All anyone wants to talk about in the games industry is AI. The technology — once a twinkle in the eye of sci-fi writers and futurists — has shot off like a bottle rocket. Every day we're greeted with fascinating and perturbing new advances in machine learning. Right now, you can converse with your computer on ChatGPT, sock-puppet a celebrity's voice with ElevenLabs, and generate a slate of concept art with MidJourney.
It is perhaps only a matter of time before AI starts making significant headway in the business of game development, so to kick off AI week at IGN, we talked to a range of experts in the field about their hopes and fears for this brave new world, and some are more skeptical than you'd expect.
Pawel Sasko, CD Projekt Red Lead Quest Designer: I really believe that AI, and AI tools, are going to be just the same as when Photoshop was invented. You can see it throughout the history of animation. From drawing by hand to drawing on a computer, people had to adapt and use the tools, and I think AI is going to be exactly that. It's just going to be another tool that we'll use for productivity and game development.
Tim Sweeney, Epic Games CEO: I think there's a long sorting out process to figure out how all that works and it's going to be complicated. These AI technologies are incredibly effective when applied to some really bulk forms of data where you can download billions of samples from existing project and train on them, but that works for text and it works for graphics and maybe it will work for 3D objects as well, but it's not going to work for higher level constructs like games or the whole of the video game. There's just no training function that people know that can drive a game like that. I think we're going to see some really incredible advances and actual progress mixed in with the hype cycle where a lot of crazy stuff is promised. Nobody's going to be able to deliver.
Michael Spranger, COO of Sony AI: I think AI is going to revolutionize the largeness of gaming worlds; how real they feel, and how you interact with them. But I also think it's going to have a huge impact on production cycles. Especially in this era of live-services. We'll produce a lot more content than we did in the past.
Julian Togelius, Associate Professor of Computer Science at New York University, and co-author of the textbook Artificial Intelligence and Games: Long-term, we're going to see every part of game development co-created AI Designers will collaborate with AI on everything from prototyping, to concept art, to mechanics, balancing, and so on. Further on, we might see games that are actually designed to use AI during its runtime.
I think it's a pretty intriguing direction because it opens up the doors that you wouldn't think of.
Pawel Sasko: There's actually many companies doing internal R&D of a specific implementation of not MidJourney especially, but literally just art tools like this, so that when you're in early concept phases, you're able to generate as many ideas as you can and just basically pick whatever works actually for you and then give it to an artist who actually developed that direction. I think it's a pretty intriguing direction because it opens up the doors that you wouldn't think of. And again, as an artist, we are just always limited by our skills that come up from all the life experiences and everything we have consumed artistically, culturally before. And AI doesn't have this limitation in a way. We can feed it so many different things, therefore it can actually propose so many different things that we wouldn't think of. So I think as a starting point or maybe just as a brainstorming tool, this could be interesting.
Michael Spranger: I think of AI as a creativity unlocking tool. There are so many more things you can do if you have the right tools. We see a rapid deployment of impact of this technology in content creation possibilities from 3D, to sound, to musical experiences, to what you're interacting with in a world. All of that is going to get much better.
Julian Togelius: Everybody looks at the image generation and text generation and say, 'Hey, we can just pop that into games.' And, of course, we see like proliferation of unserious, sometimes venture capital found that actors coming in and claiming that they're going to do all of your Game Arts with MidJourney — these people usually don't know anything about game development. There's a lot of that going around. So I like to say that generating just an image is kind of the easy part. Every other part of game content, including the art, has so many functional aspects. Your character model must work with animations, your level must be completable. That's the had part.
Tim Sweeney: It's not synthesizing amazing new stuff, it's really just rewriting data that already exists. So, either you ask it to write a sorting algorithm in Python and it does that, but it's really just copying the structure of somebody else's code that it trained on. You tell it to solve a problem that nobody's solved before or the data it hasn't seen before and it doesn't have the slightest idea what to do about it. We have nothing like artificial general intelligence. The generated art characters have six or seven fingers, they just don't know that people have five fingers. They don't know what fingers are and they don't know how to count. They don't really know anything other than how to reassemble pixels in a statistically common way. And so, I think we're a very long way away from that, providing the kind of utility a real artist provides.
