No Priors Ep. 58 | The argument for humanoid robots with Brett Adcock from Figure

In this talk

1. 3:07

   Why build humanoid robots

   Adcock argues that labor is a much larger market than transportation and that robots able to perform useful work could reduce the cost of goods and services. He says a human form provides a general interface to spaces already built for people, avoiding the need to develop a separate robot for every task.

2. 7:15

   Why the timing may work

   Figure had been operating for about 21 months, with Adcock describing autonomous bin moving and speech-guided tasks such as picking up an apple or making coffee. He points to improvements in batteries, motors, bipedal control, and AI as reasons humanoid robots were less feasible five years earlier. He also emphasizes major challenges: reliability, safety, sustained performance, and learning a wide range of physical tasks.

3. 15:18

   Building the team and product

   Adcock says he spent the first year defining Figure’s mission and engineering organization, then recruiting through hundreds of calls and cold emails. Figure prefers to buy components but builds some itself because mature suppliers do not meet its needs. Its product process sets customer and safety requirements, uses design reviews and trade studies, and relies on fast prototyping to find problems sooner.

4. 18:28

   Commercial deployment and AI

   Adcock says companies were paying for robots as Figure began deliveries, with plans for robots to work in client facilities over the following 12 to 18 months. He describes BMW’s Spartanburg plant as Figure’s first announced commercial customer and says the companies had selected five initial use cases. Figure’s work with OpenAI combines language-level task planning with Figure systems that control the robot’s movements and grasping.

5. 31:49

   Home use, trust, and AGI

   Adcock envisions robots taking on household tasks such as laundry, cooking, and tidying, with costs falling as production scales. He says social acceptance will depend on gradually proving safety and building trust through industrial deployments. He also argues that data from robots interacting with the physical world could contribute to progress toward AGI, while acknowledging uncertainty about that outcome.