Ya Durban, Atlas product strategy lead at Boston Dynamics, and Alberto Rodriguez, Atlas behavior team director, discuss manufacturing applications, robot training, reliability, and the path to deploying humanoids.
Rodriguez says many factory jobs remain manual because specialized automation is too costly for work with high variation. A car line may handle five to ten models, multiple trims, and many colors, creating countless small tasks. Humanoids could address that variability, with tasks such as picking up bolts and fastening them beneath a car roof as a long-term target.
The team identifies hardware, task behavior, and application software as three areas that must become dependable. Rodriguez describes three product phases: make hardware reliable enough for sustained trials, adapt software to customer needs, then scale deployments while addressing rare hardware failures. Durban says legs support a flexible platform for varied spaces; Rodriguez cites compact movement and quick turns, while noting that safety certification for balancing robots remains unresolved.
Rodriguez contrasts older systems built from many interconnected algorithms with neural networks trained through experience. He describes pretraining as building broad capabilities and post-training as adapting performance to a particular job through demonstrations and corrections. Experienced teleoperators use VR and body trackers; gathering roughly five to ten hours of examples for one behavior can support task training. A broader training corpus could reduce the work required for later deployments.
Durban says near-term applications need capable, reliable behavior that matches each customer's expectations. Rodriguez chose manufacturing for its scale, structured workflows that can support a gradual safety ramp-up, and varied tasks that reward general-purpose systems. He calls automotive assembly a strong test of manipulation skills that could transfer to work beyond factories.
Rodriguez expects thousands of robots in industrial settings within five to ten years, while Durban compares Atlas's maturity path to Spot and Stretch. They say Atlas is not yet ready for routine sale; early customer work, including an on-site partnership with Hyundai, is part of product development. Durban suggests starting with ergonomically difficult work and larger pick-and-place tasks, where reliability may be more attainable than precision fastening in tight spaces.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
A Boston Dynamics Prototype Operations team member explains how the company builds and improves Atlas, from precision machining through assembly and testing, alongside its engineers.
Boston Dynamics engineers discuss the design, simulation, and dexterity of Atlas’s new hands.
Boston Dynamics’ robotics team discusses Atlas’s path to industrial work at Hyundai Motor Group’s Robotics Metaplant Application Center, with a host visiting the facility.
The Atlas Director of Robot Behavior at Boston Dynamics talks with NEXT host John Koetsier about Atlas’s design, robot intelligence, factory integration, and why legs can suit industrial work.
Chris Thorn, Atlas hardware team lead at Boston Dynamics, discusses the new Atlas design with industrial designers Aaron Abroff and James Kucera at Boston Dynamics.
Zack Jackowski, Atlas development lead at Boston Dynamics, and company researchers discuss training humanoid robots for factory work; Demis Hassabis of Google DeepMind explores AI’s capabilities and risks; Palmer Luckey of Anduril describes autonomous weapons; and artists and engineers in Carrara debate robots in sculpture.
