A brief reprieve from travel this week, and a few days to catch my breath. The Automated team will be back on the road in no time, booking a week’s worth of podcast interviews with some of the biggest names in Boston robotics.
As mentioned on yesterday’s livestream with Joyce and Tom of MassRobotics, we’ll be celebrating our arrival in Beantown with the first-ever Automated Happy Hour. This socially awkward robotics blogger is looking forward to channeling his inner extrovert for a few hours to discuss your startup, research, fund, etc.
More info (and where to sign up) soon.
This is the first of what we hope will be many of these trips. Our friends at MassRobotics are already giving us the hard sell on a return visit. We’ll definitely be back, but there are a bunch of cities we need to get around to first.
As mentioned in last week’s newsletter, we’re still working through GTC interview overflow. This week, we’re featuring chats with Physicl, Flywheel, and Roboforce. I closed out the week by chatting with Kewazo’s CEO, following the German company’s recent $16 million raise, which brings its total funding up to $35 million.
You’re also getting our third-ever in-person podcast recording this week. This one was recorded with PlusOne robotics CEO and cofounder, Eric Nieves, back in January at the A3 Forum event. He’s a good guy to discuss robotics with, if you get the chance.
Lastly (for now), thanks to everyone who came up to me and shouted out Automated at GTC. It really means a lot. Double points if you did so while we were standing within earshot of A3’s president. We’ve covered so much ground in the six months since we launched this project, and we’re growing in all sorts of fun ways – most of which I can’t talk about. So, stay tuned, I guess.
Construction is high on the list of things I don’t miss about city living. For the better part of a decade, I could see and hear a seemingly endless project roughly 200 feet from my window. More often than not, the streets of Manhattan are lined with scaffolding, and for years, LaGuardia was the airport to be avoided at all costs, due to on-going repairs. Now that those are done, New York has transformed JFK into a similarly post-apocalyptic nightmare. That’s life in the big city. It’s ever-changing, always under construction, and the rodents are always looking for a new place to live – sometimes in your kitchen pantry.
Construction has always been – and will almost certainly always be – a massive business. The global market was reportedly valued at $16.45 trillion last year, and is set to hit $17.26 trillion for 2026. Projections put that figure at $21.73 trillion for 2030. These sorts of unfathomable numbers are precisely what caused me to do a double take when Kewazo CEO Artem Kuchukov told me the startup had shifted from its construction focus since the last time we had spoken, and had begun, “really honing down on industrial because we learned that industrial space is much more ready to pay because they have more capital.”
The German company had continued to occupy a construction robotics role in my brain – and why not? There’s plenty of money to be made, as countless founders and investors have pointed out to me over the years. A quick visit to Kewazo’s site surfaces the names of some massive industrial customers for its Liftbot System, including Chevron, Intel, and Dow.
Kuchukov concedes that, while construction holds massive potential for the company, industrial plants currently make more sense for its immediate roadmap.
“There are multiple components to this,” he explains. “One of the biggest ones we saw is this connection of different actors, stakeholders in the value chain. In the construction industry, owners of the building don't care how their buildings are being built. They only care about the cost, and the cost is dictated by what general contractors are going to tell them. But in general, building owners are not going to tell their general contractors how to build. The same goes with general contractors and sub-contractors. General contractors are not going to mess with specific trades and dictate them how things should be done. It becomes a lot of disconnect.”
Asked for differentiators, Leo Ma offers two words, “scalable and deployable.” RoboForce is far from the only humanoid company with such stated objections, but such goals are frequently deprioritized, as companies seek massive funding for long runways, rending real world use a distant dream.
“If you look at the whole landscape, there are companies trying to build a very scalable system that are limited by how much actual value will be created, in terms of deployment,” says Ma. “The other extreme – if you look at traditional systems, they are focused on one task and one machine, that is very deployable. What we have been focusing on is the intersection. We have to be extremely smart in focusing on anything that checks these two boxes.”
I confess to being thrown by the…practicality of it all. RoboForce shares a name with a handful of media properties, including an animated series with shades of RoboCop, executive produced by The Rock that is currently streaming on Tubi. Here’s the synopsis: “In 2089 Detroit, RoboForce was rendered obsolete by the more advanced UA101 bots and forced into menial jobs with no hope of being heroes. That is until a mysterious code virus infects the 101’s, and no one besides RoboForce can stop them.”
The design of the startup’s Titan robot also doesn’t do much to beat the enforcer charges. When a friend at GTC asked if the robot had any lasers, he wasn’t inquiring about LiDAR. RoboForce’s tagline is a bid to dispel that initial rection, noting that the Bay Area company produces, “Robotics for Humanity.” Similarly, when I ask Ma about the name, he assures me that it’s a force for good.
