SIMA 2

SIMA 2 is a generalist AI agent designed for virtual 3D worlds, capable of understanding and accomplishing long and complex tasks. It is the next milestone in research creating general and helpful AI agents, integrating the advanced capabilities of Gemini models. SIMA 2 can follow human-language instructions, think about its goals, converse with users, and improve itself over time.

SIMA 2 works by integrating Gemini models, which enables it to understand more complex and nuanced instructions than its predecessor. It can transition to learning in new games exclusively through self-directed play, developing its skills in previously unseen worlds without additional human-generated data. The addition of Gemini has also led to improved generalization and reliability, with SIMA 2 being far more successful at carrying out instructions, particularly in situations or games on which it’s never been trained.

Researchers, academics, and game developers who need to create general and helpful AI agents in virtual 3D worlds can get the most value from SIMA 2. It provides a significant step towards generalist, interactive, embodied intelligence, with important implications for the future of robotics and AI-embodiment in general. SIMA 2’s capacity to transfer learned concepts and apply them to new situations is foundational to achieving broad generalization seen in human cognition.