2013 QUT Grand Challenge Lecture – Intelligent Machines and the Future of Work – P Corke

2013 QUT Grand Challenge Lecture – Intelligent Machines and the Future of Work – P Corke


Thanks very much Ian, and thanks everybody
for coming on this Friday afternoon. Hopefully it’s a fun talk. Learn some things about a
whole bunch of topics, robots, and my view about how they fit into society, why we need
robots and why they’re useful. I titled this talk the, about “Intelligent Machines and
the Future of Industry”, but I actually think that that was the wrong title. I think the
title really should be this, I think it should be about “The Future of Work”. Because the
industries are not going to change on a timescale of ten years. In ten years Australia’s going
to be dominated by the same industries that it’s had for the last fifty years. It’s going
to be agriculture and mining and whatever. But the way we work in those industries is
going to change very rapidly over the next decade so I took the liberty of re-titling
the talk. Ok, a little bit about me. I was a child in
the 1960s. It’s a wonderful time for technology development and when I was a kid I wanted
to be an astronaut. Who wouldn’t? My parents gave me this book, and I’m not sure how old
I was when I got this book, but I was somewhere between four and nine, and this image of the
MOBOT is sort of seared into my brain. This robot is strong enough to bend iron bars,
but gentle enough to handle laboratory glassware. And so I didn’t become an astronaut. I became
a professor of robotics instead. In the 1960s images of robots were everywhere and sometimes
you’d get robots and astronauts in the same thing and that was pretty cool. And what’s
interesting about these robots, the robots around the 1960s, is that they were enormously
capable and they were benign, you know, they meant only the best for people. One of them
was a little bit too literal, in the way he understood language, but generally they were
helpful. They were forces for good. These are probably the most famous robots that exist.
Two famous robots from Hollywood, also in this benign and capable mold. So C3PO could
talk his way out of almost any situation. R2D2 is a very handy guy to have when your
X-wing Star fighter is broken, you’ve got to hack a Death Star or something like that.
But there have been quite recent movies. I’m not sure, how many people in the room have
seen this movie? This gorgeous movie, and it doesn’t have spacemen or aliens or X-wing
star fighters. It’s about a gentleman, an elderly gentleman, an ex-burglar and a domestic
robot and it’s about the interaction between the robot and the aging burglar and it’s a
lovely, it’s a lovely tale. But there have been lots of movies about robots,
even back from the 1920s, believe it or not, through the 50s, and then right through to
quite recent times, and there’s a whole bunch more that I’m sure I’ve missed. I’ve probably
seen most of these. The word “robot” is actually a Czech word. It was coined in a Czech play
in 1920, (possibly coined by his brother, there’s a bit of a debate about that), in
a play called Rossum’s Universal Robots. And the plot of this play is like the plot of
a lot of robot stories. Some people develop robots to relieve themselves from the burden
of work. The robots rebel and they kill all of the humans. It’s a bit like the Frankenstein
story, which was written about one hundred years before this and it was quite a common
recurring theme in robot stories. I was in Prague a few years ago and had the opportunity
to make a pilgrimage to his grave, and some kind person had left a little robot on his
headstone, which I thought was a very, very nice touch. So it’s a Czech word that means
“serf” or “worker”, something like that. The person who probably made the biggest impact
in getting the notion of robots into popular culture was Isaac Asimov. So he’s dead now,
but from starting around the mid 1940s he wrote literally hundreds and hundreds of stories
about robots and the common theme through all of his stories is these three laws of
robotics which was embedded in the positronic brains of the robots, so that they could do
no ill. And they were actually not a bad set of laws, but they’re not perfect, because
of its incompatibilities with the law’s strange loopholes and wrinkles that allowed him to
write all those hundreds and hundreds of stories. So Asimov is probably single-handedly responsible
for getting robots into the public consciousness. But it’s all a big con. So, in this movie
from 1956, the robot is a man in a suit. But you think “well ok, 1956, pretty rubbish technology,
what would you expect?” Sorry to pop your bubble, but that CP3O: man in a gold suit.
R2D2 is a gentleman inside the robot. And even the lovely “Robot and Frank” movie that
I just mentioned is a rather short lady called Rachel, inside a plastic casing. She said
it was filmed one summer and it was really, really hot inside the robot. She didn’t enjoy
it. So we’ve been brought up with these images of robots that are very capable, mostly they’re
benign, and they’re not eliminating humanity, they’re helping us. And as a roboticist these
images are enormously unhelpful, because that’s the bar that we are held to. And so you build
anything and it’s not as useful as R2D2 or CP3O, people look and they say “that’s rubbish”.