“A cup’s really easy,” says Jonathan Stevens. “It’s a solid object. There’s not really much happening there.” Lightwheel’s chief evangelist gives me the quick and dirty of generating one in simulation using a sensor. A refrigerator door, on the other hand, can be a surprisingly complex bit of work.
“Everyone has opened one of those. It takes a lot of force because it has got that seal,” he adds. “But when that seal breaks, all the force it takes to open that door is almost gone. It just swings open.”
Lightwheel is capturing that data so that you – and your robots – don’t have to. Stevens alludes to deployment of robot arms tasked with opening and closing fridge doors, over and over again, generating sensor data to hew as closely as possible to the real world, when mapped out in simulation. In other words: limiting the real-to-sim gap, to help eliminate the sim-to-real gap, to better train robots that execute real world tasks, in order to continue to refine that world through real-world data collection. Simple as.
Founded in 2023, Lightwheel is one of a growing number of startups that don’t build the simulators that train the robots, so much as populate them with a digital approximation of the assets in which those systems will interact.
“Think of NVIDIA as a company that creates 75% of what you need, because everyone has some different problem,” says Stevens. “Why would they want to create all the assets for simulation when they don't know what asset you need? That should be up to you. But then we're finding companies so they don't have the expertise to create physics ready assets, so they come to people like us who have been doing it for a while now and have it way as repeatable and predictable.”
“Digital cousins” is a new one to me. A quick Google search surfaces a research paper from 2024, coauthored by AI pioneer, Fei-Fei Li, titled “Automated Creation of Digital Cousins for Robust Policy Learning.” The research addresses the expense and other issues involved in generating digital twins, proposing as a stop-gap, “the concept of digital cousins, a virtual asset or scene that, unlike a digital twin, does not explicitly model a real-world counterpart but still exhibits similar geometric and semantic affordances.”
So, digital assets that are, you know, related, but not that related. That is to say that, when constructing simulation, there are those times when pretty close has to be close enough. And hey, if it was good enough for Patty Duke, who’s to say it’s not good enough for the occasional world model?
From what I can gather, two and a half years later, the term has yet to catch the world ablaze, though Alex de Vigan did casually drop it in conversation last week, which is precisely why we spent the last two paragraphs talking about the thing. He was, of course, referring to its use – along with digital twins – in the context of his recently announced startup, Physicl.
“Basically, you have inputs in a real object, but you don't have all the inputs to create the exact specific,” he notes, contrasting the digital relatives. “So there is going to be from 10-20% approximation on some of the inputs. Based on that level, depending on the type of input, you can create the different types of output.”
This specific aspect of Physicl’s work, in turn, came up during a conversation about the company’s fascinating history. The startup is a kind of spinoff of existing company, Nfinite – a kind of company cousin, if you will. It was founded by members of that team, including de Vigan, who remains CEO of both.
We know Amazon’s had its eyes on humanoids for years now. The retail giant toyed with the form factor well before announcing its initial pilots with Agility’s Digit systems. In addition to potentially deploying bipeds in amongst its millions of warehouse robots, reports have suggested that company could also deploy systems to help packages get that final mile. Though its recent acquisition of RIVR may be doing a lot of that heavy lifting in the meantime. This week, Amazon announced the surprise acquisition of Fauna Robotics, a company we’ve had our eyes on since its debut.The New York City startup’s first product, Sprout, is far from an industrial system. In fact, my initial excitement around the product was a focus on developers – something this industry rarely sees for any sustained timeline. I've also been hearing good things about the robot from those with first-hand experience. Could that point to Amazon jumping back into the developer side, which has largely been dominated by NVIDIA? What about home robots? Astro has been hanging in limbo and the iRobot deal is long dead. Sprout, meanwhile, is one of the closest things to Unitree’s robots produced in the U.S. thus far (though it’s still far from an accessible price point at around $50k).
"To our community: please know that we are still selling the Sprout Creator Edition robots to new customers and will provide continued support to all our existing customers—essentially, no change to the work we're doing together," cofounder and CEO Rob Cochran noted in on LinkedIn. "We are thrilled about what joining the Amazon team means for our future. Going forward, we will proudly operate as Fauna Robotics, an Amazon company." The startup's 50-person staff – including both cofounders – will become Amazon employees in NYC. Amazon tells me, "We are excited about Fauna's vision to build capable, safe, and fun robots for everyone. Together with Amazon’s robotics expertise and decades of experience earning customer trust in the home through our retail and devices businesses, we're looking forward to inventing new ways to make our customers’ lives better and easier.”
