The Horizon Guide to AI (2018) - full transcript

Artificial intelligence.

What is it? How does it work?
What can it do?

By looking back through
almost 60 years of Horizon

archive, we can trace its growth,

and we can also track how our
relationship with artificially

intelligent technology seems to be
following a pattern.

First comes bright-eyed
optimism and idealism...

The only thing we can be sure of
about the future is that it

will be absolutely fantastic.

Then fear and caution.

The artificial intelligences of the
future might keep us as pets.

Then it's back to hope and belief.


This pattern seems to have emerged
because of the practical

applications and limitations of the
evolving technology around us.

Hello, Ken. Say, I was wondering...

With each new technological leap,

we're able to invent machines that
can do new and different things.

It caught it!

And we therefore develop different
fears to deal with them.

Human life has been wiped out

by the nano bot.

You've got to admit that computers
can get out of control.

It wasn't so long ago we thought
artificially intelligent

robots would take over the world.

Now it's more likely to be a
mainframe with access to our

banking details.

Are we about to enter a new decade,

optimistic about all that AI
can deliver?

Or fearful of its ability to control
vast areas of our lives?

The roots of the modern computer
were largely developed during

the Second World War.

But it was really during
the 1950s and '60s that scientists,

engineers, and governments

just how important this new
technology was going to be.

These we call computers.

They can be made as big and
complicated as that

installation over there, or as small
but still complicated as that.

And that little computer is
capable of logic.

This is a typical, small, modern,
commercial computer.

It cost about £35,000,

or you could hire it for something
like £20 an hour.

It runs off an ordinary mains plug,

and it's capable of doing something

like 1,000 calculations every

These were much more than just
fancy adding machines

and were becoming capable of an
ever-increasing array of tasks.

Scientists use them more and more to
simulate real-life situations,

from the acoustics of a concert hall
to electronic circuits.

To study the nature of the
human voice,

they programmed a computer to sing a
song and accompany itself.


# Give me your answer do... #

This is the master science fiction
writer Arthur C Clarke on the

set of the film
2001: A Space Odyssey.

For him, the best scientists were
those inspired by classic sci-fi.

One mark of a first-rate scientist
is an interest in science

fiction, and conversely one mark of
a second-rate scientist is a

lack of interest in science fiction.

Science-fiction was important in
shaping our visions of the future.

In the top of his house in Newark,
New Jersey, Sam Moskowitz has

filed away copies of every
science-fiction magazine and

book ever published in the
English language,

and he's read most of them.

Science-fiction was an expression
of 20th century man's hopes,

dreams, and aspirations.

A heavy percentage of
science-fiction is merely rubbish.

Bug eyed monsters, space battles,
or something of that nature.


However, science-fiction has one
thing in its favour.

That even a story which may be from
a literary standpoint

absolute trash
may prove very prophetic.

The George Washington of
science-fiction, the father

of science-fiction, is generally
acknowledged to be Hugo Gernsback.

Hugo Gernsback published in
April 1926 the first science-fiction

magazine in the history of
the world.

It was called Amazing Stories.

Anything you can think of that is
being done today in the space age

we have done before in some way or
another in my magazines.

If by some miracle a prophet
could describe the future exactly

as it was going to take place,
his predictions would sound so

absurd, so far-fetched, that
everybody would laugh him to scorn.

The only thing we can be sure of
about the future is that it

will be absolutely fantastic.

This is a city of the near future,
planned by scientists and designers

for the General Motors exhibit at
the World's Fair in New York.

They see a future where man will be
making fuller use of the

world's, at present,
untapped resources.

The future was a shiny place in
which anything was possible,

and all thanks to science.

Improved technology will make it
possible to penetrate jungles

and build roads with tools so
efficient that, from tree

cutting with a laser beam to laying
road foundations, will be a

matter of only a few short
hours with equipment like this.

These tools, it was imagined,
would be powered by some kind of

machine intelligence, which was
surely just around the corner.

Science-fiction was already
becoming science fact.

The development men are always
on the lookout.

Recently, they produced a new
version of an old idea.

The picture phone.

And last year, they put it into
experimental service between

three cities and the
World's Fair in New York.

Hello, John. Hello, Ken.

Say, I was wondering if you had yet
seen Winteringham's

demonstration of frame storage.

Society hoped technology could
improve everything,

even this man's driving.

We're driving along a section
of Route 128.

And I'm trying to get an estimate of
the demand made upon a driver

by this particular section of road.

It was a time of scientific
optimism. A technological

utopia was emerging in which our
every material need would be met.

In the garden of his house,

Professor Thring tests out the
other half of the robot, its legs.

An engineer himself, he believes
that engineers are the only

people who can really change
the future, for good or for evil.

The evil he wants to eliminate is
the boredom of drudgery,

and the robot he hopes
will do just that.

Mr Ford, would you set the robot to
clear the table? Yes.

Some Americans may believe there is
no future for a domestic robot,

but here in England,
in this house on the outskirts of

Epping Forest,
there is one actually working.

