Advertise on podcast: Automated with Brian Heater

What if the biggest bottleneck in robotics is the way robots are taught? Jagdeep Singh believes the industry needs a much larger and more diverse source of training data to build robots that can handle the unpredictability of real work. At Rhoda AI, that starts with internet video. In this episode of Automated, Brian Heater speaks with Jagdeep, co-founder and CEO of Rhoda AI, about the company’s approach to physical AI, why he sees limits in vision-language-action models, and what it will take to move generalist robots from demonstrations into factories and warehouses. Jagdeep explains how Rhoda pretrains its model on hundreds of millions of video clips to learn how the physical world moves. The goal is to build an understanding of motion, forces, and object behavior before teaching a particular job. He says that foundation has allowed Rhoda to learn specific tasks with 10–20 hours of robot data, rather than the much larger datasets associated with other approaches. The conversation explores why footage of a specific task is not always the best starting point. Jagdeep describes how broader video datasets performed better than more narrowly curated collections of people doing jobs. Task-specific learning comes afterward, followed by additional learning from the unusual situations robots encounter with customers. That distinction became clear when Rhoda took a container-folding task from the lab into a factory. Broken latches and rusted equipment changed the conditions the model had learned. Jagdeep says the team adapted using roughly three hours of additional factory data, illustrating why deployment requires more than repeating a successful demonstration. Brian and Jagdeep also discuss how learning can compound across customers, why access to deployment data is central to Rhoda’s business, and the difference between customers owning their data and allowing the company to use it for model training. The conversation then turns to hardware. Jagdeep outlines four requirements driving Rhoda’s development: payload capacity, reliability, safety and dexterity. He explains why the company is building components it could not buy, how it is weighing the tradeoffs of three-finger hands, and why its mobile manipulators use wheels. Finally, Jagdeep explains why customers ultimately judge robots by success rate and cycle time. He shares Rhoda’s plans to deploy its model on existing hardware while developing its own platform, the role of human demonstrations in teaching new tasks, and why reliable performance in customer environments is the test that matters. Connect with Jagdeep Singh https://www.linkedin.com/in/jagdeep-singh-3b2a9222 Learn more about Rhoda AI https://www.rhoda.ai/ Explore Rhoda AI’s research https://www.rhoda.ai/research/direct-video-action We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected]. You can find the transcript and more episodes of Automated at automated.fm. Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Humanoid robots are receiving more attention than almost any other form of physical AI. But Andie Zhang says factory buyers still care more about speed, precision, reliability, safety, and ROI than a human shape. In this episode of Automated, recorded at the inaugural Davos Tech Summit, Brian Heater speaks with Andie, Global Portfolio and Technology Manager for Collaborative Robots at ABB, about why industrial robot arms are not going away, where humanoids still fall short, and how physical AI can unlock factory tasks that traditional automation cannot solve. Andie explains why the form of a robot should follow the job it needs to perform. Humanoids are designed to do many different things, but that versatility does not automatically make them the best option for industrial manufacturing. When a specialized robot can complete the same task faster, more precisely, and at a lower cost, companies have little incentive to pay for additional joints or capabilities that do not improve the application. The conversation also explores how physical AI can expand automation without replacing the systems already working reliably inside factories. Structured and repetitive tasks can still be handled through traditional programming. AI becomes valuable when robots must work in unstructured environments, interpret sensor information, adapt their behavior, or handle flexible objects that cannot be positioned the same way every time. Andie describes how simulation, digital twins, synthetic images, and AI vision are helping manufacturers prepare and test automation systems before deploying them on-site. She also explains how ABB developed an AI-based system that can recognize and pick flexible clothing packages, a task that has traditionally required large numbers of workers because every package looks and behaves differently. Brian and Andie also discuss why labor shortages are accelerating automation around the world. After visiting more than 100 customer plants, Andie says manufacturers consistently turn to robots because they cannot find enough qualified workers. She shares how cobots are helping experienced welders teach robots to perform repetitive welds, allowing companies to increase productivity without removing those workers from the process. They also examine the safety questions surrounding humanoids, the difficulty of developing dexterous five-finger hands, and the importance of making increasingly complex robots usable for factory workers and system integrators. Finally, Andie reflects on becoming the first person responsible for cobots at ABB, building an internal robotics incubator, taking robots directly to small businesses, and creating more opportunities for women and people with transferable skills to enter the robotics industry. Connect with Andie Zhang https://www.linkedin.com/in/andiezhang/ Learn more about ABB Robotics https://www.abb.com/global/en/areas/robotics Explore ABB collaborative robots https://www.abb.com/global/en/areas/robotics/products/robots/collaborative-robots Learn more about RobotStudio HyperReality https://www.abb.com/global/en/areas/robotics/innovation/robotstudio-hyperreality We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected]. You can find the transcript and more episodes of Automated at automated.fm. Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter You can also find us on: LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/ LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

AI should not be judged only by how advanced it is. It should also be judged by who it helps and who its designers remembered to include. In this episode of Automated, Brian Heater speaks with Karen Panetta, Distinguished Professor and Dean for Graduate Education at Tufts University and founder of Nerd Girls, about using engineering, robotics, and artificial intelligence to solve problems with real human impact. Karen traces that philosophy back to student projects involving a solar car and Thacher Island, where her team helped power historic lighthouses with solar energy. The goal was never simply to build one car or install one system. It was to show young engineers that the same technical skills could move from one application to another and improve people’s lives. That approach now runs through her work in food safety, firefighting, traffic management, marine health, infrastructure, and assistive technology. Some projects have become startups, while others serve populations too small to attract traditional investment. Karen explains why universities remain essential when an important problem does not come with an obvious market. The conversation also explores her early work at Digital Equipment Corporation, the simulation tools that helped lay the foundation for digital twins, and the moment at NASA Langley when software she was accustomed to testing in isolation began controlling real jet and propulsion hardware. Brian and Karen also discuss bias in physical AI, accessibility, and the business cost of designing for only one kind of user. From early crash-test dummies to modern interfaces, Karen argues that excluding people with different bodies, abilities, or ways of learning is both a design failure and a missed market. Her larger message is simple: engineering is not reserved for people who fit one academic mold. Creativity, imagination, determination, and a willingness to work across disciplines are what turn technology into meaningful change. Learn more about Karen Panetta at Tufts University: https://engineering.tufts.edu/about/undergraduate-and-graduate-deans/dean-karen-panetta Learn more about Nerd Girls: https://nerdgirls.com/ Explore Karen Panetta’s research: https://www.karenpanetta.com/ We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

AI and robotics may look like software revolutions. But every breakthrough depends on the semiconductors underneath them. In this episode of Automated, Brian Heater speaks with Dr. Ahmad Bahai, chief technology officer at Texas Instruments, about the chips, power systems, sensors, and edge-processing technologies enabling the next generation of AI and robotics. Ahmad explains why companies cannot predict the next major market a decade in advance. What they can see are the underlying trends, including the growing demand for power density, bandwidth, timing accuracy, and local processing. Those needs now extend from massive AI data centers to mobile devices and robots that must react without waiting for the cloud. Brian and Ahmad also trace his path from early Wi-Fi research at Bell Labs to Kilby Labs and Texas Instruments. Ahmad reflects on what made Bell Labs unique, why that model is difficult to reproduce today, and how companies can connect academic research with technology that can be manufactured at scale. The conversation also covers the semiconductor supply chain, the search for more efficient AI algorithms, and why technologies like autonomous vehicles and humanoid robots take longer to reach the mainstream than early predictions suggest. Finally, Ahmad explains why supposed physical limits rarely end technological progress. Scientists once argued that chips could not scale below 55 nanometers. Engineers found another way, and today the industry has reached two nanometers. Learn more about Dr. Ahmad Bahai and semiconductor innovation at Texas Instruments: https://www.ti.com/video/6372229233112 Learn more about Texas Instruments: https://www.ti.com/about-ti.html Read TI’s insights on scaling humanoid robots: https://www.ti.com/about-ti/behind-chip/articles/what-will-it-take-to-bring-humanoid-robots-into-the-real-world-read-our-experts-insights.html We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected]. You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

