Artificial intelligence has been on a roll in recent months. The development of programs such as ChatGPT, DALL-E, and Stable Diffusion that can generate text and graphics with human creativity is already inspiring many innovators looking for new uses for technology. The next testbed for these advanced algorithms could be the virtual worlds of video games. Microsoft recently unveiled an internal demonstration of an AI program that allows the user to create things within the sandbox game Minecraft by giving him a single command, rather than performing separate tasks to build as in current gameplay. showed off. Sony is now allowing players of its racing game Gran Turismo to compete against its own AI program, GT Sophy. And ‘Roblox,’ a game based on players building and sharing new virtual worlds, plans to use this technology to help creators bring their visions to life. According to Dr. Frank Lee, Director of Entrepreneurship Division at Drexel University, video games have been using less sophisticated versions of his AI algorithms while Ghost is chasing Pac-Man, but advances in technology and processing power creates exciting new opportunities for game designers. A game studio in the ExCITe center. Lee recently shared some insight into the role AI has played in gaming and how it could shape the future of gameplay.
How is AI already built into games?
The use of AI in games was fairly rudimentary. An example is the A* algorithm. This is a fast algorithm for finding the optimal path of her computer-controlled NPC (Non-Player Character) in a virtual game world. The other is an alpha-beta pruning algorithm that allows NPCs to make decisions based on a set of options. Both of these algorithms have been around since the 1960s, and game AI will make use of both algorithms in some way. But even as algorithms evolved into machine learning, neural networks, and what we call artificial intelligence today (programs like ChatGPT and DALL-E), there was little effort to integrate them into games.
One of the more modern examples of what we now think of as artificial intelligence, or machine learning, in games was in the 1990s when IBM developed Deep Blue, a machine learning algorithm for playing chess. Human Grandmaster.
However, this was specific hardware designed at the chip level to quickly perform alpha-beta pruning searches on limited data sets (chess moves in this example). But we haven’t seen such sophisticated use of AI in years. In the early 2000s, the “Halo” franchise consolidated its behavior tree algorithms to make the behavior of non-playable characters less predictable, but they were still based on very limited conditions.
So there are bits and pieces of the game that are considered AI, but are really more like rules coded into the game to make it look more intelligent or more responsive.
How would using this type of AI in a game change it?
We are currently focusing on deep learning. Thus, instead of programs sifting through hundreds or thousands of pieces of information, these “neural network” programs can process millions or billions of data and generate responses. For example, ChatGPT’s neural network algorithm can be trained by taking any document available on the Internet, and when a user asks a question, it uses the question as a prompt to search this large corpus of training material. and a response similar to the trained example.
If these AI programs were built into the game, they could potentially steer the game’s narrative in new directions in real time, much like having your own personal DM. However, this can leave the game designer with little or no control over the game and the player’s playing experience. Therefore, similar situations can arise that produce unsettling answers, as reported on ChatGPT.
We’ve already seen this used on the design side to make characters more realistic and save some of the labor intensive work of designing environments. But for the most part, integrating it into gameplay is still at the theoretical stage.
What does AI mean for sandbox games? Microsoft’s Minecraft demo is a shortcut to building things, but doesn’t it ruin the game?
The question is, is that good game design? For example, a role-playing game may include elaborate crafting systems that allow you to create all sorts of things. And that’s a big part of that game’s gameplay. If the player doesn’t want to go through all the steps until the stage they aren’t playing, the game his designers just include the button as a shortcut. Part of the fun is knowing and learning about these elaborate recipes and how-to sets.
However, we found that a more accurate AI language model was integrated to help from an accessibility perspective. Being able to take voice input and translate it into appropriate actions has the potential to improve gameplay for players with various disabilities.
What are the challenges of introducing AI into gameplay?
The reason I think the AI is not so integrated is that gaming is a very fast-paced industry. As consumers, we expect a new game or a new update of the game to occur roughly once every year or two he. So even though a developer is given six months or he a year to make a game, graphics have been, and still are, the most metric by which people judge games. That’s true for most of gaming history.
As a result, most of the development and advancement of games has been in graphics engines and hardware acceleration of graphics, and the focus has been on making things look pretty, so much time is spent building AI into gameplay. and very little effort.
Large language models like ChatGPT enable the creation of conversational, non-playable characters that feel like you’re talking to a real human. The challenge, however, is that the developer has no control over its content. From a game designer’s perspective, the goal is to create a compelling narrative, but it’s hard to design and drive a narrative when you’re giving control to the AI and you have no idea what it’s saying. is.
what’s next? How do you foresee the latest revolution in AI impacting future games – from design to creation to play?
I believe that the potential of AI lies on the development side. AI assistance could significantly reduce the time it takes to generate more complex and detailed scenes. We are already seeing examples of programs being used to relieve the artist of the heavy lifting of designing environments and creating more human-like character movements and behaviors.
“No Man’s Sky” generated a lot of buzz due to its use of generative AI. In the game, the complete world was generated spontaneously as part of gameplay, again through a set of procedural algorithmic rules rather than a neural network AI program.
But this shows the possibility of using AI to generate new levels specific to the player. There are now games that generate new levels, but they are crowdsourced from other players who create them. But in theory, AI could be used to generate an infinite number of new levels. We could also see him using things like DALL-E during the idea generation process to quickly mock up ideas in the early stages of conceptualizing the game. By simply entering a few ideas during the brainstorming stage of your design, you can spark your creativity to come up with another idea or see if it works right away.
Media interested in speaking with Lee should contact Britt Falstick, Executive Director, News and Media Relations at 215-895-2617. bef29@drexel.edu