This was a future promising
unbridled leisure time,

a society freed from the
shackles of work.

For example, you can envision in the
future a housewife dialling

up a grocery store, then pressing
some more buttons to indicate what

groceries she wants
delivered to her house.

She can authorise the bank,
again by dialling a code,

to automatically remit from her
computerised checking account

to the checking account of
the grocery store and so on.

And the primary agents of all this
change would be the robots.

Every day, in our homes and offices,
as well as in our motor cars,

hundreds of these little robots are
doing more things for us than

we realise,

taking care of the routine tasks and
leaving us free to live and

work and play in greater ease and
comfort and safety.

Of course, in the '60s, it was only
women who did the housework.

Is the housewife also safe
from losing her time honoured jobs,

scrubbing, dusting, washing?

Or can she look forward to having a
robot, a robot about the house?

If I put the robot in the house and
I finally teach the robot

to program and dust the entire
house - let's say I've taught it to

walk around and perform
all the dusting features -

how long do you think it'll be
before the wife wants to

change the furniture around?

Even the drudgery of applying
lipstick would be confined to

the history books.

The most intelligent inhabitants of
that future world won't be men.

The remote descendants of
today's computers.

I suspect that organic or biological
evolution has about come to

its end, and we're now at the
beginning of inorganic or

mechanical evolution, which will be
thousands of times swifter.

Then there's the biological

the development of synthetic life.

The control of living organisms,
genetic control,

tremendous possibilities,
and tremendous perils here.

And the cybernetics,
the coming of the robots.

And, in fact, to some extent,
the robots are already here.

They certainly were.

And like a toddler standing up for
the very first time,

they were learning to keep
their balance.

Now, this machine works just the
same, in the same way that

you gain your balance when you're
standing on the ground.

When you fall forward,

you push down on your toes to
regain your balance.

Then if you were falling

you would put on your heels so you
could come forward.

I need not look to have spatial

All I have to do is close my eyes,
because you and I can

naturally balance ourselves
with our eyes closed,

and I can feel myself going back and
having the same pressure on

my heels as I would when I was
standing on the ground.

Some visionaries, like writer and
professor of biochemistry at

Boston University Isaac Asimov, were
way ahead of their time,

even envisaging a distant future
where robots and humans would

merge and become one.

We will have a robot becoming less
metal, more organic.

We may have a society in which
robots will drift away from

total metal
toward the organic.

And human beings will drift away
from the total organic toward

the metal.

Asimov's point is perfectly
illustrated in this BBC

television play, broadcast in 1966.

So, you see, nobody must know about

I mean that he's a robot.


His official designation is TN3,
but he answers to Tony.

Him? If you think Tony is human,
then try and feel his pulse. No.

Go on, try. No, stop.

Tony, will you give Mrs Belmont your

I wonder if we will make robots
so much like men and men so much

like robots that we'll eventually
lose the distinction

altogether and have a
combined culture.

This may be the best after all.

Maybe humanity itself will die out
as humanity, and sort of melt

into this machine culture, which
won't look like a machine culture,

but to an untutored Martian may look
like a human culture all along.

As technology changed the world
around them,

scientists of the time knew that
robots were becoming a reality.

Asimov even invented three laws of
robotics to restrict what

they'd be capable of.

The first law is as follows.

A robot may not harm a human being
or through inaction, allow a

human being to come to harm.

Number two. A robot must obey orders
given it by qualified personnel,

unless those orders violate
rule number one.

In other words, a robot can't be
ordered to kill a human being.

Rule number three, a robot must
protect its own existence.

After all, it's an expensive
piece of equipment.

Unless that violates rules
one or two.

A robot must cheerfully go into
self-destruction if it is in

order to follow an order or to
save a human life.

Although the technology was
relatively primitive,

it was developing at
breakneck speed.

So the first tangible fears about
what it could do began to set in.

Would robots take our jobs and
our livelihoods?

It was certainly looking possible.

Industrialists of the day wondered
how their employees would

react to automation in an interview
clearly of its time.

I mentioned before the three
laws of robotics.

We're designing the robots so
that they're...

So that they'll obey those laws.
They won't do away with people.

And what they'll do is they'll
enable people to do more

human tasks.

The human being... the moron level,

and I bring that up only because so
many people say that, "Well, gee

"whiz, automation and things
like that, what are we going to do

"with our lower class citizens,
so to speak, lower class citizens?"

Well, a moron is actually a highly
developed creature.

And only a person that's designed a
robot can develop a real

respect for a human being.

People will somehow feel that robots
will turn upon them.

a robot turns upon its maker and
rends him limb from limb.

However, in a very abstract sense,
this is precisely what may happen.

Robots as a whole will turn on
humanity in the sense that

a robot culture will be developed

that will make human culture
meaningless, let us say,

and they will come to play a part

in more and more specifically
human environments.

But you must recognise that
behind every act of that robot

there's human thought
that prevailed.

And there is nothing in the robot's
current intelligence

that would enable it to come
even close to independent thought.

