The Man Who Knew Too Much: Jack Copeland on the Life and Work of Alan Turing
Nov 16, 2024
Alan Turing (1912–1954) was an English mathematician, logician, pioneer of computer science, and wartime code-breaker. He is credited with creating a design for the Automatic Computing Engine (ACE), the early electronic stored-program computer, as well as the Bombe—a decryption device that the British government used during WWII to crack the German “Enigma,” machine, which encrypted secret messages.
Turing invented a machine that changed all our lives.
He called it the universal Turing machine.
I guess everyone who owns a personal computer nowadays
knows that the way to make the machine do what you want,
word processing say,
is to look around in memory for the right program
and then click on it.
It's called the stored program concept
and it's, you know, the fundamental feature of modern computers.
And that's what Turing invented in 1936,
the stored program concept.
There was no actual engineering in those days.
He was just working on paper.
And he showed that by inserting different programs in memory,
you could make the machine carry out any task
for which a computer program could be written.
Not that he would have expressed it that way
because they didn't have the term computer program in those days.
He talked about instruction tables.
That was his word for what we now call computer programs.
So Turing's fabulous idea really
was for a single machine of fixed structure,
a single slab of hardware,
that could change itself chameleon-like
from a machine dedicated to one task
into a machine dedicated to a completely different task.
There was no need to change the wiring of the machine.
You just typed in a new program into the machine's memory.
I guess nowadays,
when almost everyone owns a physical realization
of a universal Turing machine,
Turing's idea of a one-stop-shop computing machine
is liable to seem as obvious as the wheel or the arch.
But back in those days,
when engineers thought about designing
specific machines for specific purposes,
different machines for different tasks,
Turing's idea of a single machine that could see off the lot
was revolutionary, startling.
So that was the basic principle of the modern computer
invented by Turing in 1936.
He was born about a century ago in London,
about a mile from Paddington Station in the centre of London.
His parents, his mother Sarah and his father Julius,
his father worked in the Indian Civil Service
and so didn't spend much time in Britain.
Julius and Sarah would come to Britain on leave,
once every 18 months or so,
and then they'd go back to Madras, Chennai,
as it's now called, where he had his job.
So baby Turing, baby Alan,
got left behind in London and the south of England
for all his young life.
He never went with them to India.
And he lived the life of an orphan,
an extremely privileged orphan.
There were servants and there were foreign holidays,
which was unusual in those days.
There was enough money,
so it's not that he went short on anything except parental love.
They just were not there for him.
And he was brought up in a series of foster homes.
I can remember reading something his mother said.
They'd just driven a taxi to his school
and dumped him there before driving off to the docks
to get the ship to India.
And Alan was running down the drive
with his arms spread out after the taxi,
you know, just full of horror
of being dumped in this alien place.
And she says that he used to be
such a cheerful and sociable child.
And then the next time she came back,
after another absence of a year or whatever it was,
and she found him completely changed.
He was introverted
and no longer his previous outgoing self.
So he went to two schools,
boarding schools,
he started boarding school at a young age.
And an adolescence of fagging
and deprivation and bullying
awaited him at boarding school.
He seems to have borne it all fairly cheerfully.
But it was far from an ideal childhood, I would say.
But of course he flourished academically,
particularly at his second school, Sherbourne,
down in Dorset in the west of England.
He was taught maths by a canon,
Canon Epperson,
who was a great fan of Alice in Wonderland
and who made a film at one stage called Alice in Numberland.
And he taught Turing maths.
And of course Turing went on to become
one of the greatest mathematicians and mathematical logicians
that Britain has ever produced.
So some credit has to belong to Epperson.
Epperson was rather sniffy about Turing's schoolwork.
Obviously he recognised Turing as a brilliant pupil,
but he said that his methods were
sometimes brilliant but unsound
and sometimes clumsy and ungainly
or something like that.
I can't remember the exact words.
But he certainly wasn't full of praise for Turing.
And I can remember
in Turing's mother's account of his school days,
she mentions that one of his masters at Sherbourne
offered to donate a billion pounds
to any charity that Turing named
if he passed his Latin exam.