And so it’s what we aspire to. It’s the beacon on the hill, and one day we will get there,
but it’s a little ways off yet. So the type of robot technology that’s very, very mature
is this kind of robot. Call it the industrial robot and they’ve been around for more than
fifty years now. And many products that you own have been touched by robots somewhere
along their path. Robots have been involved in manufacturing it. If it’s a car robots
have been involved in welding and painting. If it’s a computer, it’s perhaps been the
circuit boards that perhaps have been assembled by a robot. It may have been put in a box
by a robot, and that little box been put into a bigger box by a robot. Many things today
have been touched by robots. And there’s a lot of them on the planet. There’s about a
million of them, robots of this particular type at work today. We wouldn’t have a car
industry in a high wage country like Australia today if we didn’t have robots like this doing
a large amount of the mechanical work, physical work. That technology came from a little high-tech
startup in 1956 in Connecticut called Unimation Inc. Two people: the top one, George Devol,
who died last year aged 99, was the inventor. Hundreds and hundreds of patents to his name,
and mostly in archaic things like gramophone technology. And then he, this wonderful insight
about how to automate the unloading of machine tools. So, this robot here is unloading a
die-casting machine. The guy at the bottom, Joe Engelberger, is still going, and he was
the entrepreneur. He was tireless, going around the world prophesying about robots. So, he
was the one who convinced General Motors to buy the first one, convinced lots of other
people to buy more, convinced the Japanese that they should set up a robot industry,
which eventually eclipsed the American robot industry, which is a little ironic. Engelberger
had robots in the Jonny Carson show playing drums and opening bottles of beer. He was
the ultimate showman. And the pair of them had the right mix of skills to do this, so
in their day they were the “Steve Wozniak” and “Steve Jobs” of robotics. So this is a
typical sort of industrial robot, first generation robot and it’s doing a pretty simple thing.
It’s lifting boxes off a conveyer belt and putting them in another box. So I can think
of a robot as being a pretty simple machine that moves something from A to B, but it can
do it very precisely, it can do it forever, doesn’t get tired, we can work with the lights
off, it doesn’t care. It’s a very simple thing that moves something from A to B. But there
are lots of jobs in the world that can be done by picking up something and moving it
from one place to another. This class of robot excels at that kind of work. This is another
class of robot. It’s a robot vacuum cleaner. Who’s got a robot vacuum cleaner? I used to
have one. Ok, not many. There are 6 million robot vacuum cleaners that have been sold.
It’s fantastic, it’s awesome. It’s six times more than the industrial robots, which is
a technology which is fifty something years old. These things in less than ten years have
sold six million. So these are the harbingers of a new type of robot we call “service robot”.
So that’s a robot that delivers a service to an organization or to an individual. And
it’s a very simple robot. It needs to be in order to make things to a cost. And this is
a time-lapse photo of what a Roomba robot does when its vacuuming your room. On Flickr
there’s a lot of pictures of people taking time-lapse photos of Roomba robots, and you
can see that it’s basically random: it goes in a straight line until it hits something
and it bounces off. If it senses there’s enough space it will do a big and excited spiral
until it hits something and then goes back to doing lines. So that’s how they work. Almost
no memory. It doesn’t know what a room is. It doesn’t know where the furniture is. It’s
got no map. Almost like somebody with no eyes and no brain. Now that’s maybe not a very
good analogy. So let’s look at the world robot population. The latest numbers I could find
were for 2008, and I’m frustrating Tim by pacing. The industrial robots that I mentioned
first, the arm-type robots that are building cars. There’re about 1.3 million of them.
They’re not, 1.3 million robots rather, they’re not growing particularly quickly, but this
“service robot” sector is really taking off. Most of that’s vacuum cleaners, but there’d
be surgical robots in there and other sorts of robots that I’ll show you shortly. So the
market is really accelerating. It’s not in the industry applications, it’s in these service
applications. And they’re more challenging because a robot has to work, perform its service
in an environment where human beings work, and most of our environments are very chaotic
and changing, and this is the big research challenge for roboticists today, is how can
we make robots perform in environments that we take for granted and can work in effortlessly.