And we don't even contemplate giving
the robot independent thought.

There's no need for that.

What you can't have in a robot
is any motivation.

It cannot have its own aim
which it sets for itself,

because for this, an emotional brain
is necessary.

And we could never put emotional
brains into a robot, in my opinion,

and certainly if we did know how to,

we wouldn't be so foolish
as to do so.


By the 1970s, things were changing.

Computers were getting smaller

and were running increasingly
complex programs,

able to outperform humans
on many tasks.

And with true
artificial intelligence

a big step closer to reality,

our concerns deepened,

because it seemed
computers were catching us up.

Benjamin Landey has been playing
tournament chess

with his fellow humans for 26 years,
with great success.

Now he's pitting
his ten million nerve cells

against the 40,000 microcircuits
of a computer.

OK, I'll type in the command
to play black for the computer

and then I'll type in his move,
which is Pawn to King 4.

Last year, the computer,

a fellow member of
the American Chess Association,

beat Mr Landey.

There was an outcry.

American chess journals
described Landey's defeat

as a disgrace to the human race.

What sort of mechanical brain
can challenge us in this way?

Some of us think of them
as inhumanly dull.

To others, it seems that computers
are baffling their creators

with prodigious intellectual feats.

We are warned that machines must
be restrained, lest they take over.

Computers draining all the fun
out of chess

sparked fevered debate
in the nation's pubs.

Where would it all end?!

To these programmers,
a computer presents

a constantly changing image.

Is it a good servant,
or a bad master?

Is it more than it seems?

It's something to fear.

You've got to admit
that a situation can arise

with a computer
which gets out of control,

and computers can
get out of control.

But we must have more strict control

over the way in which
they're programmed,

which is an entirely different
kettle of fish. Oh, absolutely.

So we should keep half an eye
on what they're doing

so that when they make silly
mistakes, we can catch them.

Of course they should be afraid.
These things CAN take over.

And all you should do is build up

a complete and utter scepticism
about their abilities.


The thing is, you see,
nobody's really come up with

an intelligent machine yet.

Yes, but how do you define

It always comes back to
this one question.

Always, invariably,
back to the question,

how do you define intelligence?

An intelligent machine is a machine
which can do something

which, until the machine did it,

was regarded as the perquisite of
intelligent man.

What were once thought of as
purely human abilities

were now actually being
programmed into computers.

One particular feature
of a computer program

makes it a really powerful basis for
creating artificial intelligence -

its ability to make a choice.

If you have a programme
that you feed to a computer,

there will be instructions
all over the place

saying, "If such-and-such
is now the case, then do this,

"but if not, then do that."

And to "do this" or "do that"

may mean move to some other point
in the programme

and start from there.

Rather like a recipe that runs,
"If the milk boils over, then..."

Er, "..but if not, carry on."

The hand-eye machine
in the computer science department

of Stanford University
is entirely computer-controlled.

There are no human hands
at the other end.

By handling objects like this,
it's building up

its own body of knowledge
about the effect of its actions.

But the implications of a machine

which is building its own model
of the environment

is that it could operate
independently of man.

Such a device would not be built
in man's image.

It might be better to think of it
almost as an executive

rather than just a robot.

If I hire an executive,
I don't expect to have to tell him

exactly what to do about
everything that comes up.

I don't, in short,
have to program him.

This is the point of the executive.

He has a degree of autonomy,
and you let him rip.

Only when he fails do you fire him.

Now, if you have
a very intelligent machine...

..I think one would like to
do the same thing,

and mathematically this appears to
be a possibility,

that programming, as we know it
today, whereby every movement

of the machine has to be put in by
a human being,

will get put in several orders
removed by the machine itself,

and this is perhaps part of what we
mean by artificial intelligence.

However, if we suppose
that this machine

is going to be many times cleverer
than we are,

we really can't afford to let it
go away and carry on,

because we shan't understand
its behaviour.

It was beginning to dawn on
the scientists of the '70s

exactly what computers
could be capable of.

They could make simple choices.

Some had a degree of autonomy.

But one day, it was hoped
they might achieve much more -

perhaps even abstract thought.

Suppose there were to be machines

which did have something
corresponding to

what one regards as facial
expression in another human being,

which did, as it were,
cry out in pain,

which did talk a great deal
about their inner life,

which recounted the dreams
they'd had, and so on -

if machines performed like this,
then one might, I think,

be as inclined to attribute
consciousness in that sense to them

as we are, as one
is to one's fellows.


It's much easier to imagine
a machine creating works of art

than appreciating them, in fact,

because whereas you can to a certain
extent create works of art

by formulae - there already
is a poetry machine in Paris,

which produces not bad
symbolist poetry -

not very good, but not bad.
I once had an eminent critic

mistake one of his performances
for one by Mellarme,

which isn't too bad.

It's very difficult to have machines
doing aesthetic criticism

because aesthetic criticism is
something you can't easily formulise

and machines operate
according to rules,

and there, really, up to now,
no rules about aesthetic criticism.

