of physical device.

They're not the limits of concrete or steel

or anything like that in the physical world,

the limits to computing are the limits of your imagination.

When the first computers began to appear

in the late 1940s, early 1950s,

people were fascinated by these incredibly complex machines

that could do things like process huge numbers

of mathematical equations incredibly quickly.

And so there was a buzz at the time

around these electronic brains.

Lots of people thinking around

could machines really be intelligent?

So Alan Turing, I think, was one

of the most remarkable people in the 20th century.

What he did was he invented a beautiful test.

He said, "Look, here is this test.

If we ever get something that would passes it,

then just stop asking that question

because you can't tell the difference."

He never really expected that anybody was seriously going

to try it out, but actually people did try it out.

But it's been wildly misinterpreted, I say since then.

What Turing's most famous for is working at Bletchley Park,

a code breaking center in the United Kingdom

throughout the Second World War.

And actually, if that was the only thing he'd done

in his life, he would have a place in the history books.

But almost as a side product of his PhD work,

he invented computers, which are just any one of

the most remarkable sort of, you know, quirks of history.

And with incredible precociousness,

he picked one of the biggest mathematical problems

of the age, the Entscheidungsproblem,

which means decision problem;

whether mathematics can be reduced to following a recipe.

So the question that Turing asked was, is it the case

that for any mathematical problem that you might come up

with, you can find a recipe which you can just follow

in the same way that you would follow for arithmetic?

Incredibly quickly, within about 18 months,

Turing solved it.

And the answer is no,

mathematics doesn't reduce to following a recipe.

But the interesting thing is what Turing did

is to solve that problem, he had to invent a machine

which could follow instructions,

and nowadays we call them Turing machines,

but actually it's basically a modern computer.

And he was one of the first serious thinkers about AI.

And so in 1950, he published what we think

is the first real scientific work

around artificial intelligence.

If we ever achieved the ultimate dream of AI,

which I call the Hollywood dream of AI,

the kind of thing that we see in Hollywood movies,

then we will have created machines

that are conscious potentially in the same way

that human beings are.

So Alan Turing, I think, was really frustrated

by people saying, "Well, no, of course these machines can't

be intelligent or creative or think or reason," and so on.

So Turing's genius was he invented a beautiful test.

We call it the Turing Test in his honor.

Okay, so here's how the test goes.

You've got somebody like me,

who's sitting at a computer terminal with a keyboard

and a screen, and I'm allowed to ask questions.

I type out questions on the keyboard,

but I don't know what's on the other end, right?

I don't know whether it's a computer program

or another human being.

So Turing's genius was this.

He said, "Well, look, imagine after a reasonable amount

of time, you just can't tell whether it's a person

or a machine on the other end.

If a machine can fool you into not being able

to tell that it's a machine,

then stop arguing about whether it's really intelligent

because it's doing something indistinguishable.

You can't tell the difference.

So you may as well accept that it's doing something

which is intelligent."

And I think Turing never really expected

that people would seriously try it out.

And so there are annual Turing Test competitions

across the world where people will enter computer programs,

and there will be judges who will try

and tell whether they're a computer program

or a human being.

Most of the entries in them are like these kind

of crude internet chat bots.

And what these chat bots do is they just look for keywords

like sad or family or lonely or drunk

or something like that.

And they plug that keyword into a canned response.

And so they're really just trying to fool the investigators.

And I have to say- there's not really much AI

in most of those programs because they're doing something

which is really much more superficial.

It's a boom period for AI now because a very narrow class

of techniques have turned out to be very successful

on a wide range of problems.

Modern AI is really focused on doing very specific tasks,

and in those specific tasks like playing a game of chess

or something, it might be better

than any living human being,

but it can't do anything else.

Those programs are just tuned, very finely tuned,

to do one tiny thing very well.

But going back to Alan Turing's invention,

he invented machines that for following lists

of instructions, but the computer

can adapt those instructions.

It can learn to adapt those instructions over time.

And that's basically what you're doing in machine learning.

He's one of the founders of the field,

not just of computing, but also of artificial intelligence.

One of the most remarkable people of the 20th century.