I wasn't sure what to expect when I sat down for a conversation with MIT professor, Yoel Fink. Interviews take all manner of forms, and from time to time, they feel like holding on for dear life. There’s no way to anticipate where the next question will take you, so you might as well just enjoy the ride.
Apologies in advance to the folks who leave negative feedback whenever an episode doesn’t dive too deep into the hard tech, but Professor Fink has seen stuff and done things and lived to talk about it on a podcast. It’s a fun episode, and I think you’ll get a lot out of it – I certainly did.
Vibes-wise, it’s the ideal episode for me this week. Much like the podcast, I had no idea what I was launching into these past few weeks. Some high highs, and some (aforementioned) unexpectedly low lows. I’ve walked barefoot through a waterfall in the Alps, explored Paris on a Lime bike, and rode through Germany on a highspeed train, sandwiched between a professional violist and a pair of British birds who gave me a great tip about the Louvre.
I’m typing this intro looking down on the inaugural Machina, the physical AI spinout of the Raise conference that kicks off the following day (Weds). I was chatting with some of the old heads – founders and CEOs of companies like Boston Dynamics, Agility, Apptronik, and ANYbotics, and there’s a universal surreality to the moment.
I’ve felt it the each of the last three weeks, in fact. In Chicago, in Davos, and now here in Paris. They’re dramatically different shows in key respects, but in their own ways, each is a clear indicator of how far this industry has progressed over the last few years. These are the overnight successes that are decades in the making people use to self-identify with a good bit self-effacement.
The humble ones will also tell you about the giant shoulders on which they’re perched, plenty of those shoulders are thankfully still in the room, breathing in the same recirculated air today.
Roboticists are also a largely pragmatic bunch, as well, and for every mention of a massive round, verboten terms like “bubble” and “crash” are invoked. Even the most optimistic among us agrees we’re probably in for some manner of self-correction.
My take at the moment: Just as it followed generative AI on the way up, physical AI will follow it on the way down, though this industry has some insurance in the form of those robotics companies that are deployed doing the hard work everyday.
The landscape 10 years from now will look very different from today’s. Familiar names will fall away and new giants will emerge. Physical AI will be a massive disrupter, but all disruptors eventually become the disruptee.
The next few years are going to be wild. High highs. Low lows. Might as well just enjoy the ride.
Ali Agha started with the hardest problems first. There’s the Mars helicopter and the search for extraterrestrial life, for starters. The researcher turned FieldAI founder says planetary caves are a prime spot to go looking, as the structures shield any potential life from harmful cosmic rays. We’re talking moonshots on moonshots here, adding another level of complexity of the hunt for alien life.
After an early career like that, this whole bit about scaling generalized AI ought to be a walk in the park, right?
“We were exploring solutions with different types of mobility systems, including four-legged robots, to go into these environments on Mars, and also on Titan,” Agha told me at a live episode of the podcast recorded at A3. “We published quite a bit on this, and did the first deployment in human history of legged robots traveling several kilometers with zero human touch — no prior map, no GPS, no predefined information — into Mars-analog caves, from cave networks in Oregon to Kentucky and beyond.”
It was a kind of robotics kismet when DARPA launched a Subterranean Challenge focused on outcomes of the variety Agha’s team was already determined to solve. Missions focused on industrial settings, underground environments, digital twins. And a variety of different operation.
“We submitted a proposal, were one of the very few funded teams, and our team grew at one point to more than 70 or 80 people — we ended up winning the urban phase of that competition worldwide, right before COVID,” Agha says. “That was an incredible experience, because we were putting brains and autonomy onto a heterogeneous set of assets. At the peak of our operations, we sent 11 robots — four with legs, four with wheels, three flying — into fully unknown environments, discovering them, carrying out a variety of missions as they explored the world and autonomously dispatched themselves.”
Agha and company soon moved onto another DARPA challenge whose goals aligned with the team’s work.
“Look,” Jonathan Hurst gestures toward the stage below. 1X CEO, Bernt Børnich, is up. Behind him a massive screen is splashed with the phrase, “the holy grail of robotics.” Said holy grail is an image of 1X’s tendon drive system.
Tendon drives have been a longtime interest for the Agility founder. If you want to write a research paper on the topic, you’ll likely find yourself citing something the Chief Robot Officer has written over the years.
These days, however, Hurst wants to talk physical AI. It was the subject of a panel he and I bookended late last year, that also featured researchers from Physical Intelligence and Dyna Robotics at an event in the South Bay. It was also the focus of the keynote presentation he gave at this week’s Machina Summit titled, “Humanoids Today: From Ambition to Real World Impact.”
When I note that his attention is laser focused on the subject these days, he’s quick to dispel any notion that he hasn’t been into physical AI since before it was cool (my words). Hurst cites attitudes toward his pre-Digit robot’s record breaking run back in late 2022.
“Physical AI is the path forward,” says Hurst. “There's really no other path. And we were very early into that. I mean if you remember Cassie had the 100-meter dash world record, but at the time the reason we had to get the Guinness record folks out to do the world record is because otherwise nobody would have noticed or cared. You know if we said ‘hey we've put reinforcement learning to control balance on a bipedal robot.’ Like seven people in our community would have cared. It was pretty early.”
That last sentence might as well be Agility’s tagline. Early to legs, humanoids, early to pilots, early to deployment. As for how the Oregon-based firm delineates the latter two, Hurst explains, “A pilot is something that sometimes a lot of times it's with an RD group at a company. It's also not on the production line. You’re not moving active products. If the robot breaks and fails, it doesn't really affect [production]. It’s very similar to or a replica of what the use case would be And it's usually like a limited amount of time. They'll do it for some number of weeks. Deployment in our case is a long-term contract — like three years, and the robot is doing work where it's handling product.”
