This audio is auto-generated. Please let us know if you have feedback.
CreateMe, an automated soft materials manufacturing company, has created systems that use physical artificial intelligence to revamp how textiles are made. To do so, the company replaces traditional sewing — which is particularly difficult to automate — with digitally bonded construction powered by robotics, proprietary adhesives and AI-driven manufacturing systems.
The company says its Modular-Engineering Robotic Assembly system is the world’s first autonomous tailoring platform, combining hardware and software that “deliver dynamic variability and flawless precision, accelerated with physical AI.” But challenges remain, especially related to handling cloth, which isn’t rigid and thus is harder for robots to handle consistently.
CreateMe CEO Cam Myers holds 25 patents in apparel automation technologies developed at the company. Prior to founding the company, he was on the founding executive team of Group Commerce, a venture-backed e-commerce platform ultimately acquired by Blackhawk Network, and previously held roles at DoubleClick and Allen & Co. He holds a B.A. and an M.A. from the University of Cambridge, and an MBA from Northwestern University.
Myers recently spoke with Manufacturing Dive about the current state of physical AI, how some companies are beginning to integrate the technology into their manufacturing processes and steps others can take to see if could work for them.
This interview has been edited for clarity and brevity.
MANUFACTURING DIVE: How does your company use physical AI now, and what are your plans for the future?
CAM MYERS: We're using some kind of foundational physical AI models, but we have quite a unique perspective because we're focused on deformable materials. We don't need to go horizontal, like these humanoids where they need to do dishwashing and then fold laundry and then take the trash out. We need to be very, very good within a much more narrow constraint — how do you deal with fabric, [including] the varieties of fabric.
We're using some of the foundational models but then specializing in a specific use case... I think this is where we're going to see the rubber hit the road in manufacturing as a more narrow application of physical AI... so it's not just point-cloud-to-point-cloud-location of moving robots. It can still be nondeterministic, but it's within a more constrained variety of outcomes.
So with that principle said, we're collecting data in three ways. We're doing your typical kind of teleoperation. We have a mold that is semi-automated with a human operator, and the human does the action. It's putting fabric on a mold, and then you fold it over to build essentially a neck band, which we see on a crew neck [T-shirt]. We [also] have human operators that actually [manufacture a] product... we’re getting training data that's being captured and that's being supplemented with two other data channels.
[We also] have a data channel with teleoperation, mimicking the same motions that the human operator would do putting the fabric on a device that the robot does. And then we've developed these interesting grippe technologies that are handheld... that sort of take some cues from medical uses of robotic hands.
We have this training data, and then at the end of the year these road maps will converge, and that's when we can start to strip out where we have steps that still have some human operators.
How do you see other other manufacturers using physical AI?
My belief is that physical AI is going to show up in the next three years in a more material way in manufacturing where the ROI is is going to be highest. What I'm really saying is more narrow applications are where you're going to see more adoption in the near term.
Some of these more general-purpose humanoid robots, I think, are a little bit further off. The [best] opportunity for physical AI applications versus other types of more traditional robotics or automation in industry is that [physical AI] can still start to get a little bit nondeterministic, but it's not completely open-ended.
You've got to pick and place these tools. But the next evolution is going to be where the variability massively increases — as an example, warehouses. It’s just finding the right stepping stones.
What impact will physical AI have on the manufacturing workforce?
I think the opportunity from a workforce development perspective is [that] the more rote, monotonous types of tasks will be taken over with automation. I also think that from a productivity perspective, you're going to see the ability to scale the number of technicians that can support [a given] number of lines.
So, for example, one of our more simplified lines only requires half a technician per line. One technician across two lines. From a productivity perspective, you're going to have a scalability factor. Humans are still going to be absolutely essential.
U.S. manufacturing should be focused on things that have high flex, high variability and can be quickly responsive to market. And it's not just apparel. Across manufacturing, you want what's closest to consumption, closest to the distribution centers, to be the highest variability. And so I think having a really strong workforce there is important. And some of the physical AI automation is going to help scale those people to support more variety across the market.
In terms of jobs, I think you're going to see many more workers at a technician level and maybe less at a production level over time. What I've seen is a range of industries are looking at moving to de-risk the supply chain [and] have some manufacturing capacity on shore that's frankly not been there for 30 or more years.
You’ve said that the AI boom is no longer just about chips, models and data centers, but also about rebuilding the physical industrial base required to support them. Can you expand on that?
I think you're seeing a lot of industry trade groups [be] supportive of that initiative. We're a founding member of Robots for America, which is a consortium of a whole series of robotics companies, but also industrial manufacturing companies. Other industry groups like the New American Industrial Alliance have been very supportive of [rebuilding physical infrastructure].
Industry said we need to be driving this because it needs to be a collective effort with different branches of government. Because if you see what's offshore in China, it's very much a top-down initiative. So I think you're seeing some stimulus more for industry to work with government.
I would say the other thing in terms of... the broadening of where physical AI is showing up across manufacturing environments is in the dynamism of new companies entering the space from an industrial technology and manufacturing technology perspective.
What's interesting in the last couple years is the go-to-market model of some of these companies, whether it's companies in defense tech, space tech or other kind of heavy manufacturing environments. [There] is a greater focus on selling the product to the end user, whether it's a government or an automotive company, rather than selling a piece of industrial technology that might use AI to another large industrial manufacturing OEM.
What steps should manufacturers take to begin incorporating physical AI into their operations or expand their use of it, both now and over the long term?
I think the biggest opportunity for American manufacturing is in high flex, [and] there is obviously a real advantage with simulating different production models... There are consultancies that work with the big software companies that build simulation software [and] help mid-small to midsize manufacturers look at digital twin simulation to help them decide where they might want to invest. There's obviously still a lot of steps involved, but I think it’s a tangible way to look at ROI and the trade-offs and things before you go and put a whole lot of money into buying robots and software.