word ‘understand’ is used. If you ask the people around you what it means
you’ll stump many of them. That’s because understanding has two very
different meanings. Most people don’t separate these meanings but the
distinction is important. Understanding means to decode information,
to comprehend – but, more importantly, it also means to absorb and
internalize information. That feeling you have when you ‘get it’.
If I say, “I understand” I mean I have taken in the question you
asked and decoded it into ideas so I can provide an answer. This can be
quite a mechanical process and computers routinely understand natural
language and answer questions – Apple’s digital assistant Siri being a
case in point.
When I say, “I understand a problem” or “understand a culture”
I mean something far less tangible. Somehow the information I have
gathered over my life is formed into a matrix within my brain that allows
me to ponder and run scenarios. I can predict the effects of my actions
before I do them, and often anticipate your responses. That’s clearly
a very useful evolutionary adaption, but is there more to it than that?
Roger Penrose and David Deutsch think understanding allows us to
transfer non-symbolic information from one brain to another. We don’t
run programs in our brains, nor do we store precise information such as
lists and tables. We have, therefore, had to evolve a creative approach to
communicating skills and understanding each other. One of the most
closely studied areas in the field of communication is when it breaks
down in the lead up to a disaster.
Understanding
“The human mind tends to look
for clear linear relationships, we
like solutions that are close to the
problem in time and space and
make sense when we think about
it quickly, unfortunately, those
simple solutions are usually
wrong and come from acting on
a complex system as if it was a
simple one.”
Brett Piersen
57
Gettysburg Address, Peter Norvig
Space Shuttle Columbia Crew Photo
“For a successful technology,
reality must take precedence
over public relations, for Nature
cannot be fooled.”
Richard Feynman
Bad Understanding
Can Kill
On January 16, 2003, at 3:39pm, the Columbia space shuttle took
off from Cape Canaveral. During the launch a small piece of
foam insulation broke off the fuel tank and hit the shuttlecraft.
The event was recorded on a few low-resolution video frames. They
show a tiny white object hitting the shuttle and a plume of dusty material
splattering outward. The shuttle made it safely into orbit and for two
weeks engineers on the ground debated what to do. In the end, it was
decided the risk was minimal and the shuttle could safely return to Earth.
On reentry, the shuttle disintegrated, killing seven astronauts.
NASA managers had decided the shuttle was undamaged based
on a series of presentations by the engineers. One image in particular
analyzed the potential
damage to the shuttle’s
tiles from an impact.
Read the slide, look
at the key frames, and
decide for yourself what
action you would have
taken.
Shuttle Tile
60 Are the Androids Dreaming Yet?
NASA Internal Slide
WHAT DO YOU UNDERSTAND FROM THE SLIDE?
Some images of the launch are shown on the right
ddd
Here is what you should have understood from the slide: tiles are
really tough but if the foam dislodged from the fuel tank broke through
the outer coating it would cause significant damage. The estimated speed
of the foam hitting the tile was 640 times greater than anything previously
tested. Worried?
Is this a proper understanding of the problem? You have the slide
and the images. Take another look and think hard. If you want, you can
check a video of a similar launch on YouTube to get a feel for the scale of
things, but the still frames shown all the information you need to make
your conclusion.
Understanding
61
Photographs of the Foam Impact from Video Footage
Frame Showing Foam Dislodging
62 Are the Androids Dreaming Yet?
Still from Ground Camera
LOOK AT THE IMAGES, WHAT HAPPENED?
ddd
The truth is you simply don’t know. If you are puzzling over the
strength of tiles, you have been misdirected. There is video footage of
some sort of impact on a wing mostly covered in white tiles, and a slide
describing the effect of a benign sounding ‘foam’ hitting those tiles. But
what is the evidence for an impact on a tile? The shuttle is certainly not
made entirely from tiles; I can see a window in the picture. You should
instead be asking more questions, “What happened?” “What hit what?”
and “How bad is that?”
It was bad. The foam, a very tough material, had hit the leading edge
of the wing, a weak point, punching a hole through it. The wing failed
on reentry and tore the shuttle apart. Clearly, a full discussion of the
possibilities did not occur amongst the shuttle team, or perhaps it only
happened amongst the engineers in private. Once the analysis was tidied
up and presented to ‘management’ it was a one-way communication of
the conclusions, not a discussion of the underlying ambiguous thought
process. The result: people passively listened to the information rather
Understanding
63
than interactively understanding it and agreed on the recommendation
that it was safe to return. Clearly they did not understand the ambiguity
otherwise they would have realized they did not have enough information
to form a conclusion. This is the tragedy of lack of understanding. If they
had known how little they knew, they could have deployed a spy satellite
to take pictures of the damage – one was available nearby and would
have taken a few hours to re-task – but they did not.
Ed Tufte served on the second shuttle disaster commission and
provided an analysis of the disaster. He views slides as a poor medium for
communicating complex problems and thinks documents are far better.