But if you had, er,
aesthetic criticism formulised,

which I suppose is conceivable -

it might lead to some judgments
nobody would agree with -

if you could do this,
then I don't see why a machine

shouldn't appreciate works of art
in that sense.

The more the technology advanced,

the weirder the experiments became.

Well, it was the 1970s.


As art embraces random association,

the random creation of a computer
becomes more relevant.

For this dance, long bits of
movements, directions and tempo

came spilling from the computer
like so many beads,

with nothing to hold them together.

It's up to the dancers
to provide the string.

Rules for creating by computer
are well-known,

but we don't know how to make the
computer evaluate what it produces.

For appreciation,
there are no rules.

The fact that they are programmed

I don't think does make
an essential difference,

because we are, after all,
programmed too.

Our programming is a much more
complicated affair -

it's a question of learning
a language and of...

being indoctrinated
in all sorts of ways

by, er...the people who do teach us.

And we have a certain
innate endowment,

but so, of course,
does the machine.

The more they're like people
then the more they're like people,

and just as I don't very much mind

where another person was born
or how he was born

or wouldn't very much mind if I
found out he was synthetic -

supposing it suddenly turned out

that you were the first
synthetic person,

I wouldn't feel at all threatened
by you - I'd be fascinated.

The more they're like people,
then the more one deals with them

like people,
and why should one worry?

But people were worried.

They could see how automation and
increasingly intelligent machines

could ruin livelihoods -

and not just those of the working
class, but white-collar workers too.

At one time, 20 men monitored
the nation's power stations

on the National Grid
from this control room.

But the job became too complex
for them.

One man and a computer took over.

100 years ago, Samuel Butler
speculated that the time might come

when man shall become to machines
what the horse and dog are to us.

What is becoming possible

will profoundly affect
our future evolution.

We're on the road
to producing something

which is, of course, dangerous.

Those concerns deepened as the rate
of technological change

continued to accelerate.

The advance in the behaviour of
machines in the last 100 years

was like a billion years of biology.

Its speed has been
millions of times faster

because we can combine
separate improvements directly,

where nature depends upon
chance recombinations.

What's more, there may not be
a point at which we can say,

"Let's stop the technology of
machine intelligence here."

What sort of relationship will man
have with intelligent machines?

Are they to have independence?

What if an independent machine
should make a mistake?

Do you blame the machine?

Now, how does one look at this?
One might say

that the machine is bound to
make a mistake every so often,

therefore it would be unfair
to blame it.

One might find that a fault
had developed in the machine,

some of its components
had gone wrong,

and you'd send the machine
back to be repaired.

But in neither case, surely,

would one want to say,
"Let's punish the machine."

But now if we're saying that
machines of an intelligent kind

in the future, if you like,
are essentially like people,

are we being equally irrational to
say, "Let's punish this person,"

in retribution, if the person
has done something wrong?


By the late 1970s, entrepreneurs
were also getting in on the action,

with their own vision
of what AI could do.

Klatu is a product of
Quasar Industries,

Rutherford, New Jersey -
no relation to the TV sets.

Who programmed you, Klatu?

"Daddy" is Tony Reichelt.

He says Quasar will soon sell robots
for $4,000 each.

In December of 1979

we're going into production of
approximately 125 robots a day

that will be used for domestic
purposes, for the home.

Domestic purposes
meaning it'll do what?

Answer the door when guests arrive,
take their wraps and store them,

vacuum the rugs,
polish the floors...

But Klatu is streets ahead of
any other robot ever built.

The experts are convinced
he's a fraud.

This, on the other hand, is genuine.

He's called the Wabot,
and in some ways

he's the most ambitious robot
ever built.

The Wabot and friends - a group
of students who built him

at Waseda University, Tokyo.

The Wabot has eyes - stereo cameras
buried in its body

which look for an object and direct
the hands where to find it.

There are touch sensors on the back
of the hand that tell the Wabot

it has located the object.
Picking it up is another matter.


The Wabot can also speak.


In Japanese, of course.

There was also another
giant leap forward.

By this time,
computers could take in

and process information
at incredible speeds.

So adept were these
problem-solving machines

that we began to call them

and feel very small and puny
by comparison.

We will have almost no communication
with superintelligent computers

because...the nature of
what they're thinking about

will be completely foreign to us.

I've thought of, as an example,

two machines,
maybe each the size of a desk,

next to each other,
communicating with each other.

I've thought of one maybe as George
and the other as Sam,

and they're talking to each other,
and you walk up and you say...

"Hey, George,
what are you talking about?"

Well, from the time
you started knocking

until you finished
asking your question,

George has already said to Sam

more words than all the utterances
of all the humans that ever lived.

So the question is,
what would it say to you?

This is how the computer
sees the human brain.

The intelligent machine that enabled
our ancient ancestors to survive

in their brutally dangerous world.

The brain's rate of evolution
is extremely slow,

although its intelligence
has stretched to its limits

to produce
some incredible achievements.

Off the ground!

