it. But the real question for us was, from a creative point of view, what's an optimal number of people? As the technology has gotten a lot less expensive in all different fronts, the tools also the number of people trained to use them is significantly more than it was before and they're very good. That's a different world that we're now in today.

That the barrier to entry, if you will, to make anything is actually very low. If you want to produce something which is B or C level, it doesn't take very much to do it. It's cheaper to do, typically a 3D film for young children on Saturday morning than it is by hand. The economics have changed and it continue to change. A Generative AI is clearly going to continue to change that. That's the course that we're on and you can't ignore that the process that's going on any more than the workstation company should have ignored what was happening to the CPUs underneath that, at their peril. But that doesn't ask the question that if you want a rich full film, what's the optimum number of people? I think that's the better question that you want to ask for a really good film. But the answer is, what would it take to produce a C level film, is the answer's going to be pretty cheap. Very cheap. A brilliant person or two, using generative AI, could do something which is great, which is what happens with arts and languages. A single person will do something. I think if you've got a very rich film, then we've got contributions from a fair number of people.

I'd always estimated and it's just a gut feeling that because you've got looks and appearances, but you're also trying to draw in ideas from a number of people. Then optimal numbers probably around 100. That when you get larger than that, it's very difficult to change course because it gets very complicated. If it gets too small in general, not always, then you don't have as much richness and depth that you'd like.

Ricky Mulvey: Please correct me. We may be a few years away from someone who can maybe draw a little bit and use natural language commands to make their own animated movie.

Edwin Catmull: Well, I don't think that's even far away. In order to be able to do something, the answer is yes. Probably a few years, but you can easily imagine that, but in order to do that, it's, you're describing something and with generative AI, it's creating the dialogue, and it's creating the imagery. You could draw a comparison. If you look back years ago at the Scooby Doo cartoons, which I watched, professional reasons, not because I enjoyed them, but I was more fascinated by the fact that there were only three plots, the four plots. There was the one that took place in the haunted mansion, there was the one that took place in the mountains with the yeti, there was one in the cave. I forget what the other one was. The settings were very repetitive with each other and the format was fairly easy.

So the cost of writing them was actually very simple and it was very low cost, but it was also only watchable by children. My wife will listen to series on her iPad, I'm in the other room, I'm just hearing the voice and listening to how are these people speaking. Well actually, all of it sounds not very good, not very realistic. I think when you actually watch it with the visuals, it's more compelling because you have something with the actors, but it all sounds like something which you really could generate because there's nothing particularly interesting that I find just in listening to them. Another question is, will that be generated? It can easily be generated, and the answer is yes. Then every once in a while, you'll see something which was really refreshing, a new and real, even though it's a fantasy in many senses, which is what R does and what R is, but it would not have been generated. We will be able to fill the air with stuff which is pre-derivative. How far away is that? I don't think it's terribly far away and it will happen.

Ricky Mulvey: I want to go to creativity in for a little bit because one part of it that I found useful for me is especially the section on mental models and how we use them whether or not we're cognizant of them and how especially creative people use mental models to stay centered. One example is a producer from Pixar, John Walker, who imagined balancing an upside down pyramid on his fingertip. You wrote that when you were at Lucasfilm, your model was riding bareback on a herd of wild horses. Why was that a useful model during your time at Lucasfilm?

Edwin Catmull: Well, for me, it was accepting that in order for us to move forward, we had to have these unpredictable talents around. There's something about just saying, I'm not really totally in charge of this group, nominally I'm the leader of it, but if I thought I was trying to control what the people were doing or where they were going, even I would actually slow things down. It was to think of in terms of like this is the wild ride. For me, a lot of times it's an acceptance of the nature of the problem. Just accepting the unpredictability of it, I've always felt to be a useful tool to have, as well as I found it was useful to know that I was frequently wrong.

Ricky Mulvey: Pixar was famous for essentially where story advanced the technology and technology would advance the story. There's a million examples. One would be figuring out subdivision in a bug's life to make essentially triangles very small. You have a smooth surfaces appeared on the insects. While we know the times that it worked, were there ever times where maybe you're in story trust meeting where you have to stop an idea because you're like, this is wonderful, but this is not technically possible at this moment. I think of as you referenced James Cameron, he had to delay Avatar for more than a decade because he was like the text just not there yet.

Edwin Catmull: Obviously for Toy Story, essentially these are characters that are made out of plastic. The humans didn't look realistic. We had the complexity where things that we'd worked out like the leaves and the trees with massive amounts of stuff, but essentially we're dealing with flat surfaces. The next film was Bug's Life, where we dealt with more characters, but even at that time, the sheer number of ants, the people working on us saying, we don't think we can do that many. It was like a question whether or not we could even do it, but the technology represented limits and so we were designing around what we knew was possible. It also had uneven terrain.

Now, it didn't take long before we reached the point where that really wasn't a major limitation. As the rate of improvement in the technology became then for the next few years of which are the things that are going to be the new problems that we should try and solve, if we focus on that. They ultimately reached the point where the technology became accessible enough. For me, in some it also became a problem because it was in excess. One of the difficulties with making a film is that working out the story is really hard. Going in the screening room and looking at all the fun visuals is just status, fun. It then becomes a distraction from the story. You've seen movies, as I've seen movies, where they look great and the stories suck. Why aren't they spending more time on the story and less time on the visuals? It's kind of getting seduced in the candy store because you don't want to do your homework. That's the range you went over from, that's our limitation to actually, we can do anything we want, and it gets in the way because we're spending time doing anything we want rather than what should be done.