The danger with slides is they force you to simplify information in a way
that destroys the essence of the information. His analysis of the failure of
communication at NASA formed a major part of the final report on the
disaster. Later he coined the paraphrase “All Power corrupts; PowerPoint
corrupts absolutely.” Good communication benefits from stories and
narrative, not bullet points and graphic fluff. Instead of using bullet
points, speak! After all, we have evolved for 250,000 years to understand
language, but only 25 to read PowerPoints. ’
If you write presentations, Ed Tufte’s book The Cognitive Style
of PowerPoint is compulsory reading. He argues that much of the
information you want to communicate is complex and interconnected.
PowerPoint or any similar presentation software encourages you to
simplify it into hierarchical bullets. The format implies simple causal
relationships where none exists. This is dangerous. Communication
should convey understanding – which is very important – and not just
information. What, you ask, is the difference?
Searle’s Chinese Room
“The hardest thing to understand
is why we can understand
anything at all.”
Albert Einstein
The Imitation Game
As an experiment, I am going to ask a student to spend a week in
a locked room. The room is perfectly nice; it has a bed, a light, a
desk, some reading matter, oh, and we’ll give him some washing
facilities too! Every now and then I post some food under the door to
keep him going, Pop-tarts and pizza (thin-crust) work well.
On the first evening a note is pushed under his door with a symbol
on it. The student puzzles for a while, then opens the book sitting on
the desk. The book says, “If you get a piece of paper with symbols on it
look them up and follow the instructions.” He looks up the symbols and
the entry in the book says, “Go to page 44, write down the third symbol
on a piece of paper then post it back under the door.” He follows the
instruction and is rewarded with another piece of paper, this time with a
larger set of symbols on it. Again he follows the instructions in the book
and posts his answer back under the door. This goes on for several days.
He is somewhat bemused, but it passes the time, and he diligently looks
up the symbols and performs all the complicated actions as instructed.
Meanwhile, I meet our new Chinese graduate student and explain
to her she needs to interview a potential translator for the department.
He has just come in from Hong Kong and there is a health scare, so we
have quarantined him in the lab room. He is bored and I have some
paper for writing messages. She writes “hello” in Chinese on a piece of
paper and posts it under the door.
66 Are the Androids Dreaming Yet?
The exchange of notes goes on for a few days and the two seem to be
getting on well. There is even a little romance in the air. When the week is
over I open the door and the two meet. The graduate student says, “Hello.
It’s nice to finally meet you in person.” The man is puzzled because, of
course, she has spoken to him in Chinese. He knows no Chinese.
“I’m terribly sorry, but I don’t speak Chinese,” he says.
She is puzzled, “But I spoke with you this last week!”
“No, I really don’t speak it,” he says.
And, of course, he is telling the truth. The book he has been using
contains the rules for answering questions in Chinese, but he has
absolutely no knowledge of the language. I’ll leave to your imagination
whether the two strike up a real relationship and live happily ever after.
This is the Story of the Chinese Room. The setup is able to fool
someone into believing there is a Chinese speaking person in the room,
yet there is not. Where does the understanding of Chinese lie? The man
definitely does not understand Chinese. And the book clearly does not
understand Chinese because it is an inanimate object. Yet the person
outside the room is convinced she is communicating with a Chinese
speaker. The analogy to a computer is clear. The book is software and
the man blindly following instructions is the hardware. John Searle,
who devised the thought experiment uses it to show computers can
never understand because there is no place in a mechanistic system for
understanding to exist.
The Chinese Room has sparked huge argument in philosophical
circles; let me boil it down to its simplest form. First, let’s refute Searle’s
position with the ‘System Argument’.
The man plus the book form a system. Systems understand; their
individual components do not. My blood does not understand. My brain
without blood would not understand – it would be dead! Plug my brain
into a good supply of blood; add a dash of glucose, and it will understand
the most complex of things.
The systems argument is elegant and most scientists think this is the
definitive argument against Searle, but Searle has a neat way to counter
it. “Imagine”, he says, “that the man memorizes the book and leaves the
room. Now there is no system, there is just the man, but the man still does
not understand Chinese; he is just parroting rote-memorized words and
rules.” Computers, Searle argues, process syntax – the rules of language;
humans understand semantics – the contextual meaning of language.
Artificial Intelligence (AI) proponents hate the Searle argument.
They believe the memorization of a set of words and rules is exactly what
gives us knowledge of Chinese. That is why we go to school!
Understanding
67
A key problem posed by Searle’s Chinese Room is whether you can
know everything about a situation from just looking at the inputs and
outputs. This is very similar to the restriction posed by the Turing Test.
In that case if we were to trace the wire from our computer terminal
to the other room we would either find a human typing messages or a
large box covered in flashing lights. This would definitively answer the
question whether we were talking to a man or a machine. Similarly, if
we opened the door to the Chinese Room we would immediately know
whether there was a real Chinese speaker in there or not. But opening
the door on both tests misses the point. The question asks, “if the inputs
and outputs are the same does it matter what is really going on inside a
closed system?”