A fairly safe bet, evidently.
So Turing, he emerged at Sherbourne
as this very unusual child,
a very gifted child.
His parents put his name down for
King's College Cambridge about a couple of years
before Turing was due to leave school.
And their dream came true.
And so eventually Turing walked through
the grand portals of King's College Cambridge
to study mathematics.
Kurt Gödel was a brilliant young mathematician
working at the University of Vienna.
And in 1931 he proved what's now
what was called Gödel's incompleteness theorem.
What Gödel showed really was that in mathematics
proof is a much narrower concept than truth.
There are mathematical truths
that you can't prove.
That's essentially what he showed.
Turing knew something about Gödel's work.
Turing went to some lectures in Cambridge
in 1935 given by Max Neumann
who later became one of the great figures
at Bletchley Park and also in the development
of electronic computers after the war.
But at this stage Max Neumann was
a university lecturer in mathematics and logic
at St John's College Cambridge.
And he gave some lectures on the foundations
of mathematics in 1935 that Turing attended.
And the lectures worked towards
an exposition of Gödel's theorem.
That was sort of the grand finale of the lectures.
But what had far more influence on Turing
was a lecture that Neumann gave on David Hilbert
who was a German mathematician working
at the University of Göttingen.
And Hilbert had the idea that...
He talked about mysterious arts in mathematics.
He thought that mathematicians should rid
mathematics of all mysterious arts.
Things like intuition and insight
and these sort of woolly and vague notions.
He thought that all of this stuff should be replaced
by what they called systematic method.
Now I guess we would call this computer programs
or lists of instructions that a computer could follow.
He thought that mathematics should be founded
on systematic method.
And so Neumann was explaining Hilbert's idea
of a systematic method which is really
the rules for long multiplication that we all learn at school.
That's an example of a systematic method
that everyone would know.
A systematic method is kind of a list of instructions
that you can follow without any insight or ingenuity.
You just follow the instructions one after another
and you get the answer.
You just kind of turn the handle on it
as long as you don't make any mistakes
and the correct answer just sort of pops out.
So that's a systematic method.
And in his lecture Neumann was explaining
Hilbert's idea of a systematic method.
And he said systematic methods
are methods that a machine could carry out.
And that was a very new way of putting it at the time.
And it grabbed Turing's attention
and it set him thinking
well what is the most general kind of machine
you could have for carrying out systematic methods.
And that was what led him to the Turing machine,
the universal Turing machine.
He was completely hooked by this problem
that Neumann had inadvertently posed in his lecture.
And so Turing went away and thought deeply for a year
and came up with the universal Turing machine.
He didn't pay much attention to what Godel had said.
He didn't really talk to anyone about what he was doing.
Turing loved to work alone.
This problem that Neumann had suggested
was Turing's problem.
He didn't see any need to discuss it with anyone else,
not even Neumann himself.
One day another fellow at King's, Richard Braithwaite,
got Turing talking at dinner at High Table in King's.
And Braithwaite sort of said
well what are you working on Turing?
And Turing started chatting about
systematic procedures and computability and so on.
And Braithwaite said something like
hmm, sounds to me like there are some interesting connections
with what this chap Godel is doing in Vienna.
But Turing just wasn't interested.
And Braithwaite later wrote a letter
remarking on how little Turing knew
about Godel's work.
Turing was a desert island worker.
He liked to work out everything for himself from first principles.
Godel on the other hand was a very sociable mathematician.
He was very interested in what other people were doing.
Like Turing he was a great eccentric but nevertheless
he always had his eye on what other people were doing.
He was part of a mathematical discussion
whereas Turing
really just talked to himself in a sense.
He wasn't much interested in what the others were doing.
Godel is full of praise for what Turing
established during his year of deep thought
that eventuated in the universal Turing machine.
Godel's 1931 result was very narrowly focused.
It was focused on a mathematical system,
a system of arithmetic
that Bertrand Russell and Alfred North Whitehead had devised
which nowadays is always called PM
for Principia Mathematica.
