Australian-born inventor Rodney Brooks shook up the world of robotics with his insect-inspired concepts in the 1980s, and remained a Professor at Massachusetts Institute of Technology until 2010.
As an inventor, he co-founded iRobots, commercialising MIT research in artificial intelligence. iRobot has achieved considerable commercial success, and is perhaps best-known for its Roomba vacuums, which have sold in millions and millions worldwide.
His current business, Rethink Robotics (which he is founder, chairman and chief technology officer of) is one of the leaders in the booming collaborative robot industry.
In this Q & A, Professor Emeritus Brooks shares his wisdom on topics including getting research out of universities and into the marketplace, why a country shouldn’t ignore the value of scientific research, and why dexterity is still a massive challenge to roboticists. (Transcript below.)
When you launched Baxter, you noted that the IT revolution hadn’t reached factory workers in the way it had for office workers. Have we come any further since then?
I think we’ve seen a big chance in the landscape since we launched Baxter. Baxter has a graphical user interface, it’s easy for people who are not robot experts to interact with it. And the idea of a robot that you can be close to as its operating – a collaborative robot that you can re-task – now you can look at all the major robotics manufacturers, industrial robotics manufacturers and they’re all talking about their own collaborative offerings. What they mean by collaboration varies a little bit from place to place, but I think this was just something that wasn’t talked about just three years ago. So the conversation has certainly changed now, and it’s still early days, but people are seeing that there’s some value there.
You did mention collaborative robots and I have noticed more and more companies offering what are marketed as collaborative robots – probably it means different things to different people – most recently with YuMi last week for ABB. I was wondering if you could make a comment on [what you’ve] suggested elsewhere might be the “knee in the curve” for collaborative robot adoption now. What’s driving that and – I know you don’t like making predictions, but – where, if you’re able to say, do you see things heading?
Well what we’re seeing now I think is manufacturing across the world is suffering from a lack of labour. It’s true even in China. You talk to a Chinese manufacturer and they’ll say their biggest problem is recruitment and retention for their labour. People in China expect a better standard of living, wages have gone up there, wages in China are actually higher than in Mexico now, but even in Mexico, people are finding it hard to get the manufacturing labour they need. So I think people are starting to see these collaborative robots, at a lower cost point than traditional industrial robots, are easier to integrate, because there’s they’re safe to be around: you don’t need the fences around them, you don’t need a partition in the workspace to say ‘this is where the humans are, this is where the robots are’. They can be inter-mixed. I think people are starting to see that it makes sense to put robots in to do the really dull, repetitive jobs that the robots can do. And have people who are much smarter than any of the robots – and much more dexterous – to do the more dexterous tasks. And that’s a way to increase productivity in a world where they just can’t get enough labour.
You’re watching China’s efforts to make more and more robots and get them into factories. Do you have any comment to add, further than what you just said, about that? Is there anything to be learned, further than your recent comments, and what you’ve [just] said about what China is doing?
Well I think – one of the reasons I started Rethink Robotics, it was [then] called Heartland Robotics, was that I was previously with a company called iRobot, which manufactures the Roomba, and we were manufacturing the Roomba in China. So I spent a lot of time in China over the years from about 1997. And by about 2005 I was seeing signs that the [seemingly] infinite supply of labour in China was not going to last.
And that’s what got me thinking that eventually the only advantage to manufacturing in China will be that they have the supply chain in place for high-volume things like iPhones etcetera. But it will be possible to do more manufacturing in more locations of the world where you don’t need that really specialised supply chain.
So, fortunately, I guess I was right. Because we see that happening all the time in China right now.
I saw you recently give an example on the dexterity that’s possible with robots, where you said you – to illustrate your point – you said you wouldn’t take $25 million from a venture capitalist to try and make a robot that would take keys out of your pocket. Why is dexterity still such a challenge in robotics and what does this mean for factory robots?
We do not have dexterous robots at all. The Wall Street Journal mistakenly said last week that the ABB YuMi was dexterous enough to thread a needle. No, it’s accurate enough to thread a needle, but not dexterous enough to thread a needle. You need to set up all the initial conditions just right and then it can retain the accuracy. Dexterity is something that we have not had because to make progress on dexterity you need to do four things at once. You need new mechanisms, you need new integrated sensors, you need new algorithms, and you need new senses of touch. And you have to work on all four of them at once. One of the things we did with Baxter was – because Baxter has got that safe arm – we started distributing that as a research robot. And a lot of universities in Australia here have bought them. And one of the hopes was that this will provide an easy platform for academics to be able to make progress on those four problems at once without having to worry about having a safe robot arm on top of that. And so I’m hopeful that now people will start to explore dexterity more than they have in the last 40 years. And we’ll see if progress is made. But you can’t just order up progress on these things. It may take a few years yet before we have much dexterity.
You mentioned Baxter being an easy platform on which to build and on which to address certain problems. Has there been any evidence about where a – for want of a better term – ‘killer app’ for Baxter might come from and [are there] any industrial applications of note that you might like top mention?