By contrast,
the dangers of the world

have begun to evolve
explosively fast.

The human brain may not be
intelligent enough

to cope for much longer, unaided.

Never before has a computer
attempted anything so profound.

We seem to be witnessing

the spawning of a new kind of

Artificial intelligence.

Already, many computers
are behaving like infant prodigies.

What will they do when they grow up?

Faced with such a rapid rate
of evolution,

it would be reckless for us
to make any predictions.

As these machines evolve and as some
intelligent machines design others

and they get to be smarter
and smarter,

it gets to be fairly difficult
to imagine

how you can have a machine

that's millions of times smarter
than the smartest person

and yet is really our slave,
doing what we want.

We've been training chimpanzees to
talk in sign language.

Of course, a lot of progress
has been made there.

But, if we ever get a chimpanzee
that can really communicate,

and we try to talk them,
what we'd find

is that he's interested
in talking about things

like where can he find a banana,

will you tickle me, or playing games
that chimpanzees like to play.

But, if you want to talk to them
about nuclear disarmament,

and who's going
to be elected president,

they simply won't be interested.

On the other hand, we would
be very little interested

in discussing for long what
a chimpanzee has on its mind.

Likewise, I think that
the artificial intelligences

of the future will be worried
about weighty problems

that we simply can't understand, and
they may condescend to talk to us.

They may amuse us on occasion,

or play games that we like to play,

and in some sense,
they might keep us as pets.

I think that what that means,
really, is they might solve

They might find it necessary
to take some of our toys away,

some of our hydrogen bombs
and things,

but there's no reason that
they would want to go after

the same things we want because
they won't be interested in them.

I once owned a Porsche,
a very high-powered sports car.

But I wouldn't have let, say
a 14-year-old boy drive the car.

A Porsche in those hands
is a dangerous instrument.

I think the state of moral wisdom,
say, of our society,

is such that it is, at best,
a 14-year-old boy,

and perhaps an 11-year-old boy.

Under those circumstances,

the very, very powerful tools
that we're making,

and I'm thinking particularly
of computers, I think have to be

looked at as at least potentially
very dangerous instruments.

The thing we call the computer
does not grow by a tablespoonful

of grey matter every 100,000 years,

which is the case in the
rapid growth of our brain,

but grows a factor of 10
in power every seven years.

The computer generation.

There's no question that it will
match us in narrow reasoning power

by 1990,

and go beyond us to become the great
new intelligent race of the future.

Once the shock passes,

the shock of knowing that machines
can do better than we can,

the beneficial effect will be that
we will have a burden removed

off of our back, which is the burden
of being the supreme

intellectual creature on the planet.

We have to worry about everything
today that isn't worried

about by God,
and when computers come along,

they'll be able to worry about
a lot of big questions

that we're basically
incompetent to worry about.

It seems that, as we journey
through the Horizon archive,

we see the same old pattern and
emotional cycles again and again.

Hello, Igor.

Hello, Igor.

Hello, Igor.

We start by feeling
really optimistic

about the prospect of great change.

Then, when we actually achieve it,
we are filled with fear.

However, the fear subsides once
we get used to the applications

of all that new tech,
and we start feeling positive again.

That's exactly what
happened in the 1980s.

The fear began to subside,
and optimism reigned again.

We'd have friendly robots,
and keep them as pets.

Genghis is an intelligent
micro robot,

that can sense
what it touches and sees.

This robot
has quite a lot of sensors.

It's got belly sensors,

so it can detect if its belly
is contacting something.

It has foot sensing
in each of the legs,

so it can tell
if the motors are stalled.

It has sensors upfront so it can
detect if there is a change

in the heat field,
and it has whiskers,

or antennas on the front, so it can
detect if the front of the robot

comes into contact with an obstacle
and take appropriate action.

So, he's moving back now. Why?

Because he sensed there was an
obstacle in front of him that

was too hard for him to climb over,
so he's decided to go around it.

You referred to it as he.

Well, Genghis was
a male marauding conqueror,

so that's why we refer to him
as a he.

We don't want them to have
a problem with a small ego,

and feel very slight,

so we give them powerful names.

Scientists and engineers began
to re-evaluate intelligence

to understand what thinking
really was,

and to programme
truly intelligent machines.

The theory of AI says that if you
analyse the world in symbols,

and put the right rules
in the machine,

then it will have a mind, and
understand the world as we do.

It will be a thinking machine.

The machine will have to know
all about the world,

the totality of facts.

It has to know that helium
balloons fly upwards,

that children need looking after,

that policemen wear blue helmets...

..that professors
wear special clothes.

Easier said than done.

But, for a brief moment in the
'80s, it seemed possible.

This was largely
due to mass-market production

of the core of all computers and

The microchip.

During the 1980s and early '90s,
the microchip became smaller,

cheaper, and ubiquitous.

Computers really entered
our own homes for the first time.

We became accustomed to having
calculators, video games,

and, as home computers became
part of everyday life,

we worried less

that artificial intelligence
would take over the world.