Black Boxes
Experiments involving closed systems are known as Black Box
experiments. They presume you can learn everything about the inner
workings of a box simply by probing it from the outside. Young electronic
engineers are often given black boxes as a test. Electronic components
hidden in the box are connected to three external terminals on the
outside. The student is asked to deduce what is in the box using only
an electric meter to probe those terminals. Here are a few examples of
the possible contents of a black box. They would all show up identically
on the student’s meter. Although internally different they are externally
identical. Even my ‘silly’ fourth choice with a cat in the box does not give
Black Box Equivalence
68 Are the Androids Dreaming Yet?
itself away if all you have to go on are electrical readings. (I dare say the
cat would make its displeasure know if left in there for any time.) The
contents are, therefore, said to be black box equivalent.
The reason for teaching engineers about black boxes is to help them
understand how to simplify things. We could construct option four, with
a cat and some food, but it would cost a great deal of money. Option 1 is
functionally identical from an electrical point of view, but for a fraction
of the cost. Steve Wozniak and Steve Jobs were so successful when they
started Apple because Wozniak was brilliant at simplifying logic circuits.
He could take a design with thirty chips and come back with a black box
equivalent solution using only five. It was a fraction of the cost and far
more reliable.
Scientists put great store in black box equivalence because of
a principle called Occam’s Razor. William of Occam was an English
Franciscan friar living in the fourteenth century. He proposed the idea
of minimal explanation. It states that, ‘among competing hypotheses,
the hypothesis with the fewest assumptions should be selected’. When
trying to explain the workings of a black box, the more complicated
inner workings should be discarded, as they have no externally verifiable
effect over the simpler mechanism. Our extraneous animal must be
eliminated! Sorry.
Ironically, given his calling, Occam’s Razor is sometimes wheeled
out as a disproof of the existence of God. Surely God is a complication
unnecessary to the explanation of our Universe. The argument is
illustrated beautifully in Carl Sagan’s book Contact and the film of the
same name. God gets the last laugh in Sagan’s book when the difficulty
with Occam’s Razor is brought into sharp focus. Occam’s Razor contains
an inherent paradox. At any moment in time we only have evidence to
support the simplest of explanations, yet we know many of these simple
explanations are incomplete. We regularly discover new phenomenon –
dark matter and dark energy being some recent examples. If we stopped
discovering new things, Occam’s Razor would be a good way to simplify
our thoughts. Occam’s Razor is a useful intellectual tool to prevent us
over complicating explanations, but there will often be explanations that
are correct, but for which there is not yet any observed effect.
If we go back to our black box example, we see the flaw in concluding
the boxes are identical from examining only their inputs and outputs.
Opening them would clearly show they are not identical! But, how would
this fact reveal itself if they remain closed? The answer is: over time. If
something in the box has memory or understanding, it could present
one set of results for a while and a completely different set of results later.
Understanding
69
In my trivial example, the cat could eat a wire and change the operation
of the black box. Now there is an open circuit where none existed before.
If this happened, the output would change and we would need a new
theory to explain it. If the circuit was attached to a missile control system
or a life support system, you would really want a full understanding
without waiting. It’s humans nature to try to open black boxes. This is
what MRI scans, X-rays, particle accelerators and all our other tools
of scientific investigation are for. We want to open all the black boxes
of nature and see what is going on inside: simply waiting to see what
happens is not acceptable.
In a sense, we live in a black box. We experience the world through
our senses, seeing with our eyes and feeling with our hands. The brain
never directly experiences anything; it only infers the likelihood of
Scene from The Miracle Worker. Helen Keller
pictured at the moment she understood language.
70 Are the Androids Dreaming Yet?
something from the signals it receives. This is similar to our engineer
probing the terminals of the circuit of a black box. How can we know our
experience of the world is real?
Understanding the World
The French philosopher Descartes gave us an explanation for this
paradox. He spent a long time looking skeptically at everything we
perceive. For example, when we poke a stick into a pond, the surface of
the water bends light and the stick appears to have a kink in it. Our eyes
tell us the stick is bent, but our brain ‘knows’ the stick is straight: it’s an
illusion. Descartes wondered if something so simple could be an illusion,
perhaps the whole of our experience is too.
His eventual solution underpins much of modern philosophy – ‘I
think therefore I am’, cogito ergo sum. Even if we doubt everything else,
we cannot doubt we are thinking about this doubt. At least we can rely
upon the existence of this ‘thought’ as some reality. Descartes built up
from this bedrock the real world we live in. We can be sure we experience
things and can apply logic and use thought. We can use this intellectual
faculty to tell a great deal about our Universe.
True Understanding
In the QED lecture series, The Strange Thing about Light and Matter,
Richard Feynman relates the story of the ancient Mayan astronomers.
3000 years ago they were able to predict the motion of Venus in the sky
using only pebbles. They had a simple system that could predict when
the planet would rise over the horizon. Put a stone in the jar every day,
take out a stone once a week, add a stone at every new moon. If the
number of stones in the jar is divisible by 23, Venus will rise. I’m making
up the details but you see the idea... It’s a very simple algorithm. What
should we conclude if the Mayans had perfected their calculations to
predict the motion of Venus and it proved reliable over a whole century?