So it was supposed to be a logical systemisation of the rules of arithmetic.
And what Godel showed was that
this system PM is incomplete.
That is there are true sentences of arithmetic
kind of complicated relatives of
sentences that we all know like 2 plus 2 equals 4.
There are sentences of arithmetic that are true
but are not provable in the PM system.
So truth goes beyond provability, that's what Godel showed.
But he just showed it in the context of PM.
What Turing's work
enabled Godel to do was generalise this result
to all formal systems of mathematics
containing an amount of basic arithmetic.
And Godel says something like
it was Turing's work that brought him to the right perspective.
And it was thanks to Turing
that he was able to give a general definition
of a formal system.
And then once he got a definition then he could apply
his theorem widely and show that incompleteness
was everywhere or virtually everywhere in mathematics.
Turing was writing up his wonderful results
about the universal Turing machine
in Cambridge in 1936.
And then suddenly a paper arrived from the United States
by Alonzo Church
who was a young Turk in the mathematics department
at Princeton University, a few years older than Turing.
Church had worked with Hilbert at Göttingen
in the late 1920s
and he proved a result about Hilbert's work.
And it was contained in just two pages
of sort of compressed mathematical notation.
And a copy of these two pages arrived in Cambridge
and it was immediately clear to Turing
as soon as he read these pages that he'd been scooped.
Church had proved the same result about Hilbert
that Turing had proved.
You can imagine how he would have felt.
It's the worst thing that can happen to a researcher.
Someone else gets your result and publishes it first.
Fortunately, Turing had the very dynamic
and very sensible Max Neumann on hand
to give him advice.
And it was absolutely clear to Neumann
that Turing's paper went far beyond what Church had shown.
Church didn't have the universal Turing machine.
He just had this very limited result about Hilbert.
And unfortunately, Turing had sort of angled his paper
on this same result about Hilbert.
But the apparatus that Turing used
to produce that result
was really the important part of Turing's paper.
And it was quite different to Church's apparatus.
So although there was sort of slight overlap
between the papers,
Neumann was quite clear that Turing's paper should be published.
There is kind of this rule of etiquette in mathematics
that if someone proves and publishes a result,
then no one else should publish that same result
unless there were important new insights
or the result was developed in new ways.
And to Neumann it was just clear
that Turing's paper contained very important insights
and very important developments
that were just not present in Church's paper.
Other people were certainly interested
in the general area that Turing was working in.
Hilbert was kind of the pope of mathematics
at this time. He was an older figure
and in 1900 he'd pretty much set the agenda
for 20th century mathematics.
So there were many people working on problems
that Hilbert had raised. Turing was just one of them.
And so there were other people working on the idea
of systematic procedures.
So there was a sort of common background
between Turing's work and what various other people were doing.
Most notably Emile Post, but there were others as well.
Herbrand, for example.
But none of these people anticipated
Turing's invention of the universal Turing machine
and the Storm program concept.
That was the key component of his 1936 paper
and no one else anticipated that.
It sort of came out of nowhere.
No one knows quite where Turing got the idea from.
He was looking for a device
that could carry out all systematic procedures
and Newman had given him this idea
of what it should be, a machine.
And so Turing came up with this idea
of literally a machine with a paper tape
and a scanner that would read symbols
that were on the tape and would write other symbols.
And there was also a little sort of memory device
inside the scanner that would...
So if the last symbol that it had seen was a 2,
it should go into a particular state
and if the last symbol that it had seen was a 3,
it should go into another state and so on.
So the scanner could carry along with it
a record of what it had previously seen on the tape,
a kind of memory.
Turing was giving a graduate course
on mathematical logic at Cambridge.
I guess this is early 1939
and Wittgenstein, who was in a different college,
he was in Trinity College,
he was giving a graduate course on the foundations of mathematics
and Turing attended Wittgenstein's course.
They must have been some of the most interesting
philosophy lectures that have ever occurred.
Not just Turing and Wittgenstein
but many names that are now famous philosophers
but in those days just graduate students.