I think the fear of intelligence
in robotics has subsided because,

by and large, the reality
has not fulfilled the claims,

so people have begun
to be more relaxed about it.

But I think those fears
will re-emerge as the reality

catches up with the advertising.

Scientists pushed ahead in their
pursuit of truly humanlike behaviour

in machines, trying new and
different strategies all the time.

Like simulating biological organic
brains, rather than mechanical ones.

Yes, I want to make a brain
in a culture dish. Yes.

Aizawa is convinced that
conventional computer technology

will not be able to keep abreast
of the increasingly complex

demands of the electronic society.

In spite of the huge advances
being made in today's information

technology, Aizawa believes that
the needs of the next century

will only be met by
soft, brain-like processes.

If we succeed in
artificially designed brain

at the first stage,

and then my ultimate goal
should be creating a life,

but it's a long way to achieve that.

It certainly was a long way off.

As Aizawa was trying to create
an artificial brain in the lab,

many were still trying
to get their robots

to achieve even
the most basic of tasks.

So, this is Cog.

Cog is the first serious attempt
at building an intelligent

humanoid robot.

What we're trying to do here
is build a robot which is able

to interact with people in the way
that people interact with people.

The quest that all us
science-fiction fans grew up

with, of wanting to build a robot
which was just like a human.

The enormity of the challenge
was becoming apparent

to everyone that tried.

It caught it!

As we entered the 21st century,
the pace of change

and the power of computing was set
to reach unprecedented levels.

While computer power was increasing,

the size of its hardware
was becoming much, much smaller.

Being able to make computers
smaller and more portable,

meant whole new technologies
using AI could be dreamt up.

We're making a motorcycle
that drives itself.

Like, can you get any cooler
than that?

That's why everybody on the team
puts in so many hours,

because, when you see it work,
you say, "Wow, I made that!"

But, much like the decades
that came before,

things weren't quite perfect.

TANNOY: Contact made.



While some of the vehicles
have mastered the art

of not crashing,
others clearly haven't.

But artificial vision
is notoriously complicated.

After decades of research,
it's still unproven,

and, at higher speeds,
dangerously unreliable.

It's gone up.


But despite the setbacks,
progress continued.

As aspirations got bigger,
technology got smaller.

One day, scientists
could manipulate molecules

to create tiny computers
that will fuse with our bodies.

Today it's very hard to imagine
things like direct brain links,

you get dismissed as a nutter
if you talk about these things,

but, in 40 or 50 years' time,

it'll actually be quite
possible to do that.

We can imagine molecular computers,

molecular transistors and so on,
being small enough,

we can get these things into contact
with every synapse in your brain.

Suppose in 20 or 30 years from now,

I'm wearing a really smart shirt
and I have an accident.

That shirt knows I've had
an accident because it can

measure the G-force, it might even
measure that I'm bleeding.

The shirt can tell
the ambulance in great detail

while they're on the way
exactly what's wrong with me,

so they've got the equipment
ready for when they arrive.

It might save my life.

As a result of molecular computing
and those sorts of technologies,

over the next few decades
we're going to see far more change

than we've seen
over the last few hundred,

so life will be just beyond

Ultimately, we're not going to
go to doctors and have visits

in the way that we do today.

We're going to have systems in our
body that are continually monitoring

our body, detecting problems,
and fixing them immediately.

But very quickly, our hopes for tiny
artificially intelligent robots were

replaced by a paranoia that they may
start to replicate on their own.

Human life has been wiped out... the nanobite.

We're talking about instruments that
are designed literally atom by atom

and molecule by molecule,

so they're below what you can see.

Miniature, artificially
intelligent robots,

with the power to create or destroy
whatever we ask them to.

The intelligence of nanotechnology

will not be in one nanorobot,
or nanobite.

It will be a collective intelligence
of millions, actually trillions,

of nanobites working together,
pooling their thinking resources.

If that gets out of control,
we would have essentially

a nonbiological cancer that
could just eat up, you know,

the natural world - that's the
so-called "grey goo" problem.

So great is this fear of the "grey
goo" that eminent figures around

the world, such as Prince Charles,
have raised concerns about it.

Luckily for us, and Prince Charles,

the fears of grey goo were never
realised, and, as the dust settled,

we continue to focus
on the applications

of an intelligent computer's power,
rather than its size.

The American engineer and co-founder
of Intel, Gordon Moore,

had made a startling prediction.

Every year,
the power of computers was doubling,

and Moore was convinced
that this unprecedented growth

was set to continue, without end.

Known as Moore's Law,
the prediction had become true.

Computers are about a billion times
more powerful than they were

a quarter of a century ago,
and they will become a billion times

more powerful than they are today
in a quarter of a century.

We'll have both the hardware
and the software

to recreate human intelligence
in a machine.

By creating nonbiological
intelligence machines,

that are ultimately billions
of times more capable than human

beings today, we will integrate
with this technology,

and it will enhance human potential.