Would this constitute understanding?
Feynman would say no: the Mayan understanding was not
complete. It was only black box equivalent to our modern understanding
over a limited period. We known that once the Sun begins to run out
of fuel it will swell to a red giant and explode, destroying Venus and
the Earth. Their model could not predict this catastrophic failure. Our
modern deeper understanding of the workings of the solar system allows
Understanding
71
us to predict this future even though there is no clue from the motion of
Venus today. Understanding allows us to predict discontinuous events: a
system changing its state or a star running out of fuel.
We see the same predicament in stock markets. Stock markets
normally behave in a linear fashion but, when they go wrong; they go
very wrong. Recent recessions have been made much worse by the failure
of hedging systems to handle market disruption. Some even think the
crises were caused by the automatic trading strategies of these hedging
systems.
The quants – as mathematicians in banks are called – spend
considerable effort modeling financial instruments to show that if one
stock goes down, another will go up at the same time. If the stocks
are held together your investment is safe because, on average they will
remain constant. The problem with these correlations, which often hold
reliably for many years, is that when trouble hits they fall apart. Historical
correlations don’t give us understanding of the future: something that
was only meant to happen once in a million years has happened within
six months. As they say on your investment papers, past performance is
no predictor of future results.
Do Computers Understand?
Today’s computers don’t have our general-purpose ability to understand.
Watson was thrown off by badly formatted English. The human
contestants, by contrast, had no problem with this. Just how good would
Watson have to be, to call it – or should I say ‘him’ – intelligent? How
could I judge this had happened? Alan Turing proposed an ingenious
test in his 1950 paper Computing Machinery and Intelligence using ‘The
Imitation Game.’ We now call the Turing Test.
If we ask a series of questions to a computer and we cannot tell its
responses from those a human would give, then the computer is, for all
practical purposes, the same as a human. Since we are intelligent – or at
least we hope we are – the computer must also be intelligent. QED.
That’s all there is to the Turing Test. Puzzled? Let’s pick his argument
apart.
Imagine you are chatting away on Facebook with someone you
don’t know. They may have posted a photograph so you can see what they
look like. The photo might be a fake; you have no real way to tell. What
question would you ask the other ‘person’ to prove they were human and
not a computer? There are obviously some giveaway questions. Please
multiply the numbers 342,321 and 23,294 and give me the answer. This
72 Are the Androids Dreaming Yet?
would be very hard for a human but easy for a computer. If you got a
very quick answer; the computer would have given itself away. But, the
computer has been programmed not to give itself away, and it is free to
give the answer slowly or even reply that the calculation is too hard. Our
computer can say anything it likes, including lying to pass the test! If the
computer can fool a questioner into believing it is a human then Turing
argued the computer has shown it is at least as intelligent as we are.
It used to be assumed that the field of broad general knowledge
would be hard for a computer, but Watson has shown this is not so. With
enough storage and a reasonable algorithm, winning a pub quiz is well
within the capability of a modern computer.
The really difficult questions for a computer are philosophical ones,
novel questions and things that don’t fall into a pattern. For example,
“Are you happy?”
“What do you think of Shakespeare’s Hamlet?”
“Is there life after death?”
“How went it?”
“Think Differ…”
If a computer could plausibly answer this sort of questioning for an
extended period, say fifteen minutes, should we conclude it is intelligent,
or do we need more time to be certain?
Turing’s approach to certainty was simple. Just ask lots of questions.
As you ask more and more questions, you will become increasingly
certain you are talking to an intelligent being. He characterized it
as a linear process; after 15 minutes of questioning you might be
99% certain and after a few hours 99.9% certain and after a few days
completely certain. The problem with this approach is it does not flush
out discontinuities. What if the questioning suddenly stopped without
warning or explanation? A human responder is likely to worry that the
questioner has had a heart attack and do something to find out what is
going on including leaving the room. Humans can make creative leaps,
solve non-computable puzzles or come up with a clever new joke. A
humans could even announce the test is a waste of time and walk off.
They just exercised free will! A computer cannot do these things.
Each year a group of scientists enters a competition run by
Cambridge University to win the Loebner prize, a competition to see
how close a machine can come to passing the Turing Test. If you can beat
the test you win $100,000. So far no one has come close and scientists are
beginning to realize just how hard it is.
Understanding
73
New Yorker Cartoon
With the anonymity the Internet provides we can imagine all sorts
of strange scenarios if the Turing test could be passed. You would have
no way of knowing what you were talking to. The New Yorker ran a
cartoon back in 2000. “On the Internet no one knows you are a dog.” We
come across a similar problem the other way around when we encounter
bad customer support. A few years ago, while trying to get an answer
to a computer problem, I became convinced the thing responding to
my emails was a machine. The company did use machine responder
technology so it could well have been. I asked it to prove it was human
by putting the word marmalade into an English sentence and fixing my
74 Are the Androids Dreaming Yet?
problem. The human pretending to be a machine saw the joke, fixed my
problem and replied “Marmalade is served with butter and toast.” The
test worked!