That’s the challenge. Who’s ever bought anything from Amazon? Ok, so you go to the web and
you buy stuff and a little while later you get a box back, and in the box is all the
items that you ordered. How did the things get in the box? Yeah, it’s hard. Because look
at how much product Amazon’s got. How big is their warehouse? It must be as big as Brisbane,
with all the products that they’ve got. So you order ten different things. Some guy’s
got to pick up a box and walk through the warehouse and put the items one by one into
the box. How long is it going to take you? Like an hour? A day? Is he going to ride a
bicycle? To think about it, it just doesn’t work. So what Amazon have done is they’ve
got a really amazing robotic technology. And all the shelves in the Amazon warehouse are
moving around on robots. So the guy who is doing the dispatching, the guy whose job it
is to fill your box, stands still, and one by one, the shelves that contain the items
come past him in just the right order. And so he takes the first book out and he puts
it in the box, the next shelf comes, and he takes the next book out and puts it in the
box, or toaster, or garment or whatever it is you’ve bought, puts it in the box one by
one. So their warehouse is just shelves in motion. And they can be packed in really tightly
because no people will need to walk in between them. Robots come in from underneath and pick
them up and move them out. So this is a brilliant technology from a company called Kiva Systems.
It was a little startup and Amazon recently bought them out for a huge amount of money.
So this is a way of completely rethinking the way you do dispatching, is what allows
Amazon to do what it does. So that’s an example of the service robot. Here’s another kind
of robot, of a similar vein, but much, much bigger, and these are operating at the port
of Brisbane. So when you fly into Brisbane, and there’s the port, across the water from
the runway. One of the ports is completely automated and these machines, called straddle
carriers, pick the containers up from where they’ve been dropped on the key side, by key
crane, and put it into a large stack. Now this was an Australian-developed technology.
Came from out of the University of Sydney, commercialized by a Finnish company, Patrick
at the time, who had the vision to think that it was a new way to do container handling,
and also a bit of a stick for the unions as well, but it’s a technology that’s very, very
effective. But all it’s doing is moving very, very big things, 20 foot, 40 foot containers,
from A to B. Robots come in many different sizes and shapes, so they can fly, they can
swim underwater, they can have legs and they can repel down to the insides of a volcano
to sample gases and liquids where you wouldn’t want to send a human being. So robots have
many, many different forms. It’s an important thing to remember. A big application, a big
motivation for robots in non-industrial or non-domestic applications is the class of
industries that have traditionally been referred to as the dirty, dull and dangerous. So agriculture
and mining. And these are dangerous. They have a large number of fatalities per 100,000
workers. People go to work in these occupations and are potentially injured or killed. Doesn’t
happen to us in our kind of jobs, but there are large numbers of jobs that employ lots
of people where that is a possibility. So robots have been postulated as being perhaps
a solution to this. So this is some statistics from the United States. And the two lines
here, not the bar graph down the bottom, the lines there are the fatality rates, so the
number of people killed per million tonnes of mining product produced. And we can see
that back in the early 1900s the rate, the fatality rate, was really high and the industry
learnt. It learnt about how to make itself much, much safer. It’s a learning curve really,
and it has matured, but the problem is it’s plateaued. It’s not plateaued at zero. We’ve
got the accident rate down to a level and we don’t know what to do next. You can train
people all you like, but you’re not going to reduce that last number of accidents. And
robots are being talked about as being the solution to this. So, how do robots fit into
mining? Well you can consider something like dragline excavators. I spent too much of my
life on dragline excavators. They basically move stuff from A to B, one hundred tonnes
at a time, or a wall truck moves stuff a few hundred tonnes at a time, from point A to
point B. So these machines drive material from one place to another. Lots of accidents
happen with these big trucks, running over little trucks. It seems surprising, but it
happens, a lot. And in underground mining you do the same thing. You’re moving material
from the stope to the skip, using a machine like this you see down in the bottom corner.
And at CSIRO myself and colleagues, we developed a guidance system for this class of vehicle,
which was ultimately licensed to Caterpillar and you can buy it and it’s at work at many
mines, all round the world. It’s the simple job of driving from one place to another,
with nobody on board. No talk about robots is complete without talking about human-like
robots, and everyone’s seen human-like robots. They dance, they sing, they kick balls, they
struggle to play something that looks like soccer. Why do we have human-like robots?
Well there’s always been this human fascination with things that look like us and way-back
in the 1800s, people built mechanisms that looked like humans. They were quite intricate
clockwork machines with gears and cogs and cans inside and it was fashionable in the
high end of European society back then. This HONDA ASIMO robot, probably one of the most
sophisticated humanoid robots around at the moment. It’s not very tall, and the reason
for that is you don’t want to make it feel too threatening. If it was six foot tall you’d
feel intimidated by this thing, so they make them childlike, and it can do things like
running, and it can do very small jumping, which is a new skill for ASIMO. There you
go. (audience laughter). Audience member: Toyota?
