Despite its success, SIMA isn't perfect, with errors occurring around more fine-grained understanding. DeepMind hopes to improve SIMA's performance, including its ability to follow more detailed instructions, and to develop AI systems that can act in as many environments as possible, achieve a variety of goals, and converse with the user. The company believes that games and simulations provide a great training ground for AI systems, as they are an approximation of the real world with visual diversity and diverse settings, mechanics, and graphical styles.
Key takeaways:
- Google's AI research arm, DeepMind, has released new research on its Scalable Instructable Multiworld Agent (SIMA), an AI agent that can follow instructions to carry out tasks in video games and play games it has never seen before.
- SIMA was trained using imitation learning techniques, where researchers recorded images and the keyboard and mouse inputs of human players to teach SIMA to play games like No Man's Sky, Eco, Teardown and Goat Simulator.
- DeepMind's interest in video games is because they are a good training ground for AI systems, with the hope that research like this enables it to understand how AI systems may become more helpful.
- DeepMind aims to improve SIMA's performance, including making its agents able to follow more detailed instructions, and to ultimately develop AI systems that can act in as many environments as possible and achieve a variety of goals as well as converse with the user.