Uncannily not Human
Understanding
75
The sister test in robotics is equally hard. The goal is to simulate the
physical human form, its movements and mannerisms. It’s easy to get
close, but close is not good enough. The term ‘Uncanny Valley’ has been
coined to describe the discomfort humans have with something that tries
to simulate a human being but does not quite get there. I think it is part
of the reason Madam Tussaud’s waxworks are so fascinating. Humans
have a love-hate relationship with facsimiles of themselves. They love
the flattery but feel a sense of revulsion at anything that comes too close.
Searle and Turing
In the Turing Test, we limited our senses to the purely symbolic: using
only typed words on a screen. I could break the lock on the door and go
into the room to see what was there.
“Aha!” I would say.
“I can see you’re a computer, I, therefore, know you’ll be good at
sums and bad at creativity.”
But Turing wants us to see if the difference is given away purely
through intellect. He argues there is no way to tell. But if you follow my
argument from chapter 1, there is one way: ask the computer to find a
non-computable solution to a mathematical puzzle. This is, in practice,
a difficult test to pose because it might take a very long time. Twentyfive
billion people have lived on planet Earth during the last 350 years,
and about 5 million of them were mathematicians. None of them was
able to solve the problem posed by Pierre de Fermat until Andrew Wiles
turned up but this is a clear difference between humans and computers.
However long you give a computer it would never be able to solve the
problem.
This creativity test would take centuries to run if non-computable
thought was rare, but I think we see it often – on display even when we
tell jokes. In which case computers and humans should be easy to tell
apart: humans are the funny ones. I am not saying you can’t build a brain;
our brains are physical devices, after all. I just believe a computer or a
mechanistic machine, cannot think like a human being.
I like the Searle argument but qualitative arguments are insufficient.
We need a quantitative argument. In the forthcoming chapters, I am
going to look at the mathematical argument underlying the difference
between human intelligence and computer processing. Before we do this
let’s take one last look at a qualitative difference; the way computers and
humans communicate.

Chapter 3
BODY LANGUAGE
& BANTER
Body Language
“England and America are
two countries separated by a
common language.”
George Bernard Shaw
“I speak two languages, Body and
English.”
Mae West
“The body never lies.”
Martha Graham
In the summer of 1986 Ronald Reagan and Mikael Gorbachev met
in person for their second negotiation session, this time at the
Höfði House in Reykjavik. For five days, the leaders talked alone
except for interpreters. Reagan badly wanted to develop the Strategic
Defense Initiative; known by its nickname, ‘Star Wars’. The idea was to
put smart weaponry in space that could destroy ballistic missiles before
they reentered the atmosphere. Reagan believed this would remove the
threat of imminent destruction that had hung over the world since 1945.
Gorbachev, on the other hand, felt this was just another escalation in
the Cold War, and the Soviet Union would be forced to build yet more
weapons to overcome the American defenses. He wanted Reagan’s plans
shelved, arguing that it broke the Anti-Ballistic Missile Treaty. He was
probably right. The leaders talked back and forth, unable to overcome
the impasse. At the end of the summit there was a mad scramble to
announce some sort of deal, but this proved difficult. In the last moments
before they had to conclude a communiqué, Reagan suggested they
abolish all nuclear weapons. Reagan’s negotiating team was horrified and
shut the door.
For decades, the American strategy had been to use nuclear
weapons as a deterrent against the apparent numerical advantage of the
Soviets. In all the potential scenarios analyzed by the Pentagon, Russian
forces ended up overrunning American forward positions – otherwise
known as Western Europe! The only way to stop them was through a
release of nuclear weapons, which, inevitably escalated to all-out nuclear
Ronald Reagan and Mikael Gorbachev
80 Are the Androids Dreaming Yet?
war. It was assumed this inevitable progression deterred the aggression
in the first place, and the threat of mutually assured destruction kept the
world peaceful. Giving up this tenet of defense strategy was something
the American military just could not contemplate. Many people did not
think it a rational defense strategy; it seemed appropriate the acronym
for mutually assured destruction is MAD, but this was the status quo.
We now know our worry over Russian superiority was groundless.
The West’s technological advantage, founded on the invention of
computing and sophisticated materials technology, gave us a huge
advantage. In the only battle to be fought in the 20 th century between
Russian and Western tanks, during the first Iraq war, most of the Russian
tanks were destroyed with no losses to American tanks. We know this
now, but we are talking of a time when paranoia over the Soviet advantage
was the common view.
There is speculation that Reagan had muddled intercontinental
ballistic missiles with all nuclear weapons. I do not think this is true.
Reagan was a man of vision, quite comfortable with using his folksy way
to convey sincere belief, and I think abolishing all nuclear weapons was
in his mind. It would have been a breathtaking moment.
In the end a rather feeble communiqué was put together and the
talks declared a technical failure. But, both leaders had seen eye-to-eye;
both were prepared to make major concessions and both wanted an end
to the old strategy of mutually assured destruction. Wiping each other
out was no longer considered a successful outcome! The meeting, and
Höfði House in Reykjavik
Body Language & Banter
81
the fundamental thawing of relations between East and West, was to lead
to the Intermediate-Range Nuclear Forces Treaty and the end of the Cold
War.