Since Horizon has been following
scientists trying to create

artificially intelligent robots
through the decades,

it's clear that there's been
a number of big breakthroughs.

It's also clear that progress
is much harder than first imagined.

Stair, please fix
the stapler from the lab.

Stair, please fix
the stapler from the lab.

I will go get the stapler for you.

You didn't tell me he could talk.

It turns out getting robots to talk
is not the most difficult thing.

It's blindingly obvious to you
and me where the stapler is,

but for a robot to figure this out
is actually surprisingly difficult.

Oh, he's missed it. Oop!

Stair knows the basic layout
of the building,

and has other senses
that mean it can avoid crashing

into unexpected obstacles.

Come on, left! Left!

It knows what a stapler is,

and it knows which room
the stapler is in.

But other than that,
it's on its own.

Look, he's going down,
he's going for it.

You know, when we pick up things,
there are lots of ways to do it.

When you pick up a coffee cup,
or you pick up a bottle of water,

the motions you make with your
hands are very different from

the motions you make with your hands
when you pick up a stapler.

It has to choose for itself
how to do that. It's done it.

Yes, it's picked up the stapler.
Stair has done it.

It's picked up the stapler.

It can find and pick up a stapler,
incredibly difficult task,

but not quite the home help
of the 1950s techno utopia.

These robots have been
set up to develop,

much like a young child does.

They're beginning to understand
how their bodies work by looking

at a reflection of themselves.

You've got a mirror here,
so what's the...?

Well, the experiment is about,

that a robot would learn
something about its own body.

Because, in order to move in the
world, in order to control it,

in order to also recognise
the movements of another,

you need to have some sort
of model of your own body.

And the way that the model
is going to be built up

is that the robot is doing actions,

and watching itself
performing these actions.

So, to get a relation
between the visual image

and movement of the motor.

So here,
you see it is looking at its hand,

and you also see very much how
it is trying to keep balance.

It is pretty impressive, actually.

How these motor commands are sort
of in an early phase, right?

It's extraordinary, isn't it?

It really does look like
it's encountering itself

for the first time.

But what's even more remarkable
about Luke's robots is that,

once they're able to recognise
themselves, they start to

evolve their own language,
and communicate with each other.

One of them is going to speak,

and then he's going to ask
the other one to do an action.

He is going to invent a word
because he doesn't have yet the word

to name that action. Right.

OK, so then he says the word,

and this one isn't sure
what it could mean,

it's a brand-new word.

So then he will make a guess,
and if the guess is OK,

totally by luck... Right.
..they both know this word,

and they know for the future
they can use this word

to communicate with each other.

ROBOT: Keemaboo.

OK, so he is speaking first.

He is doing the action.

That's fantastic!

You notice how he looked.
OK, so now he is...

He is showing the real action.
Ah, right, OK.

So now there is another
interaction going to happen.

Again, I don't know which one
is going to speak. Keemaboo. OK.

Is that the word it just learned?
Keemaboo? Yes, yes.

Of course he knows already,
so he's doing it.

And he will say yes,
presumably, will he? Yes.

Over the decades,
we've dreamed of a future

with artificially intelligent

As the Horizon archive has shown,
we've weaved our way through

cycles of optimism and pessimism,
fear and wonder, and back again,

as our future alongside AI
has been brought into focus.

But, while engineers and scientists
have worked to achieve the

reality of a truly humanoid robot,

an unexpected form of AI
has been creeping up on us.

This surprising shift
has changed everything.

Our future would be built
not on robots, but on data.

Now, dreams of robots are being
replaced by the realities

of cloud computing, of networks that
stretch to all corners of the globe,

connecting people and information
in ways we'd never dreamed possible.

Colossal amounts of data uploaded
and downloaded across the world

every second, allow us
to communicate and navigate,

to monitor our health
and diagnose disease,

and to track payments and shipments,

even to find love.

The reason I believe it's good
to be a technology optimist

is, throughout the entire history
of the human race,

technology has empowered us,

from the very early days,
the Bronze Age, the Stone Age,

to today's smartphones
and modern medicine,

it's freed us, it has levelled
the playing field for everyone,

it's empowered us as a human race.

Why stop that?

Artificial intelligence
has integrated itself into the

very fabric of our lives,
almost without us noticing.

And in ways we could have
never imagined.

Smart homes are now a reality,
where lighting, heating,

and other electronic devices,

can all make seemingly
intelligent decisions.

Wake up in the morning,
the house knows that I am waking up.

It can wake up with me.
When we walk in the kitchen,

it will play the local news
and greet us into the day.

Tell us the weather forecast,
so we know how to dress.

Artificial intelligence is appearing
everywhere, and powering everything.

In warehouses,
AI is enabling robots to organise,

move and prepare
millions of online orders.

This is basically
a very large warehouse,

where orders come in
and they need to be fulfilled.

So, the system figures out
which robots need to go

to which pods, pick it up,
and bring it

to the perimeter of the warehouse,

where people take things
off of the pods

and put them into the orders,
which eventually go out.

There's a lot of robots in action
here. How come they don't collide?

