First of all, yes, I’m a vain and deeply insecure man, and I chose the above photo to lead off the newsletter because it was a rare good hair day caught on camera. Thanks to the folks at the Pittsburgh Robotics Network for that (and for setting up that convo with Seegrid CEO, Chris Baker).
Second, please vote for Automated in the Signal Awards if you haven’t already. I’m told I get a trip to New York City, if I win, which I assume means I can expense the $20 Metro North ticket.
Third, I have a humanoid safety webinar coming up next Thursday. I’ll be speaking with Apptronik’s director of product safety, Shawn Schaerer, among others. Register here. I will be hosting it virtually, so there’s nothing to expense, but I would appreciate your attendance nonetheless. Also, check out this Intel-hosted webinar next Tuesday on agentic AI.
This week, we’ve got our second interview from the summer’s Davos Tech Summit. This time it’s ABB’s global portfolio and technology manager for Collaborative Robots. Fun chat. Also, happy slightly belated Women in Robotics Day.
On the feature front, we’re speaking with Hello Robot’s Aaron Edsinger and RobCo’s Roman Hölzl — both are exciting robotics startups with strong research backgrounds.
Something strange strikes me watching a handful of videos from Aaron Edsinger’s academic career: how many things haven’t changed. I am, of course, familiar with the breakthroughs — particularly those changes to how robots learn, perceive, and follow instruction. The past few years have been a prolonged deluge.
But these are the things you anticipate, when you sit down to watch circa-2007 humanoid videos. Domo (name swiped from the Styx rock opera) is a primitive robot by today’s standards, but footage of the CSAIL project is proof that many of the fundamentals have been in place for as long as the current generation of mechanical engineering majors have been on Earth.
The ability to take the long view is, at once, frustrating and hopeful. Two days after our visit to the offices of Edsinger’s Hello Robot, I find myself on stage with his former MIT professor, Rodney Brooks, with a newfound perspective on his own ultra-pragmatic views of this generation’s robot hype cycle.
If you stick around long enough, you see new projects starting with old premises and inevitably running into familiar problems. In the case of humanoids, the past couple of years have seen many of these complexities abstracted down to their industrial essentials.
“We see Stretch as robotic cubism,” says Edsinger. “We thought about the human form. We broke it apart, resembled it.
These cycles exist in all industries. Problems are corrected for, overcorrected for, re-problematized.
Those of us caught up in the language of it all are left to wonder whether bi-dexterous manipulation on a mobile base is enough to slap the h-word on something — and more importantly, whether the answer to that question means anything outside a startup pitch deck.
The question facing the industry now is whether the breakthroughs of the past several years present a large enough sea change to ensure that humanoids can push through the pre-stated challenges. Can large models and larger bank accounts move the industry toward a kind of robot singularity? Or are we doomed to live through the cycle once again?
Edsinger has spent almost a full decade of his robotics work focused on Hello. The Bay Area company’s Stretch robot lacks many of the hardware complexities of his earlier experimental projects. This is a product of the aforementioned abstraction, significant efficiency improvements to purpose-built hardware, the rapid sophistication of machine learning, and the necessities of designing a robot for scale, rather than pure scientific inquiry. Continue Reading>
Sell a solution, not a robot. It’s not a novel concept. The term “solution selling” dates to the mid-70s, while the more abstract notion of selling customers on answers no doubt stretches back further. There are moments it feels like a relic butting up against our current physical AI climate, but the equation is simple for industrial robotics: deliver or die.
A tremendous level of trust is required before a manufacturer will disrupt the line to incorporate some new piece of automation. Even well-established machine manufacturers grapple with this. How, then, is a new contender supposed to gain a toehold with a new technology?
Roman Hölzl describes RobCo’s approach as “an outsourced automation partnership.” Actuators, control modules, sensors — the topics that dominate our conversations are not the things the customer is made to care about. Technical feasibility, commercial viability, reliability — these are the foundations of trust. If a client wanted to discuss SLAM, odds are they would have planned a very different career path.