Face-to-Face Communication
What really happened between these two leaders when they met and
talked? Was it a mechanical process of offer and counter-offer, as easily
executed by fax, or is human interaction more complex than this?
Reagan, as a young man, had been a liberal, sympathetic to socialist
ideals until a painful strike in California caused him to lose faith in the
politics of the left. Gorbachev, a lifelong Communist, was desperate to
reform the Soviet economy and make it more competitive. He, also, had
come to see the hypocrisies that could emerge in far left-wing ideology.
I don’t believe this common experience could have been communicated
by fax or email. Indeed, I am sure these specific points were never made,
but the nonverbal communication must have conveyed something of
their common background and purpose.
When we phone someone or exchange emails, the interaction is
factual, there is no body language, and we rarely laugh. When we travel
to meet someone, we spend a great deal of time with them. The average
length of a phone call is two and a half minutes, but meetings, especially
when one party has travelled to see the other, can be hours long. When
humans meet they greet each other, shake hands, sit in the same room,
talk at length, and laugh. Body language is important; people mirror
each other’s postures, adopt open and receptive stances, and make
eye contact. You can see this in the picture of Reagan and Gorbachev
above. Body language allows us to convey qualitatively different things,
such as trust and happiness. It is very expressive; you can see the more
guarded postures of Yasser Arafat and Shimon Pérez below, just after
they negotiated a landmark peace deal. Can you tell if the leaders smiles
are false?
Communication
Communication is one of mankind’s greatest expenditures. The US
telephone system is arguably the largest machine on the planet, while
the world’s mobile phone networks have a capital value of $2.5 trillion,
greater by an order of magnitude than all the steel plants in the world put
82 Are the Androids Dreaming Yet?
Yasser Arafat and Shimon Pérez
together. This lifeblood of our existence – long-distance communication
between human beings – turns out to be amazingly difficult, even with
all our clever technology.
In recent years the Internet has, in theory, allowed each and every
person to communicate freely with any other person on the planet.
In some of the most distant parts of the world mobile phones, and
projects such as; ‘One Laptop per Child’ are rapidly bringing unlimited
communication to all. This communication can be personal, one-to-one,
or broadcast: I can talk to people interested in a particular topic directly.
As we watch the Arab world democratize, catalyzed by the Internet, there
is no question that digital communication has now become a major force
in the world. Yet, people don’t communicate over the Internet as much
as you would expect; they often use the Internet to set up phone calls
during which they arrange meetings! This is odd. We have a fantastic
phone system and sophisticated communication technologies; email,
video and instant messaging. Yet, we still choose to travel when we want
to communicate.
On the face of it, there should be no difference between a phone
call and a meeting. In principle the same information can be conveyed.
Yet when we want to really understand someone, we always go to meet
Body Language & Banter
83
Smiles Fake or Real
in person. No great treaty or big industrial contract has been negotiated
without a face-to-face meeting. We see this daily: people talking on the
phone get to a certain point, give up, and arrange to meet in person.
The consequence is that we spend $550 billion annually, flying
around the globe to meet each other. Each day the world’s population
takes three million plane flights. Around 80% of these are business
flights. Some are people emigrating or going to do specific manual tasks,
but most are to have meetings. We have always assumed that this is
because the parties are unable to reach a sufficient level of trust over the
phone and need face-to-face interaction to build that trust, but it may
be that the parties are not able to convey sufficient information to fully
understand each other. Face-to-face meeting may convey much more
information than we think.
84 Are the Androids Dreaming Yet?
Smiles
When we smile naturally we use a full set of facial muscles, including the
muscles around our eyes. When the smile is forced those eye muscles
remain passive and the smile, although superficially the same, is missing
something. You can’t put your finger on it, but the look is insincere. A
study of marriages in the USA analyzed smiles in wedding photographs.
The couples with false smiles divorced much earlier than the genuinely
happy couples. Similarly for high school photos; people with genuine
smiles at 18 years of age were happier later in life and in more stable
relationships. Smiling is really important. It is good to be around people
who smile, they are more successful – and nicer.
There is also a curious reverse effect. The link between our minds
and bodies is much more fundamental than we thought. If you grasp a
pencil between your teeth, it forces you to smile. Try it. The mere act
of smiling is found to make you happier, it causes the release of the
chemicals called endorphins which improve your feeling of well-being.
Micro-expression Analysis
Since the involuntary movements of the muscles around our eyes give
away genuine happiness, a whole science has evolved looking for other
biological cues to mood. The two most interested groups are the FBI,
trying to detect lies, and poker players, trying to make money! Much has
been written on the topic, including a few best sellers, but the evidence
for micro expressions is mixed. Regardless of whether involuntary
actions give away our emotions, humans voluntarily use a great deal of
body language when talking.
Body Language
A study by Albert Mehrabian is often cited to say 93% of the information
in a conversation comes through nonverbal cues. This is misquoted.
