Zachary Karabell (l) and Rodney Brooks (all photos by Asa Mathat)
Zachary Karabell (l) and Rodney Brooks
Founder, Chairman and CTO, Rethink Robotics
Head of Global Strategy, Envestnet
Read the full transcript below. (Transcript by Realtime Transcription.)
For more about Techonomy Detroit, including a complete archive of videos and transcripts, click here.
Karabell: I think we are going to start with something we were talking about before, which was the degree to which there’s a lot of emphasis on education and skills. We read lots of stories about industrial America now becoming a high tech job that actually can’t find sufficient people with sufficient skills. But I believe you have a somewhat different take on it than that kind of common narrative, which is we’re not training our workforce for the industrial and factory work of tomorrow, and that we need to first focus on that before we then create this vibrant, new industrial system.
Brooks: Yeah, so I want to start—who has something like this? Yeah.
Did any of you take a course at the community college on how to use this UNIX‑based machine? No.
It was designed so that you could learn how to use it yourself. It was designed on some precedence that were common in the culture. But it was designed on the idea that you would learn how to use it yourself.
I think in manufacturing, we have suffered from it being an engineering domain. Manufacturing engineers like to know how complicated their machines are, so they become more and more difficult to use. It’s about complexity and the complexity is exposed.
But what if, instead, we went like we did with consumer products, and hide the complexity, and, instead, make them easy to use so that factory floor workers can use the most complex machines that there are, and be productive with them?
So we bring people along on the job to be able to be even more productive by providing the right tools.
Karabell: Although you could say, to push back on that, when presented with that device, a smart phone, which has as much computing power as a large UNIX system had in the 1970s—
Brooks: Way more.
Karabell: Way more. But it does require certain skills, like being able to search for information that you can’t just acquire—
Brooks: No, I said it was based on the culture. An example that came into mind was iRobot—I am not part of that company anymore—but we made the Roomba, and military robots that we used in Afghanistan and Iraq for roadside bombs.
Initially, when we sent those robots over there, we had complex menu systems. Then when we got feedback from 19‑year‑old kids out there, with bullets flying past their heads, saying this isn’t really useful to us, we put a game controller on it, and within five minutes of play, they had 80 percent of the capability of the robots. And they never got to that with two months’ training before. So find the right cultural reference.
Now, people are generally used to, for instance, touch screens. They are generally used to pretty stuffy DVDs in their faces. So if you build on that, you can get some pretty useful tools, pretty straightforwardly.
Karabell: There’s also another common narrative, and I am probably guilty of this common narrative, about the way in which the changes in industrial America have been a real challenge to labor forces, and, in particular, to places like Detroit. Separate from global trade, the very fact that if you open a factory today, because of the robotics that you have been so central in creating, those robotics mean that you can have 500 people having far more output, far more efficiently, with just‑in‑time manufacturing and flexible floors, than 5,000 people would have been able to do on fixed‑line manufacturing 30 years ago.
So the rub has been, is technology the way someone’s been—Andrew McAfee and Erik Brynjolfsson, two of your former colleagues, professor at MIT, talked about the way in which technology is actually a middle‑class job destroyer, not a middle-class job creator. But you have a different perspective, I think, on the role robotics and automation can play in industrial societies and the way in which that links to—
Brooks: Yeah. Conventional automation has been—I liken it to the mainframe computers, which most of you are too young to know about, but back when I was a boy, computers were behind the glass wall. Ordinary people didn’t touch them. Then came the PC, and then ordinary people started interacting with PCs. And there came a tremendous flowering. So think about it that other way around.
Can I show some slides here? I’ll just show you what I’ve been doing. Do I need to press a button?
So this is Baxter, a robot my company, Rethink Robotics, has been building. We started shipping in January. It’s a production assistant robot. The key thing about it, though, is that it is safe for ordinary people to touch—that is a worker in the factory, in that particular factory.
But the workers can program the robot and learn to program the robot in minutes. So we have had cases where we will ship the robot to a factory, a factory worker has worked on the line for 25 years, has learned to program the robot within an hour of arriving, and got it to do a useful task. And they do it by grabbing the robots arms, by using the buttons on the forearms to select menu items that come up on the screen.
We don’t use touch screens because a lot of times people in factories wear gloves, which doesn’t work real well with touch screens. But ordinary people are programming the robots, and then they are safe to interact with.
This is an example from a plastics factory. You will see a couple of Baxters here with people, in close with them, sharing the tasks, doing the—some of the dull, repetitive tasks.
Interesting thing here, as we zoom in on the hands of these robots, the hands were all made locally within the factories. Sometimes in the really small factories they go out and use text shop to do the 3D printing of the fingers.
So the next panel is going to be about the maker movement. So this is a robot that we have, and what we’ve been finding at a lot of the factories is that the workers come up and ask, can we get the robot to do this particular task? Because we don’t like doing this task. It’s really boring. Can they do it for us?
Karabell: And the smiley face is to make people think we come in peace?
Brooks: No, that’s very important. It’s not about smiling. The robot glances where it’s about to move, so it’s telegraphing what it is going to do.