Hölzl, who cofounded the startup, notes that building trust with a customer begins in simulation. The ability to create task-specific digital twins is a foundational piece of the technology the team built at Technical University of Munich (TUM), before officially spinning out in 2020.
“We develop a full simulation concept of the robot that we're gonna deploy and therefore check for technical feasibility and commercial viability,” says Hölzl. “Things like cycle times, overall equipment effectiveness, output quality inspection, deep integration into existing software, machines on the floor. That's the first step. The second step is if our solution doesn't work, we're not getting paid in robot as a service (RaaS), because it's not an upfront CapEx investment. And then you go from there. With these two steps, we try and generate as much trust as possible.”
The process proved successful enough to land the young company a gig performing chassis inspections for automotive giant BMW. RobCo is expanding system use cases based on customer needs. Existing robots are also capable of things like palletizing and depalletizing, loading and unloading machines, and material handling. At this relatively early stage, many of these jobs are still fairly customized with bespoke hardware solutions, though the company anticipates a change as it scales.
Interestingly, Hölzl, who spent a year at industrial arm giant Kuka, says the company determined it cheaper to build a full stack solution, rather than rely on available off-the-shelf parts. He cites several factors.Continue Reading >
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. Watch on YouTube >
Karen Panetta (Tufts University) - From helping create the digital twin to solar-powered race cars, Karen Panetta has lived far more life than can fit into a single podcast. We still give it the old college try.
Ahmad Bahai (Texas Instruments) - Ahmad Bahai believes in the power of the semiconductor. Texas Instruments' CTO is a long-time industry vet with a unique outlook on how the next generation of hardware is fueling the AI revolution.
Tony Zhao (Sunday) - Are general-purpose home robots really that distant of a dream? Sunday is taking a unique approach to the humanoid space that challenges fundamental beliefs about physical AI.
Boston Dynamics has a new chief executive. Rohit Prasad has spent 12 years at Amazon, most recently heading the retail giant's voice assistant efforts as senior vice president and head scientist for Alexa and Artificial General Intelligence. Amazon worked to reinvent its smart assistant for the generative AI era with the launch of Alexa+, a version trained on large models with deeper contextual knowledge. Prior to his time at Amazon, Prasad worked at defense R&D firm Raytheon BBN Technologies. “Boston Dynamics is uniquely positioned to advance physical AI through its world-class robotics expertise," Prasad said in a release. "I am honored to join the company at such a pivotal moment for the industry. By bringing together advanced AI and robotics, we can build intelligent systems that create meaningful value in real-world environments, expand the reach of robotics across industries, and shape the next era of physical AI.”
The move comes amid a number of staffing changes at Boston Dynamics, as the company looks to commercialize its third product, Atlas. Parent company Hyundai views the electric humanoid as a foundation piece of its automotive manufacturing strategy moving forward. Last month, the companies announced the opening of the Robotics Metaplant Application Center, a testing and data training facility tied to the Hyundai Motor Group Metaplant America in Savannah, Georgia.
“Boston Dynamics' industry-leading robotics, Rohit's proven ability to turn AI into products at global scale, and Hyundai Motor Group's modern manufacturing, logistics, and mobility capabilities give us a powerful foundation to build and scale physical AI," said vice chair of Hyundai Motor Group, and chair of the Board of Directors at Boston Dynamics, Jaehoon Chang. "We are thrilled to welcome Rohit to lead the company into this exciting new era of real-world physical AI." The company has announced plans to deploy some 25,000 Atlas units at Hyundai and Kia plants "over the next few years," along with the goal of producing some 30,000 robots each year.