The study really stated 93% of the emotional content is nonverbal.
That’s more believable. And further studies have shown when there is
doubt, nonverbal cues win over verbal information every time. The rule
is sometimes laid out as the 7%-38%-55% rule – 7% words, 38% tone
of voice and 55% body language. Remember this is emotional content,
your conviction and sincerity. You will still have to get over the factual
information you want to convey.
Body Language & Banter
85
Learning Swedish with The Two Ronnies
Try this experiment on a friend. Tell them you like their shirt using
different tones of voice: sarcastic, sincere, amazed. Then see what they
understood. You will find it difficult to appear sincere because I have
told you to say you like their shirt – unless of course you really do. When
you use sarcasm they will find it hard to process your statement. It is
revealing how we use the information.
Interestingly, a piece of research described in Scientific American
shows even insincere flattery is effective. If you want a pay rise from your
boss, any form of flattery will do. Vanity appears to override skepticism!
Interaction
The normal cadence of communication between people includes a great
deal of mutual interruption. When a meeting breaks down we often see
people begin to say things like, “Please don’t interrupt me,” “Do you
mind, I was talking,” “Pleeeease, let me finish.” If the meeting is really
getting out of hand, third parties will often step in and tell one to wait for
the other. This is where the mechanics of face-to-face interaction fail, as
we need to interact in order to communicate effectively.
Because we have a lot more time in a face-to-face meeting people
can wander ‘off topic’. This is an important part of the process of
communicating. After all since most phone calls are 2-3 minutes and
86 Are the Androids Dreaming Yet?
most meetings an hour, there are another 57 minutes to fill! These off
topic items bring in social experience and help us form the background
context we need to properly communicate.
What is Background Context?
Alex and Bella are both fans of the British comedy duo, the Two Ronnies,
and enjoy their learning Swedish sketch. Bella asks Alex what kind of
sandwich he wants for lunch. Alex replies ‘M’. Bella laughs. If you have
seen the sketch you will understand the background context to the joke.
If not this paragraph might as well have been in Swedish. Take a look at
the sketch on YouTube and reread this paragraph... Now you understand.
Do I think in English?
Most scientists believe we think thoughts using language, but most
scientists writing about thought are linguists or psychologists. If you are
a dyslexic engineer like me, language is a long way down the processing
chain. I think abstractly and then translate those thoughts into words.
Some ideas don’t map between languages and often, one language adopts
the words of another to fill in the gaps. Some interesting examples are:
Zeitgeist
Schadenfreude
Chutzpah
German, spirit of the times
German, enjoying others misfortune
Hebrew, audacity
All of these are fully signed up, card carrying entries in the Oxford
English Dictionary.
Some languages have fewer distinctions between ideas: truth and
law are the same word, ‘torah’, in Hebrew. Languages have different
tenses and structure. In Chinese all words are one syllable and the script
is pictographic rather than phonetic. This is unusual, even Egyptian
and linear-B, which look pictographic are mostly phonetic. With single
syllable words, Chinese uses voice inflection to change meaning; a rising
or falling tone can change the meaning of a word from ‘grey’ to ‘girl’.
In many Western languages rising voice inflection is used to indicate a
question, as in Australian English or irritation, as in English English. So
how do the Chinese show if they are annoyed or want to ask a question?
They elongate their words and accentuate the changes in intonation.
An argument in Chinese can sound quite alarming to the Western ear,
with its percussive monosyllables and extreme inflection changes. This
Body Language & Banter
87
degree of inflection is used in English, but only in extreme emotional
contexts: A Chinese argument over cold tea can sound like an accusation
of murder to a Western ear.
Symbolic Communication
The earliest recorded permanent human communication is cave
painting, dating to 33,000BCE. Written communication emerged in
Sumer, the southern part of Mesopotamia (now Iraq), using a script
called Cuneiform, written on clay tablets. It was used primarily for
accounting. The Sumerians are responsible for our common use of base
twelve. Twelve hours in a day, inches in a foot, and notes in the scale; all
stem from their civilization.
Although not the first to write stories, the Greeks perfected the
dramatic forms we use today: poetry, prose and plays. Watch an episode
of ‘Law and Order’ and you are seeing a direct descendant of a Greek
tragedy, complete with suffering and justice denied. All this permanent
thought art is made possible by the translation of ideas into symbols.
Scripts and Symbols
The world supports a huge variety of scripts split roughly into phonetic,
representing the component sound of words, and pictographic, stylized
pictures of the ideas.
Chinese Traditional and Simplified
Some scripts have interesting quirks. Ancient Hebrew, although
phonetic, is a script where vowels are omitted. Modern Hebrew often
leaves them out as well. This means words can be ambiguous and need
context to decipher them. A common set of Chinese characters has long
been used by Mandarin, Cantonese, and Japanese speakers even though
88 Are the Androids Dreaming Yet?
their spoken languages are entirely different. The script languages of
these people are gradually diverging and might in time become entirely
separate languages too.
