Twelve tries.Then it knows.

CURRENT STATE

01 Locate

Autonomous manipulation study. One continuous policy, shown at actual speed.

DEMOS: 012
POLICIES: 001
TASK: NEW

01 / THE WORK

The real world does not come with a training set.

Anthrolab builds learning systems that acquire physical skills from a handful of demonstrations, then carry them into objects, grippers and rooms they have never seen.

01Few-shot: New behaviour from demonstrations you can count, not a warehouse of trajectories.
02Transfer: The same policy keeps working when the hardware, object or lighting changes.
03Generalization: The long bet is a robot that can be told what to do instead of retrained.

02 / EARLY ACCESS

Twelve tries.Then it knows.

The real world does not come with a training set.

See what the next run can do.