Peter: (laughter) No, wrong company. And it has some sort of manual dexterity. It’s got
some fingers and it can do manipulation work like taking the top off a bottle, and if we
wait long enough it will pour it. … Maybe we won’t wait that long. Ok, why would you
build a robot that looks like a human being? It’s an interesting question. I wondered this
when I first started hearing about humanoid robots. What is the use of them? So consider
the case that you want to build a robot car, and I’m going to come back to the robot car
theme a number of times. So this is a robot car, built in Carnegie Mellon University.
It won a very famous robot car competition, a thing called the DARPA Urban Challenge.
And you can see what they’ve done to it. The back of it’s all full of computers and it’s
bristling with sensor technology. So this is where you take a car, and then you turn
it into a robot by adding a whole bunch of stuff to it. That’s one way to build a robot
car. The other way, and this hasn’t been done yet, is that you can put your humanoid robot
into the car and it drives it. So if you build a machine, that, in terms of its structure,
its shape, is like a human being, it can operate all the technology that we’ve already built.
So it could hop in the seat of your car and drive your car. It could go home, it could
open your refrigerator, it could make a meal using your stove and put the dishes in the
dish washer, because we’ve built a whole world of infrastructure and tools designed for things
about this shape. You know, 1 1/2 – 2 metres tall, two arms, two legs. So if we can automate
that then by default we get automation of everything that we do now, whereas the hard
yards is to take every single machine that we’ve ever built and rebuild it, as a robot.
That’s a very high-cost path to go. So it’s a philosophical divide here about how we do
automation. This is an interesting one out of Denmark. So one of them’s a robot. The
guy on the left’s the robot, and the other one’s a mimic. He looks a little bit mad,
but, it’s a bit spooky, and realistic in some regards. And this raises a really interesting
issue. There’s a phenomena that’s being talked about lately which is called the “uncanny
valley”, and along with the horizontal axis is “how lifelike is the machine?”, and then
in the vertical axis we have “how attractive we find it?” or “how repellant we find it?”.
And if it just looks like a, sort of, a cartoon robot, that’s not very much like a human being,
we don’t mind it, but as it becomes more and more human-like it just starts to look weird,
it looks dead, it looks like a corpse, a zombie. It’s the living dead. It’s eyes don’t move
quite right and quite suddenly we find it repulsive and as the fidelity goes up again,
and keeps increasing, we then find that it’s satisfaction to us and we can live with it.
So it’s an interesting phenomenon. You either make a robot that looks nothing like a human
being or you make something that has to be an exact copy of a human being. And so you
have images like this which, to my mind, I still find slightly creepy, and repellant,
even as a roboticist. This unsettles me, and part of the reason it unsettles me and perhaps
others in the audience is the cultural baggage. So in the West we grow up with culture, cultural
factors, like Frankenstein. Right, so we’ve heard the story of Frankenstein. We have stories
of zombies and Dracula. And there’s Jewish legends of the Golem, which is a creature
made of clay that has been brought to life by a Rabbi, and one of the famous Golem legends
comes from Prague actually, which is where the “robot” word comes from. And these things
sometimes run amuck and cause problems, and so we’ve got some very bad stereotypes in
our culture. Whereas in Japan they have quite different cultural baggage. So they have Shintoism,
which sees an animate force in everything. There’s an animate force in us, in a rock,
in a mountain, so the machines are not so dissimilar to us. And they also have this
icon: mighty atom. The cartoon character from about 1952. We call him Astro Boy, nuclear
powered robot. And in war-torn Japan in the early 50s he was quite inspirational to a
lot of people. He only did good, but he was a robot and he was nuclear, and this happened
in Japan. And if you get some elderly Japanese researcher-professors over a few beers, they’ll
start to tell you about that they read the mighty atom comics when they were young. And
people now who are senior in Japanese government had the same up-bringing. So Japan as a country,
as a culture, are quite comfortable with the idea of humanoid robots. It’s progressed there
much more than anywhere else in the world and this is my speculation as to why that
might be. Ok, I want to talk briefly about the context in which robots fit into society.