Professor Fink reflects on his own unconventional path, from military service and years of backpacking to studying chemical engineering, physics, and eventually materials science at MIT. Watch on YouTube>
Russ Tedrake (Researcher/Stealth Startup)- After a decade serving as TRI's SVP of Robotics Research, the long time MIT professor is launching a new physical AI startup.
Rick Faulk (Locus Robotics) - When Amazon bought Kiva in 2012, the robotics startup's former customers were in a bind. Quiet Logistics did something about it, creating a behemoth in the process.
Aya Durbin (Boston Dynamics) - After helping deploy industrial robots at 6 River System, Aya Durbin is looking to do the same with Boston Dynamics' Atlas Humanoid.
Robotics Raises
Automated contributor Rebecca Szkutak rounds up the most consequential recent funding rounds in robotics, automation, and physical AI.
For a moment there the other week, there was some question about which humanoid company would be the first to go public. In many ways, Agility and Unitree are a study in contrasts, an American first mover with a laser focus on the industrial space and the Chinese company behind boatloads of low cost bipeds and quadrupeds. The SPAC route, which hit peak popularity during the pandemic, offered Agility a potentially abbreviated timeline to the NYSE. Unitree on the other hand, has been rumored to be aiming for the Shanghai Stock Exchange for some time now. The company just took a huge step toward that goal, earning approval from the China Securities Regulatory Commission for an IPO that would raise $160 million with the sale of around 40 million shares. According to reporting, "Unitree is now finalizing its underwriting plan, pricing and share subscriptions for a potential debut [on the Shanghai Stock Exchange] as early as late July."
China has more or less had a lock on big humanoid robot spectacles attached to major sporting events. This week during halftime at the Norway/Brazil World Cup Match at NYNJ Stadium, Boston Dynamics' Atlas gave all of those dancing Unitree systems a run for their money, demonstrating a production that remains both extremely powerful and dexterous. In its first public appearance (excepting, perhaps CES), the robot delivered the game ball to kick off the second half.
Did you hear? Mbodi AI walked away from the Automate Startup Challenge with the giant novelty check (and $10,000). Becca sat down with the company's cofounders on Automate Live to discuss the big win. Here's a quote from cofounder Xavier Chi, I'm including to keep my bosses happy. “It's our favorite show, actually, one of the biggest in the whole robotic space, and we just thought we should go on stage and pitch our product or technology and see how people react to it. We got some really good feedback from people from all across different domains, so that's a really great experience.” Mbodi is building bespoke, agentic physical AI models for different tasks, rather than the one-size-fits-all approach much of the industry is currently chasing down.
If you've (understandably) had your fill of French humanoids for the week, skip this one, and go to the story below. Liam wrote about robots and bees and mines. It's fun. For everyone else, here's a post from Remi Cadene, the former head of Hugging Face's LeRobot repository. Cadene, now serving as founder and CEO of UMA (Universal Mechanical Assistant), a Parisian startup with no less than Yann LeCun serving as advisor (good luck keeping UMI and AMI straight for the foreseeable future), gave a first glimpse at an early version of the startup's humanoid. "This robot integrates AI, software, and hardware, all developed from the ground up in nine months with a small team," Cadene says about the robot his team has christened 'Version 0.'" Designed and assembled in Paris, this prototype marks a significant step forward in our journey.
Let us take a moment to honor the humble bumble (well, honey) bee. They pollenate, they barf honey, and they take a live and let live approach to stinging, which is frankly a nasty process for all involved. We have a lot to learn from the little insects, including how to make mining safer and more efficient. Creating a swarm of robots can help decentralize operations, ensuring that when one system goes offline for any reason, the rest can keep digging.
Chalk it up to years of reporting on consumer and startup tech (or perhaps just being an American), but I have this knee jerk reflex where I ask most of the researchers I speak with about commercializing their technology. Thankfully, Psyonic founder and CEO, Aadeel Ahktar, cut out the middle man (me), and announced that this recently published paper from the company's lead engineer, Jesse Corman, and researchers at Northwestern University will form the foundation of the Ability Leg. "The research introduces wireless, skin-conformable haptic arrays that substitute missing foot sensation for people with spinal cord injuries and stroke," Ahktar writes on social media. "Pressure from a smart insole maps to vibrotactile and thermal feedback on the forearm in real time. Participants maintained balance up to 122% longer with the system. For Psyonic, this is the early foundational work on what will become the Ability Leg. The same principles that drive the Ability Hand, restoring sensation, closing the loop between the body and the world, are what we're building toward for the lower limb."
Last week, someone told me the notion of "killer robots" is something of an American — or at least Western — idea, inspired by decades of Terminators and the like. I will say that the guy I passed at the entrance of the Louvre who shouted, "it's the beginning of the end!" when he saw a small robot deployed to mark the opening of Raise had a distinctly American (or at perhaps a very flat Canadian) accent. Of course, the notion of weaponized autonomy is a very real threat. Several leading robotics companies have signed pledges promising not to do as much. And if you count drones, well, that particular ship (to mix transportation methaphors) sailed a long time ago. The U.N. is apparently drawing its own line in the sand, as Secretary-General, Antonio Guterres, called for a ban on weaponized autonomous systems. Guterres writes on the app formerly known as Twitter, "We may be the last generation able to set the terms on which humanity and machines coexist. If AI is to be powerful, it must be governed. If AI is to be trusted, those who build it must be accountable. If AI is to be global, it must be fair. And if AI is to serve the future, it must not consume the future. Let’s build a future of AI by humanity, with humanity, for all humanity."
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.
.webp?upscale=true&width=1120&upscale=true&name=Thumbnails_16X9_TRAINING_3%20(1).webp)