Okay. Well, we certainly don’t use that word. [Laughs]
It’s a bit naff, isn’t it?
Well there’s two things. In the research area, people are working on better touch sensors, better dexterity, they’re starting to work on elder care applications, interacting with people. But that’s not Baxter per se – we’re not going to be selling lots of Baxters to do that. That’s the research that will enable that pattern in five, ten, 20 years. Baxter itself in factories – we’ve seen a number of interesting applications. One I did not foresee at all in the US. It was in third-party logistics. These are companies that take some sort or product, unpack it, re-pack it, and move it off to the shelves, say, of a pharmacy chain where you get two bottles of pills for the price of one this week and they’re wrapped together. Or two shaving kits for one. All those things change from week to week and have been done largely by undocumented labour in the US, up until recently. As there’s been a crackdown on undocumented labour, those companies have had to look for other ways to do this and traditional automation doesn’t work, because the task changes every two or three days. And they’re starting to use Baxters for those tasks. I was not aware of that whole industry, but it’s an enormous industry, it turns out. We’re also seeing in the US a lot of small plastics companies doing moulded plastics. Taking the pieces that come out of the moulding machines and packaging them. A very simple task. It doesn’t require much dexterity but it varies from day to day what the plastic widgets they’re building are. So traditional automation doesn’t do so well whereas Baxter can adapt to the different shapes and sizes.
I know that Vanguard Plastics is an old and an enduring case study for Rethink, and it’s probably a great example of where your robot slots right in.
Yeah, but it turns out there are lots of Vanguards all over the US.
This question is slightly different and perhaps one you might not feel 100 per cent comfortable answering, but you’re a corresponding member of the Australian Academy of Science, and I was wondering if you might have a comment on our country’s attitude towards science at the moment, and if there’s anything we’re doing well in science policy and anything we’re doing stupidly?
Well I think it’s always a mistake to underestimate the importance in the long-term of basic science research and basic science education. That is what has changed the economy again and again in different parts of the world and certainly in the US and certainly in Australia. Never underestimate the importance. And short-term money-saving leads to long-term problems. I was part of a study of the National Academy of Engineering in the US last year, I was a member of the study for the America Competes act, that looked at the impact of research dollars in the US on the economy. And we found it to be enormous and enduring, and I see no reason to think that it would be different in Australia.
Another research question, but of a different kind. You’ve obviously received a lot of attention for your bug-inspired robots, I guess some people might call them, such as the Genghis. Although this is obviously a long time ago, could you say anything about the relationship between what was achieved with subsumption robots and if this has influenced factory robots in any way?
The work I did in the 80s on what I called subsumption architecture then led directly to the iRobot Roomba. And there are 14 million of them deployed worldwide. And it was also used in the iRobot PackBot, and there were 4,500 PackBots in Iraq and Afghanistan remediating roadside bombs – and by the way, there are some PackBots and Warriors from iRobot inside Fukushima now, using the subsumption architecture. There’s a variation of subsumption inside Baxter – we call it behaviour-based now, but its a descendent of that subsumption architecture. And that’s what lets Baxter be aware of different things in parallel, for instance it’s picked something up, it’s put it in a box, something goes wrong, and it drops the object, sadly. A traditional robot would just continue and sort of mime putting the thing in the box, but Baxter is aware of that, changes behaviour – that’s using the behaviour-based approach, which is a variation on subsumption. So it is part of Baxter’s intelligence.
One of the challenges to industry in australia that’s been identified by successive governments and addressed with various degrees of success has been encouraging closer industry and academia collaboration. You’ve probably got more experience than most on this subject; what advice can you offer?
It has been very successful at both Stanford and MIT, where I’ve held faculty positions. Not exactly solving problems directly for industry, but being aware of industry’s problems and trying to do basic research which is going to have some impact on that long-term. And then both Stanford and MIT are not afraid to let faculty and students go off and start companies around those ideas. And those companies often get bought by the older, bigger, more established companies and that’s the way technology transfer happens. So in the US economy, it’s certainly been very successful. It does rely on an ecosystem of funding, of legal resources, accounting resources etcetera. And so it takes a while to adapt that in other places. Certainly when I was on the advisory panel for NICTA, which is headquartered in Sydney, I saw a lot of efforts – both at NICTA and in the community around NICTA in Sydney at least – for start-ups, for helping people spin-out companies. So there’s an ecosystem building. I stepped off that board about two or three years ago, so I’m not completely up to date, but I saw it building when I was on that board.
Do we have the depth of skill or the clustering in Australia to emulate some of the examples overseas?
Lots of places in the US are also trying to build up their clusters. And they’ve succeeded. Silicon Valley is clearly the biggest, and Boston is a big one. New York City is doing well, Austin, Texas is doing well, the Seattle area – they have Microsoft and Amazon there, by the way – and the San Diego area. So I think it is possible to build it up. It takes time, it takes patience, and it takes encouragement.