Gaming and Interactive Agents
Gaming and Interactive Agents
Revolutionizing human interactions in games and virtual worlds
with cutting-edge AI agents that unlock new possibilities
and deeper connections for game developers and players.
Our Approach
We are transforming Game-AI beyond rule-based systems by using deep reinforcement learning to train robust and challenging AI agents in gaming ecosystems. This technology enables game developers to design and deliver richer experiences for players. As AI technology continues to evolve and mature, we believe it will help spark the imagination and creativity of game designers and players alike.
Our Work
Gran Turismo Sophy™
Gran Turismo Sophy is the result of a unique collaboration between Sony AI, Polyphony Digital, the creative studio behind the world-famous Gran Turismo games, and Sony Interactive Entertainment.
GT Sophy Tech Blog Series
In early 2020, Sony AI set out to do something that had never been done before: create an AI agent that could beat the best drivers in the world at the PlayStation game Gran Turismo. In 2021, we succeeded with Gran Turismo Sophy. In this blog series, we will try to illuminate and make more accessible some of the technical details of the work published as a cover article in the journal Nature in February 2022.
The Team Behind GT Sophy Docuseries
The docuseries, “Meet the Team Behind GT Sophy,” offers a behind-the-scenes look at the development of GT Sophy and the team who created it. The series highlights how Sony AI’s globally-distributed team of world-class researchers and engineers virtually collaborated during the COVID-19 pandemic to develop and train the AI agent.
RESEARCH AREAS
Training Reinforcement Learning Agents for Video Games
Training Reinforcement Learning Agents for Video Games Reinforcement learning (RL) agents learn how to perform tasks through repeated practice in an environment. Modern video games require precise control and creative solutions, providing a challenge for these (often brittle) RL agents. The rich landscape of tasks in modern video games inspires innovations that are potentially applicable to a wide variety of other domains as well.
Interacting Seamlessly with Humans and Other Agents
Almost all video games involve dealing with other agents or humans and the rules of engagement may not be formalized. Without a well-specified cost function or access to all possible behaviors, agents in video games need to be more robust and tunable than in traditional AI domains. Since almost all real-world domains involve multiple agents, the multi-agent learning techniques we study are also quite broadly applicable.
Constructing an Engineering Ecosystem for Game-AI Training
Modern RL and AI algorithms need large amounts of compute and data resources. Sony AI has developed a platform and engineering ecosystem that allows for the rapid deployment of learning algorithms that train on video games at scale. Work continues to add capabilities to the platform and make RL a trustworthy and repeatable process for video games and other production domains.
Awards & Recognitions
Outracing Champion Gran Turismo Drivers with Deep Reinforcement Learning
Our research, Outracing Champion Gran Turismo Drivers with Deep Reinforcement Learning was recognized as new breakthrough in AI was featured on the cover of Nature.
2022 ACM SIGAI Industry Award
Sony AI Wins 2022 ACM SIGAI Industry Award for Excellence in Artificial Intelligence for its recent AI breakthrough, Gran Turismo Sophy™
Our Team
Related Publications
Hierarchical reinforcement learning (RL) can accelerate long-horizon decision-making by temporally abstracting a policy into multiple levels. Promising results in sparse reward environments have been seen with skills , i.e. sequences of primitive actions. Typically, a skill …
Two desiderata of reinforcement learning (RL) algorithms are the ability to learn from relatively little experience and the ability to learn policies that generalize to a range of problem specifications. In factored state spaces, one approach towards achieving both goals is …
Robustly cooperating with unseen agents and human partners presents significant challenges due to the diverse cooperative conventions these partners may adopt. Existing Ad Hoc Teamwork (AHT) methods address this challenge by training an agent with a population of diverse tea…
Related Blog Posts
From Hypothesis to Reality: The GT Sophy Team Explains the Evolution of the Brea…
Since its inception in 2020, Sony AI has been committed to enhancing human imagination and creativity through the acceleration of AI research and development. One of the first exam…
Event Tables for Efficient Experience Replay
Each of us carries a core set of experiences, events that stand out as particularly important and have shaped our lives more than an average day. However, this is often not the cas…
RPOSST: Testing an AI Agent for Deployment in the Real World
Bleary-eyed engineers know the anxiety that comes with a deployment, and the importance of testing every aspect of a product before it goes to the “real world.” Will the response t…
JOIN US
Shape the Future of AI with Sony AI
We want to hear from those of you who have a strong desire
to shape the future of AI.