Key points:
- Meta FAIR announced the release of a set of new research studies to support its goal of achieving advanced machine intelligence (AMI).
- The improvements cover touch sensing, dexterity, and human-robot interaction, which are essential elements for achieving AMI.
- Meta has established strategic partnerships with GelSight Inc and Wonik Robotics to develop and commercialize tactile sensing innovations, facilitating access to technology and supporting an open AI ecosystem.
Developing embodied AI capabilities to understand and interact with the physical world Meta’s Fundamental AI Research team, in partnership with the robotics community, is working to develop embodied AI agents capable of perceiving and safely interacting with their surroundings, as well as providing support in both physical and virtual worlds. We believe this is a vital step toward achieving advanced machine intelligence (AMI).
Key innovations in touch sensing and dexterity:
- Meta Sparsh: The first general-purpose vision-based touch sensor representative, capable of operating across multiple sensors and tasks thanks to self-supervised learning techniques.
- Digit 360: An artificial fingertip that provides precise multimodal tactile data, enabling researchers to develop AI that understands the physical world in finer detail.
- Digit Plexus: A unified platform that integrates tactile sensors into a single robotic hand, contributing to the advancement of robotics and artificial intelligence research.
Strategic partnerships to drive robotics innovation:
In collaboration with GelSight Inc and Wonik Robotics, Meta is working to develop and distribute advanced tactile sensing technologies. GelSight Inc will manufacture and distribute Digit 360 to facilitate researcher access to the technology, while Wonik Robotics will develop the next generation of the Allegro robotic hand, which will unlock new possibilities in research and innovation.
PARTNR: A new benchmark for human-robot collaboration The PARTNR benchmark is designed to evaluate and develop robot-human interaction in household tasks, enabling the study of collaboration and planning with advanced AI models.
Looking to the future: Through these technologies and partnerships, Meta seeks to advance robotics and artificial intelligence research fields while opening new horizons for multiple applications, including healthcare, manufacturing, and virtual reality. We are committed to fostering an open scientific community to empower future generations of AI robotics research.