Ricky Mulvey: I think one thing Pixar also really understands is setting stakes in storytelling, and I think that's what maybe a lot of the visual feasts that you've described missed. I see executives wanting optionality in a story and I worry that comes to pass with multiversal storytelling. I guess, what was, maybe not the magic, but how did you think about setting stakes in a story at Pixar?

Edwin Catmull: Well, the reason why I think that there's a better number, it's not like there is the actual number, but you want several people involved with it. But it isn't just you've got several people involved. For us, we wanted to do research. I don't mean research on a technical sense, the research where you go out into the world and find out something you don't know. The reason is you actually need stakes to get away from the stereotypes. If your expertise is something you've learned from movies, then you tend to copy what's in movies, which makes everything derivative. For us, it was to say, well, we need to go out in the world and discover that which is not obvious.

As an example, although there are several of them, but one of them is Ratatouille, released several years ago. The premise was a hard one, and it's important for us to take on hard premises. That is, these are ideas that would not pass the elevator test, because you can't describe quickly why this is a good idea. It's a challenge that you have to go out and not only solve this challenge, you have to find out things that the audience wouldn't know. In the case of Ratatouille, we basically have typically cooking in our home, so we know what it means to cook. Some of us cook. We see cooking in restaurants. We can watch cooking channel. But basically, almost none of us have been inside of the professional high-end kitchen. What is that really like? In this case, the research, and some research isn't this nice, but you go into high-end restaurants, you get to know them. In the case of the French Laundry, for instance, they went in and they worked in the kitchen. Thomas Keller had the people in there in the kitchen, where they went to France to the high-end restaurants, and they got to know, who are these people? What are they thinking? How are they trained? Then how do you quickly describe that in a film? The thing about this is the audience doesn't know whether or not it's true, but they sense that it is.

That's what you're trying to capture, is something that they don't know, but you sense is true. That's one element. The other one is, is you really want ideas and problems that people can relate to. It isn't just you can tell a story or even a fun story. There are good movies, they are just fun, or they're comedic, or they're adventuresome. I enjoy them also. But every once in a while you get a movie which touches you or addresses the kind of thing that you're worried about. That's what Pixar was trying to do, is to say of all the kinds of films that we do, what are the ones that can touch us? Animation, it's abstract, it's an exaggeration, and it's metaphorical by its nature. In the case of Up, well, the characters in your head look like real people. They have eyes and mouth, and they're like people, but they're not, but they're representative. But you are trying to capture something and explain something which is hard to explain, which is the emotions inside a person, and the fact that the emotions are sometimes in contradiction with each other.

In an abstract level, we know that's true, but it's not obvious to some people that if a child is distressed that they've got these conflicting emotions. In order to do that, the research is to talk with child psychiatrist, as well as to think about your own children and what happens there, and then to convey that in a story. That becomes the underlying theme for anything to capture something which is real and do it in a way that draws people in and gives them something new, that's not talking down to them, it's appealing to what they know, and what they don't know, and their desire to know things that are new to them.

Ricky Mulvey: I appreciate that. I want to go back to mental models real quick. You're now an advisor at an independent game company. That game company, they have a well-known game called Sky. Are there any mental models that you find applicable to you right now in this new project?

Edwin Catmull: Well, I've only met with them a few times. The truth is often my name is used to say that I'm advisor, but I don't meet with them that much, to be honest. I don't actually have an answer for that one. Honestly, more of my advising has to do with other technical companies. How do they think about the technology and the rate of change, and what it's being used for, and how they might think about long-term implications when we don't really know. The nature of a mental model, typically, if you think about the ones that people of God is how they're dealing with the unknown.

For me, that's the more profound question is, how do you both extrapolate from where you are? How do you make an estimate about what will take place in the future and accept the fact that you really don't know a lot? You're taking steps toward something that is unknown, which means you're going to completely or frequently alter what you think is where you're going. Now, the problem is this is a hard concept because a lot of leaders want to have more clarity about where they're going. In fact, a lot of people in the companies want the clarity. The problem is often the clarity is just wrong, in which case the company goes over the cliff. Dealing with that reality of unknown change in the world in which we're judged by immediate returns, a pretty difficult position, and most people have difficulty navigating.

Ricky Mulvey: Well, Ed, I want to be mindful of your time. It's been an absolute honor to be able to have a chance to talk to you. As I'm sure many people who speak with you, Pixar was a very important part of my childhood and my understanding of how to tell stories. Thank you for your wisdom and your time spent with us listeners on Motley Fool Money.

Edwin Catmull: It's my pleasure talking to you.

Mary Long: As always, people in the program may have interest in the stocks they talk about, and The Motley Fool may have formal recommendations for or against, so don't buy or sell stocks based solely on what you hear. I'm Mary Long. Thanks for listening. We'll see you tomorrow.