What problems do robots solve? And to do that I’ve just used three graphs which are kind
of emblematic. One’s about population growth. One’s what we call the dependency ratio, and
I’ll explain all these one by one. And the other is, this is the carbon dioxide concentration
in the atmosphere: it’s about climate change and global warming. So world population, by
any model, is going to keep increasing. That means we’re going to need more resources,
more food, more transportation. This number get bandied around a lot: 70% more food production
by 2050. That’s a big lift in food production. Because there are some problems. We need food
not just for people, we need food for animals. Animals are a very inefficient way of converting
vegetable protein into another form of protein. Land is being claimed, alternatively used
by housing, so farmland’s making way for housing. Farmers are getting older. The productivity
on farms in Australia, (this is from a recent PMSEIC report), is decreasing. So how are
we going to make 70% more by 2050? So here’s a picture which is kind of interesting to
think about for a little while. So here we have a machine which is spewing chemical,
weed it is spraying, on a farm. And this machine is really massive. The reason the machine
is really massive is because the guy’s got to drive it, and the way we can make the person
most effective, how we can cover the most number of hectares per day, is to make the
machine he’s driving much, much bigger. But there’s problems when the machine gets bigger.
“A” is the spray is not applied nearly as well, because it’s big, the spray booms are
probably higher above the crop than they should be, it’s going fast and a lot of the spray
ends up getting lost, going into the environment, which you don’t want. Not landing on the plant
is pretty expensive. The machines are massively heavy, so they crush the soil, they damage
the soil structure and so they then put GPS guidance on them so they go up and down these
kind of virtual tram tracks and they only crush the same soil over and over to minimize
impact that it has. If it’s rained, the farm’s boggy, you can’t take the machine out because
it will get stuck. Machines are very expensive because they’re very big, and the final thing
is that if the machine breaks, the farmer doesn’t have anything. So there’s a big problem
with this model and the whole model is predicating the fact that there’s a guy driving it. So
if we take an alternative tact, and we have a project at QUT at the moment where we’re
looking at this, is “ok, let’s build small robots that can go and kill weeds.” Doesn’t
need a driver, so that’s got to help. It can be smaller and lighter so it does less soil
damage. It can run 24 by 7, and we could have lots of them. We could put twenty in a paddock,
in a field, if we need them. If one breaks then we’ve still got nineteen left. We can
pick them up and deploy them to another field later. So it’s a way of rethinking the way
we do agriculture. We can also treat weeds individually. There’s less spray lost, because
we just go up to a weed, or to a plant, and say, “are you a good plant or a bad plant”,
and apply the herbicide appropriately. And that’s important because there’s another whole
problem, I’ve learned a bit about this last year, is resistant weeds. So the weeds have
become immune to the most gentle herbicides that we have, like phosphate’s Roundup. Right?
Some of them are immune to that. It will no longer kill them. To kill them now you need
to use nastier chemicals, ones that are much, much more expensive. So when you go up to
a plant you really want to look at the plant, and figure, if it’s not resistant, wack it
with Roundup because it’s cheap and effective. But if its a Roundup-resistant plant, then
you want to whack it with the smallest possible amount of the herbicide that will do the job.
So this a phenominom that is happening around the world, it’s a big and growing problem
in Australia. And so this is a possibility. I lifted this off the web. Of what, a sort
of robot, urban, garden thing. Right? So you have these vegetable pods on the sides of
plants, and great big robots that lift them up and bring them down. So maybe this is the
ultimate way that robot agriculture could end up. The other problem that we have as
our population increases is traffic. Everyone knows about traffic. So here’s some interesting
statistics, from around the world. We kill 1.2 Million people a year, and it’s going
up the list of reasons by which people die. So instead of the 10th most common cause of
death, it’s moving up to the 8th. And in terms of the cost on society it’s about 2% of the
GNP of highly motorized, industrialized, Western countries. It’s a huge cost, just from fatalities.
That’s not including the cost of fuel and time lost, waiting in traffic. So here are
the road fatalities. For a number of countries they have similar sorts of trends, but the
same thing as the mining fatalities that I showed you before. They’re platuing out. They’re
getting less and less, but the rate of decrease is getting less. They are platuing out at
this number that’s not zero, and that’s a problem. The reason that happens is that people
are not very good drivers, basically. Traffic congestion. It’s an interesting graph, because
at the bottom is a number of cars you’ve got, wanting to go along the road, and up on this
axis is the number of cars per minute going along. And the curve goes up and up and up,
so we get more and more cars along the road, and then it tanks. Right? You’re in a traffic
jam, and it all starts to snarl up and slow down, so you’ll all experience this curve,
right? The good bit, and the bad bit. The reason we have this curve is because human
beings are not very good drivers, but interestingly, 93% of people consider themselves better than
average, which anybody should know is not possible. People aren’t that good, but they
think they’re that good, and that’s probably also part of the problem. So if we look at
a traffic jam, we’re in a traffic jam, it’s pretty horrible, it looks something like this.
But if I took a picture out of my window, which looks down on the Riverside Expressway
when it’s snagged, it looks like this. You could fit a lot more cars in there. Right?