Norman Malcolm, for example,
Casimir Louis, many great names.
And Wittgenstein was an unusual lecturer.
He would say things like,
I just feel too stupid today
to explain what I mean
and I just can't begin.
So there was an edginess to his lectures.
A lot of the lectures consisted of
discussion, back and forth discussion
between Turing and Wittgenstein.
It's clear that as far as Wittgenstein was concerned
Turing was the main person there in the audience.
It was Turing by and large
that Wittgenstein wanted to talk to.
And so in the notes of the lectures
which were reconstructed and published by Cora Diamond
a few years ago, there are these wonderful pages of dialogue
between Turing and Wittgenstein.
It's some of the best records that we've got
of Turing actually speaking.
And they covered the field of
fascinating mathematical ideas.
There was one group of lectures about
contradictions in mathematics.
Bertrand Russell had put the cat among the pigeons
a few years later by proving that set theory
which is this fundamental area of mathematics
was inconsistent as it was formulated then.
You could prove contradictions in set theory.
And so as far as mathematicians knew
there might be contradictions all over mathematics.
And many people who worked
in the foundations of mathematics
viewed this as a terrible danger
because these contradictions could kind of spread
to the world beyond mathematics, the real world
and might cause real damage in the real world.
Turing says at one point to Wittgenstein
well look, the bridges might fall down
if there are contradictions in mathematics.
Engineers would use the mathematics and the inconsistencies
would show up in the calculations going wrong
and the bridges collapsing.
And Wittgenstein sort of gently points out
that well nothing of that nature has ever actually happened.
So Wittgenstein is saying maybe this panic
about inconsistencies in mathematics is misplaced.
It doesn't actually lead us into any difficulties.
And then there is another
absorbing group of lectures
where Wittgenstein is talking about
experimental mathematics.
The possibility of doing experiments in mathematics.
And Turing disagrees with Wittgenstein.
In fact they mostly disagree.
Although Wittgenstein tries to reassure his class
he says actually
Turing doesn't disagree with a word that I'm saying.
It's just the ideas underlying what I'm saying
that make him uncomfortable.
I can't remember the exact words
but Wittgenstein says something like
Turing is afraid that I'm introducing Bolshevism into mathematics
but this is far from the case.
But anyway Turing disagreed with Wittgenstein
about experimental mathematics.
Turing was quite pro the idea of doing experiments in mathematics.
Wittgenstein thought he just couldn't.
He said mathematics is petrified.
It's been elevated above the realm
at which experimentation is possible.
Mathematics has just sort of frozen into place
and no amount of experimentation that you could do would change it.
And Wittgenstein said
if only I could explain my ideas clearly enough
then Turing would come to agree with me
that experiments in mathematics are impossible.
Fortunately Wittgenstein never did manage to make his ideas clear enough.
And Turing went on to pioneer
computer assisted experimental mathematics
in Manchester after the war
at the consul of the Manchester computer
the first stored program electronic computer ever to be built
the first universal Turing machine in electronic hardware
that the world saw.
And Turing sat at the consul pioneering
exactly what Wittgenstein had said was impossible
experimental mathematics.
The Turing test is the most famous
aspect of this paper.
We now call it the Turing test.
Turing called it the imitation game.
Turing was pioneering artificial intelligence at this time
so we're around 1950
that was the date that he published his paper.
The kind of myth that you read
in computer science textbooks
and newspaper articles and so on about artificial intelligence
is that it was invented in this country, America
in 1956
and that artificial intelligence made its first appearance
with a computer program
that was written over in Santa Monica at the Rand Corporation
called the Logic Theorist in 1955
and this program was
first publicly demonstrated up at Dartmouth
in 1956 at a conference
that John McCarthy and Claude Shannon
two great names in the history of computer science
they'd arranged this conference at Dartmouth College
and it was called the Dartmouth
Summer Research Project on Artificial Intelligence
and this was the birth event of
artificial intelligence according to the folk myth.
Now it's certainly true that the name
for the field artificial intelligence
originated with the title of that project at Dartmouth
but the actual field had been in existence
for more than a decade previously
under Turing's name, machine intelligence.
