The Sequence Robotics - Issue #957: Generalist and the Rise of Promptable Robots
How physical pretraining could change the way we teach machines to work
TL;DR
- Generalist is developing robot models that rely on large-scale physical pretraining to create a reusable foundation for robot behavior.
- The goal is to enable robots to learn new tasks with only a short demonstration, rather than requiring bespoke datasets.
- The company's models have progressed through GEN-0, GEN-1, and GEN-1.5, showing increasing success rates and advancements in physical prompting.
- A key open question is whether the reported gains will hold up in complete, long-horizon jobs during actual deployments.