So we’ve got big gaps between cars because people are not very good drivers. We’ve been
trained to leave three car lengths in front of us. What if you could put three cars in
that gap? Right, you’d get four times as many cars going along that piece of road. This
is probably the only way that we can improve the capacity of our roads. We can’t build
more roads, it just doesn’t work. So maybe we need to have cars that are smarter. So
most people know that Google is building a car. It’s a Google car, in fact there’s lots
of Google cars in California and Nevada now. So why is Google building a car. Google’s
a software company, and they see cars as a huge market. There’s lots and lots of cars
on the planet. Google can sell Android into a large fraction of people’s mobile phones.
If they can sell the car equivalent of Android into a large fraction of cars on the planet,
it’s good market for them and they’re a very good software company. They’re probably a
better software company than GM, and Ford, and Toyota who’re not known for software,
so that’s why Google is building a car. But the really cool thing is, if you look at that
graph that I showed you before which is here in red. If robots are driving that line just
keeps going up and it doesn’t tank. If it does tank, it tanks much further away. So
talking about an automated convoy, and your car follows the car in front, a metre behind.
It’s paying attention all of the time, and it can maintain that kind of separation distance.
And it’d be really nerve-racking I know, but I think this will happen, and the reason it
will happen is because we can’t build more roads, and people are not very good drivers.
So we may move to a model in the future where you’d have little robot taxi pods, and they
just pick you up. You tell it where you want to go and it takes you. You can read a book,
and it drops you off. You don’t have to park it, you don’t own the car, you just hire a
robot taxi service. It could happen, it’s got a lot of merits. I want to talk briefly
about this thing called dependency ratio. This is the ratio of the number of dependant
people in our society to the number of workers. So a small number is good, a big number is
not so good. And there was a big peak back in the 1950s, no 65 right, where I’ve circled
in red. The reason for that is the Baby Boom. Right? So all these babies came, and this
ratio there is the number of children plus the number of elderly divided by the number
of working-age people. So at that period, there were a lot, a lot of babies were born.
They all moved into the work force, so the numbers moved from the numerator to the denominator,
that ratio got smaller. What’s happening now is that they’re leaving the workforce. They’re
going back to the numerator, and that’s a problem. The problem is that they live long,
and so they’re going to be in the numerator for a long, long time. Fertility rates fallen,
so there are less people moving into the denominator from the top. This is a profound problem we’re
going to face because this dependency ratio is going up, and will cross 100% sometime
in our living future. And it’s at a historic low now. It will never ever be this good again.
And the problem with these people who are older is that older people require more health
care. So if we look at the hospital visits versus age, there is a big bulge at the top,
so this is a big problem for our society. Look at a hospital. It is organized chaos.
There is stuff going everywhere. There is lots of stuff going from point A to point
B. So there are lots of applications for robots in hospitals. Not just surgery, though robot
surgery is a big thing. Robots that can pick up patients, and carry them around. Now that
is literally a back-breaking or back-damaging job for people in hospitals. It could be done
by robots with cute faces, that are a long way to the left of the “uncanny valley” curve
that I showed you before. You know, these medical instruments that can just move around
the hospital all by themselves, instead of having to be wheeled by nurses. So this will
happen. Last thing I want to talk about is environmental change. This is the graph that
Al Gore famously showed, with himself standing in a cherry picker. So think of it as an asset
management problem on a big scale, and for ordinary asset management, managing our built
assets, there are three principles. It’s all about inspection. You need to inspect the
assets as much as possible. Whether it’s sewer line or a power line. Inspecting power lines
with helicopters is dangerous business. Ergon spent a fortune inspecting power lines in
this state with helicopters. This is not necessarily one of theirs, but it is a very dangerous
business. So people have been designing robots to inspect pipelines and high voltage power
lines. Robots that climb, robots that crawl, all manner of robots. But when it comes to
the planet, which is absolutely massive, robots have got a role to play as well. Because the
way we inspect the planet now is incredibly labor intensive. Right, so scientists go out,
grad-students go out, and they sample the planet. There’s not enough of them to sample
it very well, or very often, so it means we don’t really have a very good picture of the
state of our planet. So here is some more pictures of scientists doing sampling. We
don’t need human beings to do the sampling. So here’s a Mars rover. Instead of sending
geologists to Mars, which we could do, we can send a robot to Mars to do the geology
on our behalf. It’s an interesting model. Here are some of the Mars robots, just out
of curiosity. This was the first one we sent. Sojourner. Spirit and Opportunity are these
guys, and this one over here is the current one, Curiosity. It is a monster. Compared
to the first one, it was sent probably 15 years ago or something, but they’re doing
science on our behalf. This is from a colleague of mine at CSIRO, Matt Dunbabin, and what
he did is he built a boat, a robot boat, that surveys Little Marang Dam, and it looked for
methane emissions. Methane bubbles coming up. Now if you just went out in a boat and
took samples at three points, and you say, “Ok, that’s the methane emission from Little
Marang Dam”, but what he showed was the methane emissions have got a diurnal pattern, so at
different times of the day the methane comes up, and it’s not uniformly spread. Right?
