RoboBoom
Brighter with Herbert· February 10, 2024· 1:00:47

EXCLUSIVE: Why Tesla Bot Competitor Sanctuary’s Humanoid Will Shock the World w/ Suzanne Gildert

Suzanne Gildert, CTO of Sanctuary AI, discusses embodied intelligence, Phoenix’s dexterous hands and training approach, and the company’s plans for deployment with Brighter with Herbert host Herbert and guest Scott Walter.

In this talk

  1. 1:59

    Why embodiment matters

    Gildert sees robotics as a way to study how minds work, since human intelligence developed in bodies that move and interact with the world. She argues a humanlike body is especially important for developing humanlike intelligence, because touch, movement, and handling objects shape how people understand their surroundings.

  2. 5:44

    Phoenix learns reusable movements

    Phoenix is Sanctuary AI’s sixth-generation humanoid, with about 60 degrees of freedom and a focus on dexterous bench work. Its Carbon control system supports teleoperation for collecting demonstrations, then single-purpose or multipurpose AI modes. Rather than train only on complete jobs, Sanctuary is building a library of reusable actions such as grasping, turning, and placing objects.

  3. 12:58

    Hands, sensors, and training data

    Phoenix’s hands have 16 degrees of freedom apiece, fingertip haptic sensors, and visual feedback that helps estimate finger positions. Gildert describes combining sight and touch as an open research question, while teleoperation with HaptX gloves gives pilots force and tactile feedback. She says Sanctuary is gathering demonstrations and simulated examples to improve its models and narrow the sim-to-real gap.

  4. 17:58

    Deployment, scale, and social impact

    Gildert identifies retail and manufacturing as areas of customer interest and says early deployments will likely involve small proof-of-concept orders. She cautions against mass-producing robots before AI generalizes, and points to simulation as a way to test systems before scaling hardware. She predicts millions of humanoid robots by 2030, while arguing that early use could address hard-to-fill and dangerous jobs.

  5. 43:36

    Values and speculative futures

    Gildert says she is exploring whether quantum computing might relate to consciousness, while stressing that the idea is speculative. She believes imitation learning can carry human values through demonstrations, and argues that people and institutions should shape those examples. She also imagines robots supporting work beyond today’s labor needs, including running experiments in biomedical labs.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks