What happens when a robot is no longer built for one fixed task, but can potentially be retasked through software to do thousands of different things?
Brian Ringley, HRI Designer at Boston Dynamics, joins Greg for the 150th episode of Machine Minds to explore that question through the evolution of Spot, Atlas, and the software systems that connect increasingly capable robots to the people and environments around them.
Brian’s path into robotics is anything but conventional. After starting in architecture, CNC machining, industrial robotic arms, computational design, and construction technology, he found himself working at the intersection of digital models and physical automation. That journey eventually brought him to Boston Dynamics, first to explore construction applications for Spot, then into product management and industrial inspection, and now into human-robot interaction design for Atlas.
In this conversation, Greg and Brian explore:
- How Brian’s background in architecture, digital fabrication, CNC machining, and industrial robot programming shaped the way he thinks about human-machine collaboration
- Why construction became an early proving ground for Spot, from reality capture and laser scanning to safety monitoring and project progress tracking
- The critical shift from thinking of “the robot as the product” to treating the robot as an invisible physical layer that creates value through the work it enables
- Why collecting more data is not enough, and how vision-language models and generative interfaces could let operators effectively “have a conversation” with factories and construction sites
- The enormous HRI challenge behind Atlas: designing an interaction model for a robot whose potential capabilities are effectively unbounded
- How reinforcement learning, behavior cloning, vision-language-action models, reasoning systems, and traditional software tools can work together across the Atlas autonomy stack
- Why Brian sees the real promise of humanoids as an economic breakthrough: one hardware platform that can be retasked through software instead of requiring new machines for every application
- Why general-purpose hardware also demands generalizable software interfaces, rather than building a completely new application every time the robot learns a new task
- How Boston Dynamics balances natural human-like movement with Atlas’s superhuman range of motion, while deliberately designing it as an industrial tool rather than a human imitation
- Why great hardware still matters in the age of physical AI, from reliability and field serviceability to strength-to-weight ratios, range of motion, manufacturability, and uptime
- What the Google DeepMind partnership brings to Atlas, and why frontier reasoning models and high-quality robotic hardware can accelerate each other
- Brian’s vision for the software-defined factory, where humanoids work alongside AMRs, fixed automation, inspection robots, digital twins, and other specialized machines rather than replacing them
- Why teleoperation will remain valuable even as autonomy improves, especially in nuclear facilities, underground mines, collapsed buildings, launch sites, and other environments that are unsafe for people
Brian also shares why robotics consistently gets overestimated in the short term and underestimated in the long term. Building systems that can understand natural instructions, reason about unfamiliar situations, manipulate the physical world, and operate reliably at scale remains extraordinarily difficult. But that difficulty is also part of what makes the field so compelling.
For anyone working in robotics, industrial automation, human-robot interaction, or physical AI, this conversation offers a grounded look at what it will really take to move humanoids from impressive demonstrations to flexible, useful tools embedded throughout the physical world.
Learn more about Boston Dynamics: https://bostondynamics.com/
Boston Dynamics’ Youtube Channel: https://www.youtube.com/user/BostonDynamics
Connect with Brian Ringley on LinkedIn:
Connect with Greg Toroosian on LinkedIn: https://www.linkedin.com/in/gregtoroosian/