Sarah Bond, Xbox Head of Global Gaming Partnership and Development: We're in the early days of it. Obviously we're in the midst of huge breakthroughs. But you can see how it's going to greatly enhance discoverability that is actually customized to what you really care about. You can actually have things served up to you that are very, very AI driven. "Oh my gosh, I loved Tunic. What should I do next?
Tim Sweeney: I'm not sure yet. It's funny, we're pushing the state of the art in a bunch of different areas, but [Epic] is really not touching generative AI. We're amazed at what our own artists are doing in their hobby projects, but all these AI tools, data use is under the shadow, which makes the tools unusable by companies with lawyers essentially because we don't know what authorship claims might exist on the data.
Julian Togelius: I don't think it will affect anyone more than any other technology that forces people to learn new tools. You have to keep learning new tools or otherwise you'll become irrelevant. People will become more productive, and generate faster iterations. Someone will say, "Hey, this is a really interesting creature you've created, now give me 10,000 of those that differ slightly." People will master the tools. I don't think they will put anyone out of a job as long as you keep rolling with the punches.
Pawel Sasko: I think that the legal sphere is going to catch up with AI generation eventually, with what to do in these situations to regulate them. I know a lot of voice actors are worried about the technology, because the voice is also a distinct element of a given actor, not only the appearance and the way of acting. Legal is always behind us.
Michael Spranger: The relationship with creative people is really important to us. I don't think that relationship will change. When I go watch a Stanley Kubrick movie, I'm there to enjoy his creative vision. For us, it's important to make sure that those people can preserve and execute those creative visions, and that AI technology is a tool that can help make that happen.
Julian Togelius: Definitely. If you have a team that has deep expertise in every field, you're at an advantage. But I think we're gonna get to the point where, like, you only need to know a few fields to make a game, and have the AI tools be non-human standings for other fields of expertise. If you're a two-person team and you don't have an animator, you can ask the AI to do the animation for you. The studio can make a nice looking game even though they don't have all the resources. That's something I'm super optimistic about.
Tim Sweeney: I think the more common case, which we're seeing really widely used in the game industry is an artist does a lot of work to build an awesome asset, but then the procedural systems and the animation tools and the data scanning systems just blow it up to an incredible level.
Michael Spranger: Computer science in general has a very democratizing effect. That is the history of the field. I think these tools might inspire more people to express their creativity. This is really about empowering people. We're going to create much more content that's unlocked with AI, and I think it will have a role to play in both larger and smaller studios.
Michael Spranger: I think what makes this different is that the proof is in the pudding. Look at what Kazunori Yamuchi said about GT Sophy, [the AI-powered driver recently introduced to Gran Turismo 7]: there was a 25-year period where they built the AI in Gran Turismo in a specific way, and Yamuchi is basically saying that this is a new chapter. That makes a difference for me. When people are saying, "I haven't had this experience before with a game. This is qualitatively different." It's here now, you can experience it now.
Kajetan Kasprowicz, CD Projekt Red Cinematic Designer: Someone at GDC once gave a talk that basically said, "Who will want to play games that were made by AI?" People will want experiences created by human beings. The technology is advancing very fast and we kind of don't know what to do with it. But I think there will be a consensus on what we want to do as societies.
Julian Togelius: AI has actual use-cases, and it works, whereas all of the crypto shit was ridiculous grifting by shameless people. I hate that people associate AI with that trend. On the other hand you have something like VR, which is interesting technology that may, or may not, be ready for the mass market someday. Compare that to AI, which has hundreds of use-cases in games and game development.
Luke Winkie is a freelance writer at IGN.
Continue reading...
It is perhaps only a matter of time before AI starts making significant headway in the business of game development, so to kick off AI week at IGN, we talked to a range of experts in the field about their hopes and fears for this brave new world, and some are more skeptical than you'd expect.
