One of the most controversial and often cited criticisms of modern AI systems is that they are built on and dependent on the stolen works of others. This isn't the only criticism people have of this new technology, and you may be wondering why anyone would want it in the first place. Today I'd like to talk about some of the bright futures that generative AI could have for us, and why the path to getting there may be difficult to walk.
The other day someone asked me what I liked most about my job. Ten years later, my answer is still basically the same. That's a strange thing to be surprised by. That also applies to generative AI. One of the best ways to use generative AI in games is to rely on the weirdness it can provide, rather than trying to make it as human-like as possible. A good example of this is iNNk, a game developed by some of my colleagues at ITU Copenhagen. iNNk is a drawing game played in teams against AI. One person tries to draw a word and her friend has to guess the word before her AI guesses. In playtests, we've seen all kinds of things, from people doodling over images to hide what they're doing, to using visual jokes and wordplay to fool his AI's more literal brain. It was shown that he was developing an outlandish strategy. This is a great example of what AI can do. new Rather than encroach on the human presence within the games we're already making, it provides a type of gaming experience.
AI can also provide new routes for accessibility. One of my favorite papers on the use of AI in games, his “Your Buddy, The Grandmaster'' describes how so-called “superhuman'' AI systems go beyond simply supporting players, and with all levels of ability and It describes how you can use it to suit your access needs. Trying to beat the world champion. This paper describes several different games based on the hardcore platformer Celeste, where superhuman AI enables new ways to play the game. One is that the player's control over the game is reduced to pressing his one button (the jump button), and the AI controls all other aspects of the game. The AI has a high skill level and can adapt to your jumps, but it doesn't play the game alone, so both parties need to work together to complete the game. In another version, the player can ask her AI to show the ghost how to complete the level in any way the player believes possible.None of the new games are Better From Celeste – Celeste is beautiful in its own right. But they're all unique and interesting variations that can be more accessible and appealing to people with different skill levels, accessibility needs, or interests.
These kinds of applications—live AI systems that reconfigure games or play with them—are not something you often hear about when generative AI is talked about. Often the dreams we are sold into are much simpler. His recent demo for Nvidia at CES, which allowed you to converse with game characters by speaking into a microphone, is being touted by the tech giant as “the future of gaming.” Many generative AI products try to sell us a vision of a completely open and free future, but this is reflected in many of the dreams children have about games where they can go anywhere and do anything. Masu.
I think the reason we hear this kind of pitch so often is because of where generation technology has ended up. In the early 2010s, generation systems were more like procedural content generation techniques like those found in games like Minecraft and Spelunky. Although they were treated as technical processes, these techniques have become a way for designers to express themselves through algorithms. Spelunky's level generator is an extension of Derek Yu's own sense of level design in his game designer. Minecraft's vast landscapes have been painted with brushstrokes that have been refined and trimmed over the years with updates from developer Mojang. Traditional procedural generation techniques are an extension of the person who created the system.
Manage cookie settings
But generative AI is having a hard time figuring out its guide. Much of that guidance comes from that dataset, but as we saw in the previous part of this series, that data often comes from a confusing array of sources, many of which are legally unstable. Nvidia's NPC scripts don't all come from the same dataset from your favorite authors. It's a collection of all kinds of things. Perhaps this is why many of these generated AI pitches swing in the opposite direction and present the player as the guide. When there is no human author and no consistent voice in the generative system, the selling point has to come from the only human left in the process: you.
But is that really what we want? 2023's big hitters, Baldur's Gate 3, Final Fantasy 16 (and Sometimes both games come together), I think losing this would be dissatisfying for many people. If I thought I was cool enough to write and play the role of the main character in a video game, I would have already put on stage my one-man show version of Mass Kanojo no Effect.
However, I don't think this is insurmountable. I believe that generative AI systems can overcome this problem if they are re-centered around the creator. It's not just about having people donate data and building models from it. It's about giving these creators the ability to control, edit, shape, and re-create models to achieve their own goals, and do so with the help of millions of ignorant and uncredited collaborators. The idea is to make it possible without using data. If we can find a way to put generative AI firmly back in the hands of creative people, we can begin to discover what artistic and empowering uses this technology can serve.
Some of the touted benefits of AI hide other issues, which makes it difficult to talk about the benefits. For example, AI can have a significant impact on localization. AI translation has come a long way in the last decade, unlocking hundreds of thousands of games that would never have been localized any other way, including abandoned games, games only available through emulation, and hundreds of thousands of games created through enthusiast communities. The possibility of translating games is increasing. In the world. These are all games that are very unlikely to require paid translation work, so they could be powerful applications for this technology.
However, even here there is a sense of tension. AI translations often lack cultural understanding, poetry, and human-written works. Translation work, even unpaid fan translations, is an important way for people to gain experience, build a community of translators, and keep a language alive and changing. Needless to say, AI translation is much better in some languages than others, and particularly threatened languages are often poorly supported. More emphasis on AI translation could bring many new games to many new audiences and help tell the history of the medium to a wider audience, but it could also create unexpected surprises for other parts of the community. It may also have a negative impact.
One of the difficulties when thinking about the benefits of AI is that it is difficult to distribute them equally among people. If we can create tiny robots that help make games better, that could be a huge benefit to the thousands of independent game developers around the world struggling to compete with the likes of Ubisoft and Activision. not. But nothing will stop studios like Ubisoft and Activision from building thousands of these little robots, dumping them into their companies, and making the difference again. Nor does he help the tens of thousands of ambitious independent developers who can't make enough money to buy one robot in the first place.
The benefits of AI tend to be a double-edged sword. But that doesn't mean you can't identify optimistic and positive stories at the same time. Because thinking about these consequences can help guide us toward technologies and a future that are more satisfying and perhaps free of these drawbacks. . In the next and final installment of this series, we'll flesh out some of these ideas and ask how, for better or worse, AI is actually used by game developers today.