A robot does not need another robot to learn how to work in a home. That idea is at the center of Sunday Robotics. The company is building Memo, a general-purpose home robot trained with human data collected through its custom Skill Capture Glove. In this episode of Automated, Brian Heater speaks with Sunday Robotics co-founder and CEO Tony Zhao about physical AI, robot learning, and why he left Stanford without keeping an academic fallback. Tony traces the company's foundation to ALOHA and the Universal Manipulation Interface, two research efforts that changed how he thought about scaling robotic intelligence. Tony explains why data remains one of robotics' biggest bottlenecks. Sunday's glove closely matches Memo's hand, allowing people to capture high-quality manipulation data while performing ordinary tasks. This approach is cheaper and easier to distribute than collecting every training example through a robot. Brian and Tony also discuss how smarter AI could compensate for simpler, lower-cost hardware, why Memo could eventually be priced more like a gaming PC or phone, and why a home robot does not need to work at human speed. If the robot can finish the dishes, laundry, and tidying while its owner is away, reliability and capability matter more than speed. Finally, Tony explains why a polished robot video is not the same as repeatable real-world deployment. He argues that the variability of homes creates what Sunday calls research-market fit, pushing robots toward the broad, adaptable intelligence they would need to work around people and eventually move into services and industrial environments. Connect with Tony Zhao https://www.linkedin.com/in/tony-z-zhao Learn more about Sunday Robotics https://www.sunday.ai/ Apply for the Memo beta program https://www.sunday.ai/beta-program Explore ALOHA https://tonyzhaozh.github.io/aloha/ Explore the Universal Manipulation Interface https://umi-gripper.github.io/ Learn more about ACT-2 https://www.sunday.ai/blog/act-2-preview We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at https://automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Ten years ago, Google DeepMind’s robotics team wanted to build a single model that could go from pixels to motor control. People thought the idea was crazy. Today, much of general-purpose robotics is moving in that direction. In this episode of Automated, Brian Heater speaks with Carolina Parada, VP and head of robotics at Google DeepMind, about the ideas behind Gemini Robotics and the company’s plan to build an AI layer that can power many different kinds of robots. Carolina traces her path from speech recognition at Google to self-driving perception at NVIDIA and explains why she has repeatedly pursued technical fields on the verge of transformation. That experience taught her to question established approaches instead of settling for incremental improvements. She explains why general-purpose robots face an even harder challenge than autonomous vehicles. Roads have lanes, traffic lights, and established rules. Robots operating around people must navigate unstructured environments, adapt to unfamiliar objects, and understand how to “read the room.” Carolina also breaks down the family of models behind Gemini Robotics, including embodied reasoning, vision-language-action models, and reinforcement-learned whole-body control. She explains how these systems work together and why a smaller on-device model may be more useful when reliability matters more than maximum intelligence. Brian and Carolina discuss why useful robotics does not have to wait for physical AGI, how Gemini Robotics is expanding what Boston Dynamics’ Spot can do during inspections, and why logistics and manufacturing are likely to see capable systems first. They also explore Google DeepMind’s partnerships with Apptronik, Boston Dynamics, and Agility Robotics, and why Carolina believes one robot will not win it all. Finally, Carolina explains why robot data alone cannot scale and why physical AI will require a mix of real-world experience, simulation, human video, and robots learning from their own mistakes. Connect with Carolina Parada https://www.linkedin.com/in/carolinaparada Explore Gemini Robotics 2 https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/ Register for the Automated Happy Hour with Rodney Brooks https://luma.com/n688dcl5 Learn more about the Advanced Vision & AI Conference https://www.automate.org/events/advanced-vision-and-ai-conference We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Humanoid robots are attracting enormous investment and dominating the conversation around physical AI. But Clara Vu believes the form factor is a trap. The more a robot is expected to look and work like a person, the more people assume it has human-level capabilities. In reality, forcing spinning motors into a body designed around tendons can add unnecessary constraints before engineers even begin solving the task. In this episode of Automated, Brian Heater speaks with veteran roboticist and Veo Robotics co-founder Clara Vu about what nearly three decades of building autonomous systems has taught her about the gap between a compelling demo and a product that works in the real world. Clara traces her career from joining a 10-person iRobot before she was legally old enough to drink to working on oil-well exploration robots, Hasbro's My Real Baby doll, and the software framework later used for Roomba. She also shares how toy manufacturing taught iRobot to build at consumer scale, why Harvest Automation chose potted plants over fruit picking, and how her work at Rethink Robotics helped lead to Veo. The conversation gets into one of the hardest truths in robotics: a prototype that works 80% of the