So there’s a lot down in this arm of the dam, but not much in other places. If you sampled
the way we do now, men in a boat, very sparse sample, you wouldn’t even pick up this effect.
So what robots can do is continuous. They can do this in Spring and Summer and Winter
and Autumn, over many many years. After flood events and after drought events. We could
really get a good picture of what’s going on. This is the great advantage of robots.
Some work here from QUT, an unmanned aerial vehicle looking for dugongs in Moreton Bay.
Solar powered aeroplane that can just stay up for a long time and look at stuff. The
Argo float network. So there are thousands of floats, all in the oceans, all round the
world, going up and down, measuring water quality, salinity, temperature, radioing it
to satellites and building up brilliant models of the world’s oceans. This is a robot we
have at QUT that can map a block of water. You just let it loose, tell it where to go,
and it will measure all of those things. There’s some other work we’ve been doing at QUT just
by taking images from a camera, on a robot, and this is speeded up, and just from moving
a single camera along a piece of coral, we can build up this nice 3 dimensional representation.
We can add more and more levels of detail, so here you can see the model taking some
kind of form. You can zoom in on a bit of it. So this is now a three-dimensional model
of that piece of coral that I took just by moving a robot take a camera over that piece
of coral. I can measure things. If I did this from year to year, I can see how much each
of those organisms had grown metrically. How many millimeters has this thing grown since
last time I looked at it? So this is the sort of things that robots can do, they can acquire
massive amounts of data and use it for science. I want to talk briefly about robots and ethics.
It’s something I think as a community of roboticists, it’s something we don’t talk about enough.
The first and most obvious one’s warfare. Robots have an increasing role in warfare,
and it’s not just sort of science-fiction robot warriors. In the United States, now
figures are one in fifty war fighters is now a robot. Right? So we have unmanned aerial
vehicles which are firing missiles at people who are considered hostile. We have robots
that can, you know, direct withering fire at a target. At the moment, all of these robots
have got a human in the firing loop. So information goes back to a remote operator who says whether
the robot should fire. There is interest in removing the human from the firing loop and
I think that’s really dangerous and really scary. Robots should not be making those kinds
of decisions. Cars is another one. Big ethical issues I think around cars. So let’s say there
was a robot car, and you could go and buy it, and let’s say that they eliminated 95%
of all road fatalities. It would save almost a million lives a year. But it didn’t eliminate
all of them. There’d be still some 5% of accidents left, but now they’re the robot’s fault. Right?
So you have a technology that saves a million lives a year, there would still be some fatalities
that would occur. Would we as a society adopt that, or would we just say that those robots
are dangerous, sometimes they make mistakes, and therefore we won’t use them? It’s an interesting
question. Let’s say you got a Japanese car with American software in it, and it kills
someone in this country, who do you sue? How do you get the thing registered, or roadworthy?
There are all sorts of issues, legal and ethical issues, that come up just in the context of
robot cars. And I think as a society, we’re going to face this in the next ten years.
I think we will have to. And everyone’s just waiting for someone else to go first. Privacy
is a really big issue. Anybody now can buy a gadget like this for two hundred bucks,
and you can fly it from your iPad, and it beams down everything that it can see from
its camera. You can go buy them from the shop. You don’t even need any particular skill to
fly it. And so there are big privacy issues. There was something in the newspaper on the
weekend, I think, about Queensland Police, and drones and eyes in the sky. The need to
work. I’m not sure if you eliminated all work, and robots did everything, whether we’d be
any happier. That’s a big question, probably not one for engineers. This is an interesting
one though. This is the one that often doesn’t get talked about. Warfare is the very obvious
ethical question, but this one: is it appropriate to have our children raised by robots? If
you can’t get enough childcare workers, is it ok to plonk them down in front of robots,
without any kind of human company or role-modeling or whatever. The same for the elderly. So
in this particular movie, without being too much of a spoiler, this gentleman’s son finds
it’s too much of a drag to drive up from New York and visit his Father, buys him a robot.
Can we just buy out of responsibility of looking after elderly people by buying them a robot?