What do you think the biggest impact AI technology will have on the video game industry?
Pawel Sasko, CD Projekt Red Lead Quest Designer: I really believe that AI, and AI tools, are going to be just the same as when Photoshop was invented. You can see it throughout the history of animation. From drawing by hand to drawing on a computer, people had to adapt and use the tools, and I think AI is going to be exactly that. It's just going to be another tool that we'll use for productivity and game development.
Tim Sweeney, Epic Games CEO: I think there's a long sorting out process to figure out how all that works and it's going to be complicated. These AI technologies are incredibly effective when applied to some really bulk forms of data where you can download billions of samples from existing project and train on them, but that works for text and it works for graphics and maybe it will work for 3D objects as well, but it's not going to work for higher level constructs like games or the whole of the video game. There's just no training function that people know that can drive a game like that. I think we're going to see some really incredible advances and actual progress mixed in with the hype cycle where a lot of crazy stuff is promised. Nobody's going to be able to deliver.
Michael Spranger, COO of Sony AI: I think AI is going to revolutionize the largeness of gaming worlds; how real they feel, and how you interact with them. But I also think it's going to have a huge impact on production cycles. Especially in this era of live-services. We'll produce a lot more content than we did in the past.
Julian Togelius, Associate Professor of Computer Science at New York University, and co-author of the textbook Artificial Intelligence and Games: Long-term, we're going to see every part of game development co-created AI Designers will collaborate with AI on everything from prototyping, to concept art, to mechanics, balancing, and so on. Further on, we might see games that are actually designed to use AI during its runtime.
I think it's a pretty intriguing direction because it opens up the doors that you wouldn't think of.
Most people tend to think of AI through the prism of visual tools like MidJourney and Dall-E, but how do you expect the technology to impact game development in other, more under-the-hood ways. Like, say, sound, or animation, or optimization?
Pawel Sasko: There's actually many companies doing internal R&D of a specific implementation of not MidJourney especially, but literally just art tools like this, so that when you're in early concept phases, you're able to generate as many ideas as you can and just basically pick whatever works actually for you and then give it to an artist who actually developed that direction. I think it's a pretty intriguing direction because it opens up the doors that you wouldn't think of. And again, as an artist, we are just always limited by our skills that come up from all the life experiences and everything we have consumed artistically, culturally before. And AI doesn't have this limitation in a way. We can feed it so many different things, therefore it can actually propose so many different things that we wouldn't think of. So I think as a starting point or maybe just as a brainstorming tool, this could be interesting.
Michael Spranger: I think of AI as a creativity unlocking tool. There are so many more things you can do if you have the right tools. We see a rapid deployment of impact of this technology in content creation possibilities from 3D, to sound, to musical experiences, to what you're interacting with in a world. All of that is going to get much better.
Julian Togelius: Everybody looks at the image generation and text generation and say, 'Hey, we can just pop that into games.' And, of course, we see like proliferation of unserious, sometimes venture capital found that actors coming in and claiming that they're going to do all of your Game Arts with MidJourney — these people usually don't know anything about game development. There's a lot of that going around. So I like to say that generating just an image is kind of the easy part. Every other part of game content, including the art, has so many functional aspects. Your character model must work with animations, your level must be completable. That's the had part.
Tim Sweeney: It's not synthesizing amazing new stuff, it's really just rewriting data that already exists. So, either you ask it to write a sorting algorithm in Python and it does that, but it's really just copying the structure of somebody else's code that it trained on. You tell it to solve a problem that nobody's solved before or the data it hasn't seen before and it doesn't have the slightest idea what to do about it. We have nothing like artificial general intelligence. The generated art characters have six or seven fingers, they just don't know that people have five fingers. They don't know what fingers are and they don't know how to count. They don't really know anything other than how to reassemble pixels in a statistically common way. And so, I think we're a very long way away from that, providing the kind of utility a real artist provides.