Turing first glimpsed the possibility
of machine intelligence at Bletchley Park
when he was working on his so-called BOM
a machine that he built for decrypting enigma messages
and the BOM worked by what we would now call
heuristic search
and heuristic search was the mainstay of artificial intelligence
for the first decades of the subject's existence
and heuristic search is still an important concept in AI
and there was Turing implementing
heuristic search on his high-speed computing machine
the BOM, not a stored program computer
not a universal Turing machine
but nevertheless a computing machine
it was carrying out systematic procedures that could be done
by a human calculating clerk, by a human computer
although too slowly to be any use
you couldn't do these procedures with paper and pencil quickly enough
to actually get the intelligence out of the messages
in time for the intelligence to be useful
so Turing had devised this high-speed computing machine
which was based on relays
and he was running heuristic searches on this machine
to crack enigma messages
the first appearance of machine implemented heuristic search
and so Turing glimpses the idea
through the BOM
of building intelligent machines
that search heuristically for problems
of a much more wide and varied nature
than just finding the key settings to enigma messages
so if anywhere has a right to the title
of the birthplace of artificial intelligence
it's Bletchley Park, not Dartmouth College
more than a decade earlier
so Turing had been working on artificial intelligence
since the early 1940s
and in the 1950 paper
he published his test
for whether a machine is thinking
whether a computer is thinking
he was a bit cautious about saying that passing his test
would show that the machine was thinking
and at one point in a BBC radio broadcast
that he made
he said we should probably avoid begging the question
by saying that a machine that passes the test is thinking
and it's better to call them grade A machines when they pass the test
and he said that the question
can machines pass the test
is not the same as do machines think
but he says it's near enough for our purposes
so he's putting forward this imitation game
we now call it the Turing test
as a way of telling whether a machine thinks
or as a way of telling whether it's a grade A machine
and the idea is that a judge or interrogator
is faced by in two closed rooms
a computer and another human being
and the judge is supposed to try and figure out
by just question and answer
which of the contestants is the computer and which is the human being
and if the judge can't tell
then the computer is a grade A machine
a thinking machine
ever since Turing published
Computing Machinery and Intelligence in 1950
it's been hotly debated
the idea of the imitation game, the Turing test
and whether it is a way of showing that computers think
the most famous objection nowadays is due to
the Californian philosopher John Searle
who's put forward an argument that's known as the Chinese room argument
if Searle's Chinese room argument works
then it shows that the kind of processing
that happens inside a computer can never add up to thinking
all the computer is doing
really is shuffling symbols
shuffling binary digits around, moving them from one memory location
to another or performing basic computational operations
on strings of binary digits
manipulating strings of binary digits
is insufficient for thinking
Searle says that his Chinese room example
shows, he thinks, that someone could
pass the Turing test
just by manipulating Chinese characters
in accordance with a computer program
written out in many volumes of a rule book
that this person in the Chinese room can read
the rule book is in English
the Turing test is being carried out in Chinese
and the person in the room is able to
pass the Turing test as a result of following the rules in the rule book
and manipulating the Chinese symbols
the questions that the experimenters pass into the room
the person in the room manipulates the symbols
and produces intelligible Chinese answers
in Chinese symbols to the questions
and these are the output
so to the people outside the room
it looks as though the room is passing the Turing test
it's producing intelligent responses to their questions
but the person in the room is doing all this
just by following rules for manipulating Chinese symbols
and this guy has no idea
what the Chinese symbols mean
Searle says as far as the guy in the room is concerned
it's all just squiggles and squoggles
he doesn't even know that they're symbols of a language
so if Searle's right, Turing's wrong
because something that doesn't think
doesn't understand the questions and answers
can pass the Turing test
so the Turing test can't possibly be a diagnostic for thought
if something that doesn't think can pass it
so if Searle's right, Turing's wrong
it's an open debate
in my view, Turing's right, Searle's wrong
but probably this isn't the place to go into that
Turing took up residence at Bletchley Park
on the first full day of the war