Interesting question. Ok, nearly done. The title of the talk is “Intelligent machines”,
so I should talk about intelligent machines, and “intelligent” is an interesting word.
It’s not a word that I like in the context of machines. I got a Google Car, again, and
it can drive, like a person drives a car. A person is intelligent, so if the machine
drives as well as a person driving a car, is the car intelligent? I don’t know. The
dictionary definition of intelligence is actually a very, very low bar. It says “the ability
to acquire and apply knowledge and skills”. So the Google Car has knowledge and skills.
It didn’t acquire them itself. The Google engineers acquired the knowledge and put it
into the computer. But you could argue, yeah, it probably is intelligent. But this is an
interesting definition. A bit more utilitarian, a bit like a Turing test. It says “Intelligence
is in the eye of the beholder”. So if you think the car’s intelligent, you can call
it intelligent. But I struggle with term “intelligent” and “artificial intelligence”. I still don’t
know what it means. The future of work, is the other thing I said I was going to talk
about. The traditional model has been that the worker is at the workplace. So if you’re
a driver, you sit in a car. If you’re a machine operator, you sit at the machine. If you’re
a farmer, you’re at the farm. Right? That’s starting to change, so what’s currently possible
now is the worker can be remote from the machine. They can remotely control it, and in some
examples : a deep sea vehicle, exploring the wreck of the Titanic or something. Can’t put
a human being there, it’s too expensive. Human beings get crushed easily, so we have a machine
doing the work. The worker is actually remote, sitting on the boat. Surgery. The surgeon
is sitting over here, operating some levers and looking at a screen. The patient is over
here being worked on by the robot. All that’s between them is wires. So we can actually
separate them and this has been done. Operations have been conducted across continents. Surgeon
in France does an operation in America, so you just have a good optical fibre link in
between. Now the ethical question: a surgeon in France, operating on a man in the United
States, using a robot that comes from Germany. Patient dies: who do you sue? I don’t know.
So in the future, what we’ll move to is probably one worker remotely managing many machines,
and this model is already happening. These rovers on Mars are not remote-controlled.
You can’t joy-stick this thing, because it takes eight minutes for the command to go
to Mars and for the imagery to come back to see how it works. You just can’t joy-stick
something with a sixteen minute delay, during when you move your thumb a little bit, and
when you see what happened. You’d go completely mad. They can’t even do it to the Moon and
that’s only like a half-second round trip. So these machines are remotely managed. Every
night they’re given their orders for the next day: “now try and get to that rock over there”.
And then it’s the robot’s job during the day to try and get to the rock over there. If
it can that’s great. If it can’t then it reports back “failure” and they plan another mission.
Yep, so that’s the way it works. So this machine is managed. To work you give it commands and
it will carry them out. The underground mining machine that I showed you earlier is the same.
This lady is not actually driving the machine. She is managing the machine, she is telling
it what to do. Most of the time the machine’s driving itself, so what they’re looking at
here is that this lady could be controlling a fleet of machines. Right? So she’s a manager
and she’s got five robot workers, and most of the time the robot workers can do everything
by themselves, but occasionally they get stuck or confused, because they’re only robots and
so she may have to intervene and help it along, as a manager has to do. An employee comes
along with a problem, and they try and help them fix it. That’s the model that we’re going
to move towards. So, future case: mine in Australia, optic fibre network, driver sitting
in China, joysticking the machine in the Pilbara. There’s no reason that the worker needs to
come to this country. They don’t need a 457 visa. They need a joystick and a decent home
broadband system. The worker could be anywhere in the world. And there’s all sorts of implications
at this. What if something bad happens? What if the excavator falls down and kills somebody?
Your workers in China on the other end, you don’t even know who he is, or she. There’s
all sorts of authentication issues, there’s security issues. What if your robot gets hacked?
All this is now possible, and the nonsense of fly-in and fly-out is really expensive,
disruptive to families, disruptive to the towns where people are flying in and flying
out of, could all be a thing of the past. We don’t need to move the workers to do the
work. Worker and workplace have become separated. So that’s it. Take-home messages. Not too
many. Robots are real. Right? They exist, and every day they can do more and more than
they used to be able to do. Very diverse in form and function. They’re part of the solution,
I think, to a lot of problems we face as a society and as a planet. They’re not the solution
to everything, they’re solutions to some things that we face. But I think it is time now to
start to think about the wider implications, in terms of law, in terms of ethics, and all
the sorts of things that we think are important. I will stop there, thank you.

Leave a Reply

Your email address will not be published. Required fields are marked *