time may represent only 2% of the work required to ship. Clara explains why edge cases, environmental variability, reliability, and exception handling consume nearly all the effort, and why a task that takes three to five years to commercialize should look like something a robotics graduate student could prototype in three to five weeks. Brian and Clara also examine the current humanoid boom. Clara argues that venture capital rewards companies that can claim enormous markets, which makes a general-purpose humanoid an appealing pitch. The problem is that the more environments and tasks a system must handle, the harder it becomes to make that system reliable enough to deploy. Instead, Clara makes the case for a new version of systems integration that combines existing robot arms, sensors, computer vision, AI, and software with application-specific engineering. That model may not produce one machine that does everything, but it could automate far more of the difficult work happening in factories, farms, warehouses, and other environments today. The technology already exists to automate many of those tasks. The bigger challenge is choosing the right problems, allowing for profitable customization, and building funding models that support focused solutions. As Clara puts it, if the job is moving a pallet, make the pallet jack intelligent instead of building a humanoid to operate it. Connect with Clara Vu https://www.linkedin.com/in/clara-vu-244941/ Learn more about Veo Robotics and its acquisition by Symbotic https://www.symbotic.com/news/symbotic-acquires-veo-robotics-to-enhance-efficiency-and-safety-innovation/ We'd love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter You can also find us on: LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/ LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

AI can speak fluently, reason through problems, and describe the world. Getting it to do something in the world is a much harder problem. In this episode of Automated, Brian Heater speaks with Dhruv Batra, co-founder and chief scientist at Yutori, about the gap between language models that generate answers and intelligent systems that can take action. Dhruv explains why many post-ChatGPT robot demos are still little more than a physical wrapper around a language model. A robot may be able to hold a convincing conversation, but that does not prove it can navigate, manipulate objects, understand physical space, or respond to the movement of people around it. The conversation begins with an even more fundamental question: What does it mean for an AI system to have a belief? Dhruv traces the idea back to Bayesian probability and explains why a rational decision-maker should never assign exactly zero probability to an event. Once a possibility reaches zero, no amount of new evidence can change that belief. Brian and Dhruv also examine the human language used to describe artificial intelligence. Words like “belief” and “thinking” carry rich meanings in everyday life, but AI researchers often use them in much narrower technical ways. Dhruv argues that this trade-off can create confusion, while also giving researchers something precise enough to measure and improve. They trace Dhruv’s path from probabilistic machine learning and computer vision to visual question answering, Grad-CAM, embodied AI at Meta FAIR, and eventually Yutori. Along the way, he explains how his team trained virtual robots to navigate directly from pixels to actions without building a map, using LiDAR, or separating perception from planning. The conversation also explores why simulation has earned a complicated reputation in robotics, where sim-to-real transfer works well for locomotion but remains far more difficult for raw camera images and dexterous manipulation. Dhruv recalls the blunt question roboticists asked whenever he presented work in simulation: “Did you touch a robot?” At Yutori, Dhruv is now applying many of those embodied AI ideas to web agents. These systems perceive a browser through screenshots and take actions such as clicking buttons, entering information, and completing tasks. He calls them “robots of the web.” Unlike traditional automation, they do not depend on a website remaining perfectly structured or unchanged. The result is a wide-ranging conversation about uncertainty, paradigm shifts, robot learning, web agents, and the difference between an AI system that can answer a question and one that can act on your behalf. Connect with Dhruv Batra https://www.linkedin.com/in/dhruv-batra-dbatra/ Learn more about Dhruv Batra https://dhruvbatra.com/ Learn more about Yutori https://yutori.com/ We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