he'd been working on Enigma for a few months
before war broke out
but from September 1939
it was a full-time job
it occupied the majority of his energies
through until 1945
Enigma was the German code generated by the Enigma machine
which was a typewriter-like cipher machine
it was old technology by the time the war broke out
Enigma was marketed commercially in the early 1920s
and during the 30s the German military
added some bells and whistles
to improve the security of the machine
but it wasn't leading-edge technology by the time war broke out
and there were other cipher machines that Turing attacked
that were much more interesting than Enigma
and probably much more significant
in terms of the intelligence that they delivered
he also worked on Japanese cipher machines
not too much is known about this
but there is clear evidence in some of the reports that he wrote
that he did know about
the attack on Japanese cipher machines
and he knew about some particular machines
that had been built at Bletchley Park
to attack the Japanese naval cipher JN25
from which it can be deduced
that Turing himself was involved in that attack
because nobody was told anything that they didn't need to know
at Bletchley Park
so if he'd been told about it then he was involved
no doubt the details of Turing's attack on the Japanese codes
will emerge over the years with further declassifications
but at the moment not too much is known
Turing built his bomb
the first milestone on the road to modern artificial intelligence
in my view anyway
he built his bomb to attack Enigma
and his bombs turned Bletchley Park into a code-breaking factory
by 1943
they were breaking two Enigma messages every minute
by means of Turing's bombs
there were teams of bomb operators
working in shifts
all of them women
and living in horrendous domestic situations
small rooms full of bunks
peeling wallpaper, damp, bad food
and they worked long hours on the bomb
and they were bused in and out of Bletchley Park for their shifts
so there was that side of Turing's contribution to the code-breaking
the other side
was his attack on the German Tunny machine
which was a thoroughly up-to-date cipher machine
which first went on air in 1941
unlike Enigma
it was completely up-to-the-minute technology
Enigma was slow and clumsy to use
it required a staggering total of six operators
in order to cipher and then decipher a message
you needed three operators at the sending machine
and another three operators at the receiving machine
the Tunny machine only required one operator
at each end of the link
a sending operator and a receiving operator
with Enigma, the actual ciphertext
the coded form of the message
lit up at the so-called lamp board of the machine
it lit up letter by letter
and one of the operators had to note down the letters of the ciphertext
one by one as they lit up
a painstaking process
and then the operator passed the ciphertext
to the radio operator, the third player
who then translated the letters of the ciphertext
into Morse code and transmitted it by radio
to the receivers
there was nothing like that in Tunny
under normal operating conditions
the sending and receiving operator
never even saw the ciphertext
everything was automatic
as far as they were concerned, German went in at one end
and German came out at the other end
it was a much more sophisticated machine than Enigma
and in 1941 the British knew nothing about it
they knew all the details of the Enigma machine
because machines had been captured
and they had taken them back to Bletchley Park and examined them
but no Tunny machine was captured
until the very end of the war
but nevertheless, astonishingly
they were able to reverse engineer the Tunny machine
without ever having set eyes on one
and Turing contributed the first method
for reading the daily Tunny traffic
there are two steps to breaking a new cipher
first of all you need to work out the details of the machine
that produces the cipher
and Turing wasn't involved in that
the second step was that once you know the details of the machine
you need to work out a method for actually reading the daily messages
that you intercept
and that's what Turing did
he worked out the first method for breaking the Tunny traffic
the reason that was so important was that
unlike Enigma, Tunny carried just the highest level traffic
messages between Hitler
and members of the German army high command in Berlin
to the generals at the front lines
so Hitler and the front line generals
would be sort of chatting about the situation
and about strategy
about the strength of the enemy
about what divisions the Germans had in the area
about what they were going to do
was it going to be a pincer movement, were they going to go here
or perhaps Hitler thought they should attack some other town first
so there was all this to and fro dialogue on Tunny
once Turing was into it, it was as though the code breakers
could listen to what the top German players
were saying to each other over breakfast
it made a tremendous difference to the war
and saved countless numbers of lives