Currently filling up my speaking schedule through the end of the year, so if you're interested in having me or the Automated team involved in something, sooner is definitely better. Here's one I'm looking forward to next month, which includes (among other great chats) my first on-the-record chat with Russ Tedrake since Walden Robotics' official launch. I'll be hosting two sessions at the upcoming SOSV Deep Tech Summit:
November 6 @ 8:40 a.m. PT Russ Tedrake: Teaching Machines to Learn from Humans: Massachusetts Institute of Technology professor Russ Tedrake has spent decades working on one of robotics’ hardest challenges: building machines that can learn, adapt, and act autonomously in the physical world. After co-founding Walden Robotics and raising $300 million at a $1.1 billion valuation, he is putting machine intelligence to the test in the real world. Can robots learn from humans and from their own deployments, and ultimately handle the complexity and variability of real-world work?
November 6 @ 11:10 a.m. PT Re-Booting Industrial Robotics: Agility Robotics, Inbolt and Standard Bots: Industrial robotics is entering a new era as advances in AI, vision, and robotics expand what machines can do. VC investment is surging, and so are startup valuations in the category. Agility Robotics is preparing to go public at a $2.5B valuation, Standard Bots raised $200M this year at a $1B valuation, and Inbolt has raised over $25M while deploying across 100+ factories. These three robotics experts approach this future differently. Damion Shelton, who recently took on a role at Carnegie Mellon University, co-founded Agility Robotics, where he was a pioneer in bipedal humanoids. Standard Bots’ Leif Jentoft is building AI-native industrial robots from the ground up. Inbolt’s Albane Dersy adds AI-powered vision to existing robots. We’ll dig into how the latest wave of physical AI advances are changing the outlook for industrial automation. Are we approaching an era of general-purpose bipedal robots on factory floors, or will specialized machines dominate? How close are we to solving the human-robot interaction challenges?
This is the first time I recall publishing a story to Automated that involved a reporter penning their own disclaimer about squeamishness. Becca's warning should be heeded, of course — generally speaking, we don't publish a lot of stories about...humane dispatch. I definitely learned something editing the piece — specifically about Ikejime, a traditional Japanese process that inspired Shinkei's robot.
The rest of the world was content with a Thing from The Addams Family reference, but Liam went deep with a very specific episode of BBC cult sci-fi sitcom Red Dwarf. Given that the entire series' existence is just a PBS-fueled fever dream for my smooth American brain, I'm going to take his word for it [Please submit all feedback in writing to BBC Complaints PO Box 10031, MPN2, Lutterworth, LE17 4UE.] The thing I appreciate the most about this story, however, is that the team at ETH Zurich apparently saw a fictional depiction of a disembodied hand and wondered what it would be like to make one themselves. Big win for science, if you ask me.
Known knows? Unknown knowns? Some days you just want to toss up your arms and say "Who knows?" Maybe we can all take a cue from this team of MIT researchers and use a little more “SANDO” (for “Safe AutoNomous trajectory planning for Dynamic unknOwn environments”) in our lives. [As an aside, we're really stretching the bounds of acceptable vowel selection with acronyms these days.] According to the team behind it, the UAV autonomous navigation system offers "a mathematical guarantee of safety, even when encountering unpredictable obstacles in unknown environments. Per lead author Kota Kondo, "The only thing the planner needs to know is the top speed the obstacles could reach. Given that, you could use it in any environment, without a map, and you know the UAV is not going to crash into anything.”
Seems like every time I turn around there's another YC-backed robotics startup taking its first steps into the larger world. Rhem is further proof that assistive home robots are still receiving investment, though the company's approach is more along the lines of an Elli-Q or earlier efforts like Jibo than, say, Hello Robot or Weave. It's a desktop robot in the age of agentic AI. With that comes more conversational interaction, health monitoring, fall detection, and other pieces designed to help older adults retain relative independence.
The Association for Advancing Automation (A3) is North America’s largest automation trade association representing more than 1,400 organizations involved in robotics, artificial intelligence, machine vision & imaging, motion control & motors, and related automation technologies.
.jpg?upscale=true&width=1120&upscale=true&name=gr3-comparison-16x9-1%20(1).jpg)