The future of robotics may not belong to one machine that can do everything. It may belong to teams of specialized robots that learn how to work together. In this special episode of Automated, recorded at the Davos Tech Summit in Switzerland, Brian Heater speaks with Amanda Prorok, professor of collective intelligence and robotics at the University of Cambridge and founder of the Prorok Lab. Amanda explains why she believes the world is built on collective intelligence. From ants using pheromone trails to find the shortest path to food to robot teams dividing complex work among specialized members, nature repeatedly shows that cooperation can produce capabilities no individual agent possesses alone. That principle produced a surprising result in Amanda’s own research. Her team gave a group of AI agents one simple objective: score a goal. Without being taught human soccer strategy, the agents organized themselves into goalkeepers, defenders, and attackers. They independently discovered the same role specialization humans developed for the game. Brian and Amanda also explore the challenges of coordinating robots with different bodies and capabilities. Amanda explains how drones, wheeled robots, and legged robots can share information even when they move through the world differently, and how a “blind” robot can navigate using cameras distributed throughout its environment. The conversation then turns to one of the biggest unresolved questions in physical AI: trust. Many systems are being deployed based on empirical performance rather than mathematical guarantees. A robot may work well most of the time while still failing in ways researchers cannot fully explain. Amanda shares a personal moment that brought that uncertainty into focus after she stood beside a humanoid robot with her three-week-old baby. She also describes her lab’s work on remotely detectable robot policy watermarks, which could allow someone to record a robot with a phone and verify the origin of the policy controlling its behavior. Connect with Amanda Prorok https://www.linkedin.com/in/aprorok/ Learn more about the Prorok Lab https://www.proroklab.org/ View Amanda’s University of Cambridge profile https://www.cst.cam.ac.uk/people/asp45 Explore the robot policy watermarking research https://www.proroklab.org/publications/iclr2026-watermarking/ We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Get a direct line to the biggest names and brightest minds in robotics, Physical AI, and automation. Automated with Brian Heater brings you long-form conversations and unfiltered insights into how we got here, where we’re going, and what’s behind the technologies impacting how we live and work.  The show features conversations with visionaries from AWS, NVIDIA, Universal Robots, and beyond. From groundbreaking research in humanoid robotics to real-world industrial applications, each weekly episode explores the people and ideas shaping the future of work, industry, and society. Subscribe and be the first to get new episodes on Wednesdays. LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Humanoid robots can walk, gesture, and look astonishingly capable in a polished demo. Then Jan Liphardt bought one for his house, along with three quadruped robots, and discovered how quickly the illusion breaks. After charging the batteries and getting the humanoid to stand, it simply stood there. It could not become the R2-D2-like helper, companion, teacher, or coworker he had imagined. In this episode of Automated, Brian Heater speaks with Jan Liphardt, founder and CEO of OpenMind and associate professor of bioengineering at Stanford University, about the software, intelligence, and trust systems that robots still need before they can become useful parts of everyday life. Jan explains why the final 3% of humanoid deployment can consume nearly all the work. A robot that performs well in a lab still has to handle traffic, animals, wet pavement, moving leaves, different languages, and countless other edge cases before it can operate safely in an unstructured environment. The conversation explores why OpenMind is deliberately not trying to solve every robotics problem. Rather than building hands, arms, or high-speed control systems, the company combines its own models with technology from other teams. Its OM1 runtime and FABRIC coordination layer are designed to help different robots become more capable, understandable, and useful around people. Brian and Jan also discuss why social intelligence requires much more than speech. A compelling humanoid has to track attention, move its head and shoulders, use its hands, understand body language, remember preferences, and respond in ways that feel natural. Jan even argues that a robot’s imperfections can strengthen the bond between the machine and the person helping it. They also dig into one of OpenMind’s most distinctive ideas: connecting sensors and models through natural language so people can inspect how a robot reaches a decision. That approach trades some speed for intelligibility, giving developers a clearer way to debug behavior, add guardrails, and improve governance. Jan also challenges the idea that robotics has one universal data problem. Teaching math, folding a T-shirt, and completing an assembly task require very different data and system designs. In some cases, he argues, a robot may learn more effectively from physical constraints than from watching millions of videos of humans. The conversation also covers Jan’s path from physics and bioengineering into robotics, why elite degrees do not always predict great engineers, what a 17-year-old calling from a robot field trial taught him about hiring, and why combining academia with the real world can produce better research questions. Connect with Jan Liphardt https://www.linkedin.com/in/jan-liphardt Learn more about OpenMind https://openmind.com/ Learn more about Jan’s work at Stanford https://profiles.stanford.edu/jan-liphardt Register for the Automated Happy Hour in Mountain View on September 22 https://luma.com/n688dcl5 Learn more about the Advanced Vision and AI Conference on September 23 and 24 https://www.automate.org/events/advanced-vision-and-ai-conference/register We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Chef Robotics had millions of dollars in contracted