To everyone who made it out to last week’s happy hour and live podcast recording with Rodney Brooks: you rock. To everyone on the 100-person waiting list: apologies. We’ll be back soon. We’ve got more live events planned before the end of the year. More info soon.
All in all, it was a triumphant return to the Bay, and a great way to reflect on how much this project has grown in the past year. You can help us keep up the momentum by voting for Automated in this year’s Signal Podcast Awards. I’m sure there’s gratification in a job well done, but frankly my Webby Awards can use a little company.
This week we’ve got a pair of interviews from last week’s trip. I chatted with RealSense CMO Mike Nielsen about the Cognex acquisition after his panel at our Advanced Vision and AI event. You’ll also find a piece based on our recent trip to Maven Robotics’ office.
Dyna made a big announcement, as well, and we happened to be around to capture some of its new model in action. Another one of last week's subjects, Hello Robot also shared some news you'll find below.
In 2008, futurist Paul Saffo broke down the processes behind his forecasts in a blog post titled, “Strong Opinions, Weakly Held.” The eponymous philosophy, which he identified as a defining characteristic of his approach since the mid-1980s, remains an oft-repeated Silicon Valley truism decades later.
“Allow your intuition to guide you to a conclusion, no matter how imperfect,” Saffo wrote. “This is the ‘strong opinion’ part. Then — and this is the ‘weakly held’ part — prove yourself wrong. Engage in creative doubt. Look for information that doesn’t fit, or indicators that point in an entirely different direction. Eventually your intuition will kick in, and a new hypothesis will emerge out of the rubble, ready to be ruthlessly torn apart once again. You will be surprised by how quickly the sequence of faulty forecasts will deliver you to a useful result.”
Hamza Derbas offers a paraphrase as we discuss the work leading up to Maven Robotics’ recent $100 million Series A. “Strong convictions, loosely held” is one of “many” mottos for the young physical AI startup, albeit one of the ones he says his team enjoys the most.
“At Maven, we like it because especially in this age of AI, things move so fast,” Derbas tells me during a recent visit to the firm’s Bay Area offices. “You can be staring at a problem, thinking of the solution, and just lock in on something — and some really smart engineer or AI scientist somewhere comes up with a new way to think about the problem. If you're not ready to shed the past and chase the more scalable future, you're going to be left behind.”
For nearly nine years, spanning 2015-2023, Derbas served as the director of engineering for Apple’s Autonomous Systems, Special Projects Group (SPG). As for what, precisely, the job and the eponymous “special projects” entail, Derbas offers next to nothing. As with many other internal Apple projects, SPG appears to share its first two rules with those governing Fight Club. Continue Reading>
I thank Mike Nielsen for showing up in the first place. Over the years, I’ve dealt with several instances of speakers coming down with a well-timed illness just as some big news was about to hit. When I make some off-handed comment about it being a “bit of a busy week,” he smiles and notes that the previous day was “much worse.”
RealSense’s CMO happily obliges when I ask for a bit of his time following his presentation at A3’s Advanced Vision and AI event. After my quick introduction, Nielsen prefaces the talk by addressing the gorilla in the front row, noting that he’ll leave plenty of time for audience Q&A, including yesterday’s acquisition news.
Turns out a technical audience at a technical stage asks entirely technical questions. We all have our bailiwick. As someone who’s been writing about both M&As and robots for a long time, there are a couple of topics I’d like to broach once we step into the foyer.
Prior to joining RealSense, Nielsen served as senior director of Intel Incubator, an internal division of the silicon giant focused on facilitating the creation of startup-style projects. The division was also home to RealSense, which originated as a research project inside the company around a decade prior. The original depth sensing technology remains central to the brand’s core.
Intel spun RealSense off last July, in a bid to recenter its efforts on its bread-and-butter semiconductor business. The timing wasn’t ideal, given the massive acceleration of the robotics and physical AI sector, though the company maintained equity in its former division.
Startup mode lasted just over a year for RealSense.
“I've done startups before and know how painful it can be,” says Nielsen. “But it was, it was a startup with the complication of having to do high-volume manufacturing customers. We did, but I think we didn't really give enough credit at the time to how much of that stuff was actually being done by the company, being Intel.”
Over the following six months, according to Nielsen, RealSense “came together as a team. We brought in a bunch of new people from the outside, and we started to kind of rethink how we did stuff.” Continue Reading >
Panetta is Distinguished Professor and Dean for Graduate Education at Tufts University and founder of Nerd Girls, which uses engineering, robotics, and artificial intelligence to solve problems with real human impact. Watch on YouTube >
Ahmad Bahai (Texas Instruments) - Ahmad Bahai believes in the power of the semi-conductor. 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.
Carolina Parada (Google DeepMind) - Sometimes things require a fundamental rethink. During her time as VP and head of Robotics at Google DeepMind, Carolina Parada has challenged longstanding orthodoxies.