So if you are around it, you are never surprised by its motion. Just as you glance where you are about to reach out, it’s telegraphing its intent. It’s aware of people and it will slow down. If people do get in its way, it stops. So it’s being aware of people, and it’s signaling them what it’s going to do.
Karabell: If we come back in 10 years to Techonomy 2023, will I be engaged in a conversation with Baxter? Is that your view?
Brooks: No, Baxter is pretty dumb. It’s got a little bit of common sense. It’s not a trajectory controlled robot, like all the robots are in Detroit; instead of following trajectories, it is doing tasks. If it drops something, it doesn’t keep moving with an empty hand. It goes back to pick it up, because it has an understanding of the task level of what it’s doing. So it’s not truly intelligent.
Karabell: So in what way will that facilitate a revitalization?—Not just of manufacturing in terms of output, because it’s very clear that automation, robotics, all this software, all these systems allow for much greater productivity. It is still less clear how it leads to greater opportunity of employment and activity, particularly in an urban or any area.
So how does this then translate? I mean, we talked a little bit about small factories would have been the norm prior to the late 19th century, because it was easier to organize a factory of 50 people. We didn’t really know how to do huge factories. But that was a real source of economic vitality. Do you see these being a way to create smaller manufacturing hubs, literally?
Brooks: Yeah. I see what we have gone to is a very top‑down command control sort of architecture. But information technology, as it comes to manufacturing and many components, allow new sorts of business models. I use this all the time. Does anybody use MFG.com website? Anybody heard of it?
Okay, well, I looked today. There was about $73 million of business being advertised there. It’s very small facilities, some just two people, some 10 people, who build small rooms of parts. If you want prototype parts or smaller parts, you go to the website, you put up the CAD drawings, you bid on it, you select who will do to work, and 48 hours later, they arrive at your place.
Why does this work? It works because of standard CAD drawings, PDF, Web, credit card payment services on the Web, and FedEx. When you put all those things together, little tiny companies that used to have an addressable market of about three miles, now have a national market. And it’s really changed the dynamics for rapid prototyping and small‑batch manufacturing.
So the new tools put together imply new business models. And I think we often forget about how radically the PC and the progeny of the PC completely changed business models. I know later today the CEO of Twitter is going to be here.
One of my friends, a few years ago, Deb Royce, established a company called Bluefin Labs, which does Twitter analytics, and it got bought by Twitter. Who would have thought there would be a Twitter analytics company five years ago?
If we can go back to the slide, I don’t think this particular one will happen, but it’s an example of what could happen.
Everything is going up a little slower than I’d hoped. This is the big model we talked about. A product company designs some product, and then they ship that design to a big factory, which happens to have a Chinese flag on it. That factory has thousands of people doing really mind‑dulling jobs. It makes lots of product. The product gets sent to retail, either big bucks retail or, now, virtual retail, and that gets shipped out to consumers across the country.
As we get into more intelligent CAD, which includes manufacturing instructions and manufacturing lines that are quick to set up, not the current manufacturing lines that current robots take a year or more to set up, here is an example of how that model might change. The product companies might develop the product design indirectly to the retailer.
So maybe the retailer is, say, Wal‑Mart. Then Wal‑Mart says, okay, I need 50,000 of these units in Seattle. I need 100,000 in Kansas. Then little companies, little manufacturers, bid on those, bid them locally, don’t have to ship them all over the place, so the supply chain becomes much, much shorter and much more able to be responsive to need. That’s an example of what could happen.
Karabell: Is this an example of advanced robotics toy that you are developing with your company, or will it also be a product of 3D printing?
Brooks: I think it’s all of that. It’s 3D printing, it’s better CAD, and there will also be new sorts of robots where you don’t have to have tremendously concentrated, sophisticated capability in everything. You can put the stuff together yourself, as we have been able to do with PCs in offices for years. Graphics people used Apple computers for years and years because they were easier to use. And that’s really what kept Apple alive during the dark years.
Karabell: I hope we can get better graphics because right now this looks a little like carrots and castles will be proliferating through the country.
Brooks: It’s folksy.
Karabell: It is. So in this model, will you have factories?
Brooks: I don’t know it will be this particular model. It’s an example.
Karabell: You will have small, flexible—will each of these be able to produce multiple types of products?
Brooks: Yes, they will be bidding on stuff all the time. And one of the glamorous things about international travel in the old days was you’d go to different places and you could buy different stuff. Now every store and every airport is the same. This sort of will enable things to be more culturally local, and for people to be proud of what they’re doing.
Karabell: So it’s a re-shoring and a re-localization of manufacturing?
Brooks: I see both as possible, and it’s not just robotics. It’s not just 3D printing. It’s not just new tools on the web. It’s all of those things combined. And stuff that we can’t even imagine just yet.
Karabell: So what are the obstacles to this? I mean, what prevents this—is it purely a technology issue right now?
Brooks: No. I think people are working on the technologies. I mean 3D printing is just going wild. Tech shops. Tech shops have become places where entrepreneurs go and develop their product, and then they use MFG.com. So I think we are starting to see this have an impact.