The robots have to generate
trajectories and plans,

and those plans are then shared
to a coordinator,

which then figures out
how they should go,

and execute their plan
so that they don't hit each other.

Artificial intelligence
has changed the world of medicine.

Some old fears have crept back in.

Should I be worried as a GP?

The thought that a robot,
or artificial intelligence,

could take my job just seems crazy.

I'm going to pose as a patient and
give myself an imaginary condition.

And then we can see just how
accurate the machine really is.

May I ask, please,
what's troubling you today?

I am feeling tired all the time.

So, as well as feeling tired,
I've had been feeling kind of weak.

Let's tell the computer that.

Do you get breathless on exertion?
Yes, I do.

Thanks, I have noted this.

So, I've given the computer
all of my symptoms now,

and it's come up with a diagnosis.

You can see that I have put down

and the computer has said
uterine leiomyoma,

which is actually the same thing.

What the circles represent
are diseases,

symptoms and risk factors,

and what the lines represent are
the relationships between those,

so based on that, the computer
has taught itself actually

how strongly related those diseases,
symptoms and risk factors are.

OK, so that's how it's termed,

the probability is from looking
at past real-life cases.

Absolutely, and that's why
this is machine learning.

But machines can only learn from
information that they are given.

And that information comes from us.

Facebook today could not exist
without AI. It's as simple as that.

Over a billion people
use Facebook everyday,

and they load their news feed a few
dozen times every single day.

And, you know, if you imagine
how many people it would take,

if you were to line up all the
pieces of content that are available

to you every day, if I had
to sort out how relevant this

story is going to be to this person,

you multiply that by a billion
people, we see that for a human,

this would be a task that
is absolutely impossible to do.

Every day, AI technology
is learning from us.

The more it learns,
the better it becomes.

But, as this rapid change
is occurring, we are once again

becoming fearful of its
potential impact on our lives.

They thought this was going to be
funny, they were teenagers,

so they didn't think
about the implications

of deleting everything
someone owns,

and how much precious data
you may have in your life.

I mean, data is quite precious
to people now, it's valuable.

Artificial intelligence,

machines that seemingly
think for themselves,

are already changing our lives.

Looking back through
the Horizon archive,

it's clear our relationship with
AI has followed a pattern.

Optimism, followed by
a fear of the unknown.

If this cycle continues, and today's
anxieties about AI using our

data in a bad way, soon may evolve
into something more positive.

Scientists are again dreaming
of the cities of the future,

and the wonderful things
we'll be able to achieve.

However, creating a robot that
can truly think for itself

may be decades, or even centuries
away, it may never happen.

But the merging of man and machine,
the organic and the metal,

human and AI,
is happening right now,

just as Isaac Asimov predicted.

It seems to me that, as robots
become continually more advanced,

that people will not try to keep
it entirely a matter of metal

and electrons.

We will have human beings who
will make more and more use

of artificial organs,
of metal and plastic.

Meet Erik Sorto.

Deep inside his brain
are two arrays of electrodes.

I suffered a gunshot...

..which left me paralysed
from the shoulders down.

Erik's spinal cord was severed...

..stopping the signals from his
brain that control movement

reaching his limbs.

I'm a C3-C4 complete
quadriplegic complete.

To try to restore movement
he has lost,

Erik is part of a trial
to merge his brain with a robot arm.

In short, we may have a society
in which robots

will drift away from total metal

toward the organic,

and human beings will drift away
from the total organic

toward the metal, and plastic.

And that, somewhere in the middle,
they may eventually meet.

Now, when we have a kind of
metal-organic hybrid,

will it matter
that he was originally metal

and became metal-organic,

or that he was originally organic,
and became metal-organic?

Or will it not matter?

Will we then have formed
a kind of mixed culture...

..which perhaps might be higher,
or more efficient, better?

In the beginning,
I was very conscious of them.

Now I completely forget
they are there,

until someone reminds me, like,
"What's that on your head?"

I'm like, "Oh yeah, that's right,

"I have two pedestals
sticking out of my head!"

Now, Erik is ready to try
to pick up a bottle of beer

using just his thoughts.

It's the big moment. Let's do it.
Are you ready?

All right. OK, here we go.

He thinks only of the goal of the
movement, "bring hand to mouth",

and the robot arm
works out the rest.

There you go.

First step done.

When you go to reach for something,

you don't walk it step-by-step,
you just do it.

Once the arm
has grasped the bottle,

Erik thinks
"bring hand to mouth" again.

Is there anything essentially
horrible about thinking

that man has the right
to create a pseudo-living system,

just as nature did?


All right!

Hey, you finished that thing off?
That's good!

His progress in this
extraordinary trial

has extended what it is to be human.


In the beginning, it was my brain,
my arm, and the robotic arm.

Now when I go in there, it is my
brain and the arm. We are one.

And it feels like my arm.
I think the brain is...

It's a point of fact, it is ready to
use any tool available to keep on...

To keep us moving forward,

and helping us live a better life.