revenue from fast-casual restaurant chains. Rajat Bhageria still chose to walk away. The demand was real. The model was not viable. Restaurants only operate at peak production for a few hours a day, while a useful robot may need to handle as many as 90 different ingredients. The economics and technical reality did not support the company Rajat wanted to build. In this episode of Automated, Brian Heater speaks with Rajat, founder and CEO of Chef Robotics, about the difficult decision to leave restaurants behind as the company’s starting market and focus on high-mix food manufacturing instead. Rajat explains how the team assumed food factories were already automated until it visited plants where hundreds of people were still portioning meals by hand. Unlike a restaurant, these facilities can run production for as many as 16 hours a day. A robot on a high-mix assembly line may also need to handle four or five ingredients instead of every ingredient on a restaurant menu. The conversation reveals why so much fresh-food production remains manual. Low-mix products such as soup, cereal, or bagged vegetables can justify dedicated automation. Fresh meals, wraps, salads, and prepared foods require constant changeovers across hundreds or even thousands of SKUs. That flexibility still often comes from people working in refrigerated production rooms. Brian and Rajat also examine the robotics graveyard. Rajat argues that many robotics startups do not fail because their technology is poor. They fail because they choose the wrong customer, solve the wrong problem, or never make the unit economics work. That same tension now hangs over the humanoid market, where capital runway may determine which companies survive long enough to reach commercial scale. They also discuss what it takes to win a manufacturer’s trust when a startup has no track record. Chef structured early deployments around clear acceptance tests, embedded its engineers at customer sites, and spent a year sending team members into plants before sunrise to find bugs, release new software, and test it again the next morning. Finally, Rajat explains why Chef’s lack of food-industry insiders became an advantage only because customers acted as true design partners. The team learned food safety, production planning, and line operations directly in the field. It also discovered that the hardest robotics problems were often the least obvious, including clear trays on white conveyors, sloped floors, changing line speeds, and placing food without spilling it. The result is a candid conversation about product-market fit, customer pull, real-world robotics, and why shipping a useful system requires much more than building impressive technology. Connect with Rajat Bhageria https://www.linkedin.com/in/rajatbhageria Learn more about Chef Robotics https://www.chefrobotics.ai/ We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at https://automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Humans know robots are machines. That does not stop us from naming them, caring about them, or feeling uncomfortable when someone hurts one. In this episode of Automated, Brian Heater speaks with Kate Darling, Research Lead for Robotics, Ethics & Society at RAI Institute, about why social robots may be far more valuable than their current capabilities suggest. Kate believes every social robot we have seen so far is still in the PalmPilot stage of innovation. The breakthrough may come when people stop asking what a robot does and begin recognizing companionship itself as the application. But that emotional connection also creates risks. Kate explains why empathy is not a finite resource, even if attention is, and how companion technologies can be designed to serve a company’s interests instead of the user’s. The problem is not simply that people form relationships with machines. It is what businesses may do with those relationships. Brian and Kate also discuss humanoid robot safety, including what happens when an emergency stop causes a dynamically balanced robot to collapse. Kate explains why the workers who understand factories and warehouses need to be involved in these decisions before robots are deployed. The conversation also challenges one of automation’s most familiar promises: that robots will take over dull, dirty, and dangerous work. Kate argues that these labels are rarely defined and can overlook cultural differences, underreported injuries, and the parts of a job that workers actually enjoy. They also explore the future of work and why robots are not independently coming for anyone’s job. Companies make choices about how automation is used, whether it replaces a worker or improves an entire workflow. Finally, Kate shares the unforgettable Pleo experiment that changed the direction of her career. After participants bonded with a group of baby robot dinosaurs, they refused to damage them, even when threatened with losing every robot in the room. The experiment was not a formal scientific study, but the participants’ emotional response revealed something Kate has spent her career exploring: people can know a robot is a machine and still feel deeply compelled to treat it as if it were alive. Connect with Kate Darling https://www.linkedin.com/in/kate-darling-37a181149 Learn more about Kate’s work https://www.katedarling.org/ Learn more about RAI Institute https://rai-inst.com/ Read The New Breed https://www.amazon.com/New-Breed-History-Animals-Reveals/dp/1250296102 We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