And then we are starting to see some big players. I don’t know if people noticed last week, Motorola said they will start building a smart phone, and they will build it in Texas, not in Asia. That’s a big step, I think. By the way, we build our robot in the U.S., totally made in the U.S.
Karabell: How do you build your robot?
Brooks: We have almost 20 primary suppliers around the country who build components, and then they go to about four different assemblers, who put together five subassemblies, and then they go to a final place. So it’s shipped all around the country. The reason we do that is we are using particular technologies where there’s localized capabilities.
So for instance, the gears in our gear boxes are powdered metal gears, so making things out of powdered metal where you are pressing something in a mold and center it. It is much cheaper than steel gears, but you need a lot better computation to model what’s happening. That technology came out of Penn State, so all the powdered metal companies that can do this are in Central Pennsylvania. So we go to Central Pennsylvania to make our gears rather than make them somewhere where the expertise is not localized.
Karabell: And the other thing about this model, would this model then work in the way in which you had industrial development in the 19th Century, you had factories which were actually embedded in the urban area. Right? The development in the 20th century was for factories to go outside of the area because they simply—they were too large and demanded too much resources and pollutants for them to be easily located in a downtown. Could you see this being something where you would start to have relatively small manufacturing floors that could be placed within urban environments.
Brooks: Yeah. I think that’s a strong likelihood. Most of the companies that contract on MFG.com are very strong and in urban environments, the neighborhood machine shop that you still see.
Karabell: Where do you do your production? I am just curious—of the actual robots?
Brooks: The final assembly? It’s an hour out of Boston. But the rest, it’s Minnesota, it’s Washington State, it’s Michigan, it’s Chicago, Illinois, it’s Pennsylvania, it’s all over the country, different subpieces are built.
Karabell: So your view would be, the more these things become embedded in the next generation of manufacturing, that creates enough fluorescence of its own, not one that you can easily chart, you can’t know the outcome of these; but it creates its own next generation of multifaceted creative manufacturing.
Brooks: Yes, and it’s also a services and new businesses model. If this clicks right, I will just show you some examples. We’ve had plug ‘n play equipment in IT for a long time. We haven’t had plug ‘n’ play equipment in manufacturing.
An analogy. In Twitter Analytics Company, there will be lots and lots of companies doing supply chain optimization.
I call this one delivery flow. I don’t know what it means, but I use this analogy. A few years ago, GE stopped selling aircraft engines and started selling hours of engine service on airplanes. So our whole supply chain model for manufacturing is still about the physical stuff going. I think there’s meta‑levels, but information technology combined with this Super CAD will lead to the sort of—not by direct analogy, but to the Amazons of supply chain. It will be very different from what it is now. I wish I was as smart as Jeff Bezos so I could figure out how to make this happen.
Karabell: Perhaps you should buy the Detroit Free Press. Then you could be kind of like a mini version of Jeff Bezos.
Anyone around the mics? We can have a few questions. Anyone? This is just too fascinating at the end of a long morning? Okay, we will take a question, and then we will segue into our next—sir?
Humphrey: Thanks for being here. Andrew Humphrey, WDIB in Detroit.
With regard to—actually, two questions. I promise to be brief. In robotics what has been the reaction so far from developing governments? Because in the 60 Minutes piece on Baxter and your company, it was said it might be direct competition with the low‑wage jobs that are in the developing world. So what’s the initial reaction there?
The second question with regard to automation. How do you still maintain accountability when it comes to automation, especially when you are talking about automated warfare and workplace accidents?
Brooks: Okay, first question. The reason I started the company was I was doing a lot of manufacturing in China and saw that the end of that low‑cost labor was coming, as year after year we found it harder and harder to get more workers to come, more ships to coming in, and we see that dynamic playing out in Fox Com now. Thankfully, we see wages coming up in China, people are not as desperate for the jobs and other things, so they are pushing back on their working conditions, and that will continue to happen.
You go down to Vietnam, which is where a lot of stuff is happening now. But where do you go after that? Really, Thailand last year was the fifth largest market for industrial robots because it is already becoming industrialized. So I see that low‑cost labor is finally disappearing, and it’s just not going to be there anymore.
For accountability, that’s a real issue. I—on the military side I was involved in military robots at iRobot but they were bomb disposal robots, they were not weaponized in any sense. But I will point out one thing that some people miss; that is, a robot can afford to shoot second. A 19‑year‑old kid that we put in uniform to and we send him to some place, they are going to be very worried and they are going to shoot first in many circumstances. A robot can sit there and take it and doesn’t have to shoot back. So some people will worry that putting robots into military situations will speed up things. I say it is also possible that it slows things down because it doesn’t matter if a robot takes a few hits. It does matter if one of our soldiers takes hits.
Accidents, motivation. I think that is no different from anything else. We regulate how trains are made. We regulate how buses are made. We have regulations. I think that’s important, and we will continue to do so. And that’s got to be part of the process. So I don’t see any conflict with our standard practices there.
Karabell: I think we’re at the end of our time, and I think that was a really good way of looking at the potential hopeful re‑creation of industrial society in Industrialization 3.0 via all these technologies leading to this world of creativity that has certainly not been the case for the past 20 or 30 years. So thank you.
Brooks: Thanks very much.