Have you ever wondered
why we do not have the all singing,
all dancing robots that have been
promised for many years
in science fiction and books,
plays, and movies?
We look around us and see robots
starting to infiltrate our lives,
but we still seem a long way
from our robotic companion,
who will do the dishes,
take out the rubbish,
and drive us to our destination.
I have a robotic vacuum cleaner,
which we’ve named Bennie.
Bennie is wonderful at cleaning the floor.
He does random movements.
He wanders round, and he eventually
manages to cover the full floor.
You can now get vacuum cleaners,
robotic vacuum cleaners,
that follow maps of your floor
and thus ensure
that you get the full area.
But if you put down a long-pile carpet
or you encounter an electric cable,
its sensors can’t really deal with that,
and it resorts to just sitting there,
flashing a red light and giving out a huge
long beep, calling for human aid.
And as for stairs, no chance,
So they can be Wi-Fi enabled,
and that allows you closer control.
You can move it around and steer it.
But I suspect that our little robot Bennie
is much more interested in talking
to the TV than to us.
In fact, it works well,
but within certain constraints.
Robotics and artificial
intelligence as a subject
came into being in the 1950s and 60s.
And at the time, there was a huge hype
about the possibilities that could
be achieved within the next 20 years.
Then reality hit, raised its little head,
and the promised dates
for the the all singing,
all dancing robot came and went.
The government’s Science Research Council
commissioned a report
called the Lighthill report in 1973,
and this was there to examine
the state of artificial intelligence
in robotic research in the UK.
It was highly critical
of robotics research,
in particular the area of actually
building physical robots,
and that led to a lack of confidence
in the research area,
and therefore a lack
of ongoing government funding.
And only three universities
in the UK were given grants
over the next period of time.
To quote from the report,
“In no part of the field
have the discoveries made so far
produced the major impact
that was then promised.”
And that’s that’s quite
a damning report, really.
So the actual area of building robots
was not funded very much
over the next few years.
Now in contrast,
the area of what was called computer
based control arm had a lot of funding,
especially from industrial companies,
and this was in the area
of robot arms, manipulator arms.
And this led to the successful development
and incorporation of robotic arms
in assembly and manufacture.
And you'll see lots of factories with
these huge arms and performing assembly.
Now, this meant that the robots
were highly engineered
for the environment.
They were put into highly
structured environments
and they were caged
away from human beings,
both for the safety of the human
beings and for the robot,
in that the humans couldn’t come
and interfere with what it was doing.
But in other areas -
mobile robots, flying robots -
the robots didn't leave the labs
until many years to come.
So what was the problem?
Why did these robots with human-like
capabilities fail to materialize?
Well, the extent of the difficulty
of the problem of building robots
was severely underestimated.
The thinking at the time was you can build
a system to sense what the robot’s doing,
you can build a separate system
which plans where it’s to go,
and you manipulate, you know,
you build a system to manipulate objects
You put them all together,
and the problem’s solved.
No, it has since become clear
that this is not the case,
as the real world intervenes
and throws up uncertainties
and unpredictable interactions.
You’ve got dynamical worlds.
You've got things that move and change
that cannot be pre-programmed.
So jump forward to the early 2000s.
Among other shows,
the TV program “Robot Wars” got people
interested in robots once more.
Being a judge on it for a couple of years
gave me personal insight
into how we could now provide
fairly robust engineered systems.
But it was still very much the human
being that was in control.
Autonomous systems
were still a long way away.
Again, jump forward another 20 years.
We are starting now to see robots
join us in the real world.
So why now? What has changed?
Well, one of the important things
that we see during the last 50 years
is that we have much smaller,
faster, more powerful computers.
We have better materials,
more suitable for building
the flexible, complex robots that we need
that will work in our environment.
We have much better sensor capability.
In the early days,
if you had a robot with a camera,
you would not have very good resolution.
Now we have a very good resolution,
and the problem is still how to understand
what the picture actually shows.
We can see what’s in a camera image,
you’ve got a set of pixels,
but what do they actually mean?
And that's really one of the problems
we're still encountering.
And overall,
we have a better understanding of how
all this interacts with the real world.
We have robots that can work
beside us, rather than in cages.
We have drones which are able
to do limited deliveries
in quite set environments.
And as previously described,
we have things like Bennie,
my little robot vacuum cleaner
that will do our bidding around the house.
We are told we will have autonomous cars
soon that will drive us around the roads.
But we still don’t have that initial
vision from back in the 50s and 60s
of the general-purpose robot
that can do everything.
If we think about this though,
all humans and living creatures are
actually adapted to specific environments.
Some human beings probably
are more general than others.
Even though we can navigate
a diverse landscape of abilities,
we still, if we have problems with vision,
need glasses to see.
We need binoculars and telescopes
to allow us to see much further.
And we have machines that allow us
with our limited sensory system of eyes
to see into the ultraviolet and infrared.
So I leave you with the thought
that perhaps we will never have
that general-purpose robot
dreamed of in the 50s.
This may be far
too complex and unnecessary.
Perhaps instead,
we will look to have robots
that are adapted very well
to particular tasks.
And these robots could
perhaps use their own machines
or our machines or other robots if they
need to perform more specialized tasks.
Thank you.