My bad. We probably should have waited another week before speaking withAMD'sAdaptive and Embedded Computing Group SVP/GM. AI pioneer Fei-Fei Li's World Labs came up a few times during our chat, but no one slipped up and mentioned that the chipmaker was on the verge of a little $8.2 billion purchase. AMD has agreed to buy the world model firm and —- judging from the above photo — traded outerwear as a sign of good faith. "We’re extremely thrilled to be part of a shared journey towards the future we want to build together," Li wrote in a post, where powerful AI platforms and systems are developed and deployed to help solve the most fundamental problems that humanity faces: from drug discovery to unleashing human creativity; from transforming learning to unlocking the next generation of robotics."
Psyonic has one of the cooler backstories in the world of end effectors. What began life as a prosthesis product has since entered the world of autonomous industry by way of the humanoid boom. The company's Ability Hand now has a decidedly form-factor-agnostic design. This week it announced that it is bringing the highly dexterous five-finger gripper to the end of Universal Robots arms. “Sophisticated automation shouldn’t be synonymous with difficult deployment,” Psyonic's Aadeel Akhtar says in a release. “By collaborating with Universal Robots to integrate our touch-sensing technology with the Gen 7 platform, we’re giving manufacturers and researchers a much simpler path to advanced, adaptive manipulation.”
We swung by Dyna Robotics’ Redwood City offices during last week’s Automated with Brian Heater whirlwind Silicon Valley Tour to chat with founder and CEO, Lindon Gao. The full episode is coming soon. In the meantime, we can share some iPhone video I captured of the physical AI firm’s Dyna-2.1, a physical agent that allows systems to execute multi-step processes. In this case, it's all the pieces that comprise doing a load of laundry. This week the company posted a one-hour video of the process, noting, “To the best of our knowledge, Dyna-2.1 is the first physical agent to perform hour-long, non-linear loco-dexterous manipulation workflows.” I popped my head into the studio where the video was being shot to capture some raw phone footage.
When we interviewed Josh Grunstein early this year, I inquired whether Tutor's Cassie robot was named in homage to another notable system from mechatronics' past. The CEO explained that the moniker derived from the automation work it's doing around shipping cases. On the other hand(s), Sonny, the other robot in the above image is, indeed, named for a famous robot — albeit a fictitious one. Naming conventions aside, Tutor still very much has its feet (wheels) planted in reality. Becca's follow-up with Grunstein yields some relevant quotes on the topic of promise versus performance, such as, "I speak to a lot of chief supply chain officers, who have built this worldview, which I think is right, that advanced robotics is kind of a scam, because every advanced robot needs a babysitter. Robots do run into edge cases, and pretending those edge cases don't exist is a great way to shove that burden onto the customer.”
Yes, you can, if you happen to be a Unitree G1 robot hanging around the hallowed halls of U.C. Berkeley or Stanford. The NorCal schools have helped develop BeyondMimic, a framework that combines reinforcement learning and a diffusion model to achieve that holy grail of humanoid functionality: spin kicks (plus cartwheels, flip kicks, sprints, and other fun stuff). "As well as the researchers' own observations," Liam writes, "the researchers also asked 77 other people to compare the motions of the robot using the BeyondMimic framework against the motions generated by Unitree's standard controller. The reports back showed that in 70.8% of the comparisons, the BeyondMimic’s movements were deemed to be more human-like and more natural-looking than those with the standard controller."
Hello Robot Inc Tuesday announced that it has been awarded a research grant from the National Institute of Health’s National Institute on Aging. The funding goes toward researching the use of the California startup’s mobile manipulator, Stretch, to assist patients with mild cognitive impairment and early Alzheimer’s. The grant, which finds Hello partnering with University of Illinois Urbana-Champaign, is for “nearly” $3 million stretched over a three-year period. “A robot’s value isn’t how complicated the task is. It’s how much agency that task gives someone,” cofounder and CEO Aaron Edsinger said in a statement. “This research will help us understand how Stretch can provide useful, trustworthy assistance while supporting family members and professional caregivers to help their loved ones and clients live more independently.” Last week, Automated with Brian Heater filmed an in-person interview with Edsinger at Hello Robot’s Martinez, California offices set to release soon.
Swimsuit season is now in the rearview, but here's some research out of MIT that could make you a touch less self-conscious before the next time you need to swim in front of a beach full of strangers. Turns out a little muscle can go a long way. In fact, a layer of well-positioned cells can get a robot moving pretty well along a body of water. With a skeleton comprised of "a film of gel that is about the length and width of a stick of gum," it begins to twitch when light is applied, generating enough motion to propel the system through the water. “It takes a lot of force to move through water versus air,” says study author Ritu Raman. “The robot’s quite strong, given its size.”
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.