Sarah Bond, Xbox Head of Global Gaming Partnership and Development: We're in the early days of it. Obviously we're in the midst of huge breakthroughs. But you can see how it's going to greatly enhance discoverability that is actually customized to what you really care about. You can actually have things served up to you that are very, very AI driven. "Oh my gosh, I loved Tunic. What should I do next?
A number of people have raised questions about how AI usage interacts with copyright law and the labor market for artists. What are your thoughts on those concerns?
Tim Sweeney: I'm not sure yet. It's funny, we're pushing the state of the art in a bunch of different areas, but [Epic] is really not touching generative AI. We're amazed at what our own artists are doing in their hobby projects, but all these AI tools, data use is under the shadow, which makes the tools unusable by companies with lawyers essentially because we don't know what authorship claims might exist on the data.
Julian Togelius: I don't think it will affect anyone more than any other technology that forces people to learn new tools. You have to keep learning new tools or otherwise you'll become irrelevant. People will become more productive, and generate faster iterations. Someone will say, "Hey, this is a really interesting creature you've created, now give me 10,000 of those that differ slightly." People will master the tools. I don't think they will put anyone out of a job as long as you keep rolling with the punches.
Pawel Sasko: I think that the legal sphere is going to catch up with AI generation eventually, with what to do in these situations to regulate them. I know a lot of voice actors are worried about the technology, because the voice is also a distinct element of a given actor, not only the appearance and the way of acting. Legal is always behind us.
Michael Spranger: The relationship with creative people is really important to us. I don't think that relationship will change. When I go watch a Stanley Kubrick movie, I'm there to enjoy his creative vision. For us, it's important to make sure that those people can preserve and execute those creative visions, and that AI technology is a tool that can help make that happen.
Do you think AI will democratize game development for indie studios that don't have the same resources as triple-A publishers?
Julian Togelius: Definitely. If you have a team that has deep expertise in every field, you're at an advantage. But I think we're gonna get to the point where, like, you only need to know a few fields to make a game, and have the AI tools be non-human standings for other fields of expertise. If you're a two-person team and you don't have an animator, you can ask the AI to do the animation for you. The studio can make a nice looking game even though they don't have all the resources. That's something I'm super optimistic about.
Tim Sweeney: I think the more common case, which we're seeing really widely used in the game industry is an artist does a lot of work to build an awesome asset, but then the procedural systems and the animation tools and the data scanning systems just blow it up to an incredible level.
Michael Spranger: Computer science in general has a very democratizing effect. That is the history of the field. I think these tools might inspire more people to express their creativity. This is really about empowering people. We're going to create much more content that's unlocked with AI, and I think it will have a role to play in both larger and smaller studios.
Gamers tend to be skeptical about technology hype cycles. What do you think makes the AI boom different from, say, NFT integration, or the metaverse, or 3D televisions, or other fads that have passed through the industry in the past?
Michael Spranger: I think what makes this different is that the proof is in the pudding. Look at what Kazunori Yamuchi said about GT Sophy, [the AI-powered driver recently introduced to Gran Turismo 7]: there was a 25-year period where they built the AI in Gran Turismo in a specific way, and Yamuchi is basically saying that this is a new chapter. That makes a difference for me. When people are saying, "I haven't had this experience before with a game. This is qualitatively different." It's here now, you can experience it now.
Kajetan Kasprowicz, CD Projekt Red Cinematic Designer: Someone at GDC once gave a talk that basically said, "Who will want to play games that were made by AI?" People will want experiences created by human beings. The technology is advancing very fast and we kind of don't know what to do with it. But I think there will be a consensus on what we want to do as societies.
Julian Togelius: AI has actual use-cases, and it works, whereas all of the crypto shit was ridiculous grifting by shameless people. I hate that people associate AI with that trend. On the other hand you have something like VR, which is interesting technology that may, or may not, be ready for the mass market someday. Compare that to AI, which has hundreds of use-cases in games and game development.
Luke Winkie is a freelance writer at IGN.
Continue reading...
