“Robots must be easy to use”
Rodney Brooks is one of the world’s most renowned roboticists. He was Head of the «MIT Computer Science and Artificial Intelligence Laboratory» until 2007 and founded the two American companies iRobot and Heartland Robotics. He was recently invited to ETH Zurich for the first symposium of the National Centre of Competence in Research Robotics. Among other things, he spoke to ETH Life about why robots still have trouble recognising the difference between a heap of sugar and dirty laundry.
Mr Brooks, how many robots are in use at your house?
I’ve got two: the vacuum cleaner
robot Roomba and Scooba 230, a little robot that mops the bathroom – both of
which were developed at iRobot.
You’re best known for the behaviour-based artificial intelligence you publicised
in the early 1980s, according to which robots, to put it simply, act
instinctively, like insects, directly according to their sensor data, without
an elaborate map. Are you satisfied with how this concept is being implemented
in robots today?
Behaviour-based robotics is on
the up. It’s used in the robots Roomba and Scooba we developed, for instance,
and in self-driving car or Mars robots. But the technology has far from been
exhausted, so a couple of years ago I thought to myself: either I stay at MIT
and become a grumpy old professor who gets annoyed because his ideas aren’t
being capitalised or I start a company and see them through.
You opted for industry and thus experience in both fields – research and
industry. Where do you see the biggest fields of application, the biggest
market opportunities for robots?
Most
certainly in a society where people live longer. That’s where there’ll be a
great demand, such as in China, for instance, on account of the one-child
policy. But how you’re supposed to translate that demand into products still
remains to be seen. We’ve got the technology but what are we supposed to build
with it that really helps? I believe there could ultimately be many little
devices one day – that can carry the shopping upstairs, for example. After all,
people want to live longer in the environment they’re used to.
The robot as friend and helper?
Contrary to what many Japanese
companies are predicting, I don’t believe robots will become our friends. It’s
about enabling people to preserve their dignity and independence for longer. And
robots can help them to do so.
Where else can robots help us?
In agriculture, for instance. Robot
systems in tractors control the amount of fertiliser or seed to be scattered
based on the moisture measured on the fields – and the tractors themselves are
even robot-controlled. And another major market, of course, is the military, where
robots help diffuse bombs and sweep for mines. And they’re
used in disasters like in Fukushima. Safety systems in cars is another
area that’s becoming increasingly important.
Besides price, what are the main criteria for being able to commercialise
robots?
Ease of
use can’t be rated highly enough. I don’t want to waste any time unnecessarily trying
to find out how something works. A robot should be easy to operate; otherwise
no one will buy or use it.
That’s where the developers come in. What are the biggest technological
challenges in building robots?
Perception
is still a problem. If a robot is supposed to clean a house, it needs to
recognise the difference between a pile of paper, a heap of sugar and laundry,
and treat these things differently; it shouldn’t put the sugar in the washing
machine. And we haven’t made a vast amount of progress in gripping and manipulating
objects, either.
So what we need is another visionary, groundbreaking idea like yours on
artificial intelligence, then?
Certainly not from me [laughs];
it’s up to my younger colleagues now. No, in all seriousness: you try things
out, put them on the back burner for ten or twenty years, then try them again.
Navigating robots was already attempted back in the 1960s, for instance, but it
didn’t work very well. But in the 1990s it came off and the problem with
navigation has pretty much been solved. You have to try again every twenty or
thirty years; the second or third time round you end up hitting the jackpot.
Is a programme like the National Centre of Competence in Research
Robotics in Switzerland a promising approach to succeed more quickly perhaps?
An interdisciplinary
collaboration like the one between the four institutions at the National Centre
of Competence in Research (see box) is very important. The exchange of ideas it
entails can be especially fruitful.
In which areas of our lives will robots have the biggest impact in
twenty years?
Devices
that perceive, calculate and act in the world are bound to be extremely widespread.
But it’s difficult to predict in what form as yet.
What are your dreams and goals when you think of the future of robotics?
You
could say I’ve already lived my dream – from a time when there wasn’t a single
robot in the world to the millions of robots we have today. Then again, a key
aspect is relieving people of mundane or dangerous jobs. Yes, increasingly taking
the burden off people through robotics is definitely something that would make
me happy.
National Centre of Competence in Research Robotics
The National Centre of Competence in Research Robotics – “Intelligent Robots for Improving the Quality of Daily Life” was launched by the Swiss National Science Foundation at the end of 2010. Under the programme, scientists are looking to develop robot technology that is geared towards people over the next twelve years. The programme unites leading robotics experts from ETH Zurich, the University of Zurich, EPF Lausanne and the Dalle Molle Institute for Artificial Intelligence based in Lugano. It is coordinated by EPF Lausanne.
READER COMMENTS