Autonomous driving. Virtual reality. Chatbots. Online collaboration. Livestreaming. GPU deep learning. Facial recognition. It often comes as a surprise to learn that so many of the revolutionary ideas that dominate today’s tech headlines actually have their roots in the gaming industry. But one revolutionary technology stands out above the rest for its long history in games: artificial intelligence.
“AI in games has been around since gaming has been around,” says Youichiro Miyake, Lead AI Researcher at Japanese gaming giant Square Enix and a highly respected figure in the Japanese gaming industry. “In the 80s and early 90s it came in the form of simple scripts that reacted to player movements, the scope of which was very limited. But now that most games are 3D, the technology involved has become incredibly complex.”
“I wanted to create artificial life itself – something with a body and a mind, something that takes actions in real time, reacting to its environment.”
Miyake, who began his gaming career in 2004 after completing his Ph.D. at Tokyo University, builds AI for Square Enix’s critically acclaimed Final Fantasy series. At last month’s Rakuten Technology Conference, he provided some rare insights into how he works in creating revolutionary AI technology.
Developing autonomous AIs
It was Miyake’s passion for artificial intelligence that drew him to the gaming industry. “I wanted to make ‘whole’ AIs, rather than simply develop algorithms,” he explains. “I wanted to create artificial life itself – something with a body and a mind, something that takes actions in real time, reacting to its environment. I decided that games would be the best way to do that.”
Miyake separates his work on Final Fantasy into three different categories: character AI, which functions as each individual character’s “brain”; navigation AI, which extracts information about the game environment; and meta AI, which observes the characters, environment and players all together over time, making judgments and regulating the game itself.
Developing robust character AI is of core importance, as it’s the characters that interact directly with players. Miyake likes to compare his characters to real-life autonomous vehicles. “The two are very similar,” he says. “AI characters must recognize their environment, recognize how that environment is constantly changing, and make decisions accordingly. All of this needs to be done in real time, as it does with self-driving cars.”
Miyake has serious ambitions about the future of his character AI development. “Eventually I would like to make characters self-conscious, even capable of understanding human players, just by observing their in-game behavior,” he states. “I think this would revolutionize the entire gaming industry.”
When it comes to applying in-game AI technology outside of the gaming world, Miyake says that of the three types of AI, it is meta AI that shows the most promise. “The role of the meta AI is to observe. It can observe humans, it can observe their minds by analyzing their behavior,” he explains. “This could be applied to so many different situations. Hotel lobbies, for example: the meta AI would observe guests and their movements, issuing orders to robots to help guests with their luggage or lead them to their rooms.”
The meta AI is necessary to control all of the moving parts in a complex system. “While it would be difficult for the individual robots to understand what the guests are thinking, by observing everything at once, the meta AI could handle this much more efficiently.”
The philosophy of intelligence
For Miyake, his work is closely related to philosophy, particularly the work of thinkers such as Descartes (“I think, therefore I am”) and Husserl (phenomenology). “To make artificial life, we need to answer some serious philosophical questions,” he says. “What is intelligence? What is life? What does it mean to live?”
“Our work has a deep relationship with philosophy,” he continues. “We are not just making characters, but rather intelligence itself. And intelligence is life.”
Read more posts from the Rakuten Technology Conference here.