Humanoid robots can walk, fold laundry, and even do backflips. Nic Radford wants to know whether anyone will pay them to work. In this episode of Automated, Brian Heater speaks with Nic Radford, co-founder and CEO of Persona AI and a former NASA roboticist who helped lead the development of Robonaut 2 and Valkyrie, about what it takes to turn a humanoid robot into a durable business. Nic explains why Persona is starting with welding and shipbuilding through its partnership with HD Hyundai, rather than waiting for a single general-purpose robot that can do everything. He breaks down the market questions that matter just as much as the technology: labor scarcity, task complexity, customer purchasing power, tool use, safety, unit economics, and the willingness to adopt. The conversation also gets deeply technical. Nic’s career as a high jumper, and the titanium ankle he now lives with, helped fuel a long-running fascination with biomechanics. He explains why the human ankle is so difficult to recreate, why adding motors at the bottom of a robot’s leg creates an energy problem, and how humanoid designers have to balance capability against overengineering. Brian and Nic also revisit Nic’s years at NASA. Nic explains how radiation can flip bits in a robot’s machine code, why Robonaut used three processors in each actuator, and why building a robot for the deep ocean's pressure can be even harder than building one for space. They also explore the original vision for Robonaut as an astronaut helper and emergency responder, the role DARPA played in advancing modern robotics, and why moonshot programs create lasting value even when the original goal remains out of reach. Finally, Nic shares what humanoid robotics can learn from the nearly 20-year arc of self-driving technology. The biggest lesson may be that impressive hardware is only one part of the journey. The real test is whether the technology solves a valuable problem, reaches customers, and gives them a reason to ask for another robot. Connect with Nic Radford https://www.linkedin.com/in/nicolaus-radford Learn more about Persona AI https://persona.ai/ Learn more about Persona AI’s work with HD Hyundai https://www.prnewswire.com/news-releases/hd-hyundai-and-persona-ai-sign-agreement-to-deploy-humanoid-welding-robots-for-shipbuilding-automation-302449258.html We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected] You can find the transcript and more episodes of Automated at automated.fm Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/

The ability to move is about far more than getting from one place to another. It can shape independence, confidence, social connection, and quality of life. In this special episode of Automated, Brian Heater speaks with Kathryn Zealand, founder and CEO of Skip, the Google X spinout behind MO/GO, and Jessica Bath, DPT, PhD, assistant professor of physical therapy at UCSF. Kathryn explains how a personal experience with her grandmother helped lead her from theoretical physics and evaluating moonshots at Google X into wearable robotics. She breaks down why Skip thinks about MO/GO as an “e-bike for walking,” how the company is building foundation models of human movement using real-world gait data, and why hiking offered a better starting point than the structured environments where many exoskeleton companies begin. The conversation then turns to Parkinson’s disease, a subject that has become deeply personal for Brian following his father’s diagnosis and recent passing. Kathryn and Jessica explore freezing of gait, fall risk, and what researchers still do not understand about the neurological signals behind movement. They also discuss how wearable robotics, deep brain stimulation, exercise, physical therapy, and better movement data could help people remain active and independent. MO/GO itself is a consumer recreation device and is not FDA-cleared. Skip’s work around Parkinson’s uses a separate prototype and is being studied through clinical research. Kathryn also shares the deeply personal story behind Project Stardust, a research-stage effort working toward better at-home pregnancy monitoring that she helped launch after the loss of her son, Ziggy. It is a conversation about movement, dignity, data, and what it means to build technology around problems that genuinely matter to people. Connect with Kathryn Zealand: https://www.linkedin.com/in/kathryn-zealand/ Learn more about Skip and MO/GO: https://www.skipwithjoy.com/ Learn more about Project Stardust: https://project-stardust.org/ Connect with Jessica Bath: https://www.linkedin.com/in/jessica-bath-pt-dpt-phd-a2331499/ Jessica Bath at UCSF: https://profiles.ucsf.edu/jessica.bath Support Parkinson’s research through The Michael J. Fox Foundation for Parkinson’s Research: https://www.michaeljfox.org/donate We’d love to hear from you. Have thoughts or guest suggestions? Reach us at [email protected]. You can find the transcript and more episodes of Automated at automated.fm. Unlock full access to Automated and explore everything automation. Subscribe today and leave a review on YouTube, Apple Podcasts, Spotify, and the Automated Newsletter. https://www.youtube.com/@automatedpodcast https://podcasts.apple.com/us/podcast/automated-with-brian-heater/id1837762221 https://open.spotify.com/show/60olq6brlBEIJWggx2fMR6 https://www.automate.org/automation/automated-newsletter LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/ Instagram https://www.instagram.com/automatedpod/