Henceforth, Fauna will be "Fauna, an Amazon company." Amazon tells me they're looking to learn from creators' continued use of Sprout, which could signal some interest in getting in on the ground floor with developers, be it through AWS, or in a manner similar to either NVIDIA or the work Hugging Face is doing with its open-source Le Robot community. Notably, the company's comment also mentions the home, retail, and devices – categories that include Astro and various Echo/Alexa products. This, in turn, could signal learnings from Sprout being applied across the company's consumer lines and, perhaps, retail locations, including Whole Foods. All speculation on my part here, but we're still early days with it.
Unitree Friday filed an IPO with the Shanghai Stock Exchange, in a bid to raise $4.2 billion yuan ($610 million). The Shenzhen-based Unicorn has become a massive star in the international market over the past several years, owing to the proliferation of its many bi- and quadrupedal robots made available at an (relatively) accessible price point. Founded in 2016, the company got another publicity bump during the recent Spring Festival, when its humanoid robots were seen busting moves on the world stage. Unitree says it shipped 5,500 humanoids last year alone, comprising roughly one-third of the global market. Unlike much of its high profile competition, Unitree hasn’t focused on industrial settings, opting to target research and other non-labor use.
OpenAI is taking its toys and going home. The ChatGPT-maker launched the video generating app in late September, hitting one million downloads in less than five days. Six months later, the company is saying sora long. The news follows a recent all-hands meeting, in which CEO of applications, Fidji Simo, reportedly warned against investing too much into “side quests” – a category that apparently involved that $1 billion Disney content deal. Ultimately, it seems, Sora was a massive use of resources – both in terms of the company – and the compute required to power it. The company’s video generation technology will, however, continue to play a role in its physical AI efforts, as it becomes increasingly involved in the robotics world.
Researchers at MIT have created a wristband that uses ultrasonics to track a wearer’s wrist muscles, tendons, movements, and hand gestures. Paired with AI (why not?), the corresponding output can be used to control the motions of a robot hand in the real world or simulation. The team behind the work believes such a system could one day prove useful in training robots. “We think this work has immediate impact in potentially replacing hand tracking techniques with wearable ultrasound bands in virtual and augmented reality,” said MIT professor, Xuanhe ZhaoT. “It could also provide huge amounts of training data for dexterous humanoid robots.”
Generative AI could help robots “see” through walls and into boxes, per new research out of MIT. The technology is being used in tandem with surface-penetrating wireless signals, to give computer vision systems the outline of an obscured object. The AI models essentially fill in the gaps of the shapes that can’t be discerned by wireless signal alone. “What we’ve done now is develop generative AI models that help us understand wireless reflections. This opens up a lot of interesting new applications, but technically it is also a qualitative leap in capabilities, from being able to fill in gaps we were not able to see before to being able to interpret reflections and reconstruct entire scenes,” says researcher Fadel Adib. “We are using AI to finally unlock wireless vision.”
"We're on an ambitious journey to give software a body to do physically productive work that the world needs and become the USA's leading supplier of modular drones and robots," says Lucid Bots cofounder and CEO, Andrew Ashur. For now, that journey largely features drones pressure washing the sided of buildings – and hey, finding uses for drones beyond imaging is no small feat. The Charlotte, North Carolina firm just announced a $20 million Series B that brings its total funding up to $34 million. Lucid said it currently has a bit under 1,000 systems in operator hands, generating $75 million in revenue.
A week after getting some facetime at GTC, Agile Robotics is back with some big partner news. This time, the German firm joins Apptronik in announcing a research deal with Google DeepMind. Agile has been around since 2018, focusing on lightweight industrial arms. The Munich firm announced its first humanoid, the fittingly named One, in November of last year. There’s not really a lot to go on as far as details of the partnership, though the pair says they’re focused on industrial and manufacturing tasks for starters – no major surprise there. "Agile Robots has already installed over 20,000 robotics solutions worldwide, proving intelligent automation at scale,” says Agile CEO Zhaopeng Chen. The huge opportunity ahead lies in autonomous, intelligent production systems that can transform entire industries. Integrating Google DeepMind’s Gemini Robotics models into our robotic solutions positions us at the cutting edge of this rapidly growing market."
Eric Nieves (Plus One) - Plus One's cofounder and CEO joins us backstage at the A3 Business Forum to discuss labor, humanoids, and keeping people in the loop.
Gary Cohen (iRobot) - iRobot has been a cornerstone of Boston robotics for a quarter-century. The Roomba-maker's CEO discusses its uncertain future.
Mehul Nariyawala (Matic) - A better robot vacuum was only the beginning for Matic. The startup is building a home platform.
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.