The Chinese government in Beijing has moved to using simplified
Chinese for Mandarin speakers, while Hong Kong continues with the
traditional form. Japanese has developed many new characters for
modern ideas, such as computers, that differ from the Chinese, and mixes
in a great deal of Katakana, a script allowing the phonetic representation
of foreign words. If you walk around these countries their signage looks
quite different, although I am told Cantonese speakers can still read
Body Language & Banter
89
simplified Chinese. Take a look and normally you will find them to be
quite different. Each example in the figure is my best attempt to translate
the phrase “Hello Reader” into a script and the corresponding language.
Symbols of the World
English is one of the most irritating script languages of all. It commonly
uses etymological elements, showing the history or origin of the word
that has nothing to do with the sound of the word. A word like school
has the ‘k’ sound spelt ‘ch’, showing its historical derivation from the
Greek, but confusing for pronunciation. English has 53 sounds derived
from only 26 letters, so there are plenty of letter combinations, many of
which are irregular. Because the language favors historical convention
over simplicity, sugar is pronounced “shu-gar” whereas sand is strictly
phonetic. As for Leicestershire I’ll leave that as a test for the American
readers amongst you. If you’re British, try Mattapoisett, a town in
Massachusetts named in Native American.
Yet English is also a ‘lovely’ language. Because of its richness there
are often twenty different ways to say something, and a dozen words
to choose on any topic. One of my own favorite words is ‘jump’. It is
phonetic, but also onomatopoeic and even pictographic. Jump both
sounds like a jump and looks like a jump.
Two scripts that puzzled scholars for many years are Linear-b and
Hieroglyphics. Linear-b – found on clay tablets on the Island of Crete –
turned out to be a coded form of ancient Greek with some slight quirks,
such as dropping the letter ‘s’ from the ends of words. The ‘s’ is superfluous
in most Greek words, and dropping it saved precious clay space!
Hieroglyphics was a real puzzle. It looks so like a pictographic
language that it fooled many people for centuries. The Rosetta Stone was
discovered in 1799 and became the key to their deciphering. This stone
had the same edict written out in 3 languages – Greek, Egyptian and
Demotic. The French adventurer Jean-François Champollion decoded
hieroglyphics in 1822 and although it looks pictographic, it was found to
be predominantly phonetic. Linear-a, another script found on the Island
of Crete has yet to be decoded and remains one of the world’s greatunsolved
mysteries.
All these different ways to code ideas into symbols present the
children of the world a great learning challenge. Because written language
is so young, in evolutionary terms, our brains have not had enough time
to evolve to master it. Instead words co-opt parts of our brains originally
90 Are the Androids Dreaming Yet?
evolved for different purposes. As languages differ in their construction
they co-opt different bits of the brain. It is possible to see this using brain
imaging.
Dyslexics – and I am one – have difficulty in translating between
the realm of conceptual thought and written script. This translation is
subtly different for each language. Chinese speakers use their motor
cortex to process characters. Young children write out the characters
over and over, to memorize them, so the ‘muscle’ memory is highly
involved. French and Spanish children use the audio pathways, as most
of their language is phonetic, the motor part of writing is then an addon
and does not process meaning. English children must use portions
of their visual cortex to process the meaning of words, as many words
have spelling quirks that have nothing to do with the sound of the words.
Some studies even suggest a child dyslexic in one language, because, for
example, their audio pathway is impaired, might not suffer the condition
in another language that relied on a visual or motor skill.
Can Objects Communicate?
The process of communication has many components, starting with
something capable of communicating. Communication usually –
perhaps always – is something that occurs between sentient beings. I
don’t think of my computer as communicating with me, but rather think
of it as a medium for communication or a dumb machine. But colloquial
language around the subject is a little muddled. We all agree a lighthouse
does not communicate, even though it can signal danger, but what do
we mean when we say, “That song really spoke to me.” No one believes
the song is actually communicating, but some kind of communication
was made nonetheless. When we talk of communication do we mean the
agent or the message?
Stories
Humans enjoy communicating; we create works of art, music and
literature that transcend simple analysis. The COIN dynamic slide,
which we saw earlier detailing the strategic situation in Afghanistan,
would probably have been better communicated with a story. Humans,
unlike computers, do not cope well with large quantities of unrelated
information, and studies of memory and comprehension show we
Body Language & Banter
91
benefit from a narrative structure. Let me give you a basic example. One
simple trick the human brain uses is chunking. Give yourself a moment
to try to learn this string of characters.
HALTNTIBMGTATLAMATLOLPOMSGTG
TRY TO MEMORIZE THE STRING WITHOUT READING ON
Now, if I divide it into chunks, you will see it includes meaningful
information.
HAL TNT IBM GTA TLA MAT LOL POMS GTG
You probably won’t recognize all the acronyms unless you are
under 10. Even then, you will find memorizing it hard, but if you put the
sequence into the context of a story then it is much easier to learn.
HAL uses TNT to blow up the IBM building in Grand Theft Auto.
“Three Letter Acronyms are annoying,” says MAT. I’m Laughing Out
Loud; Parents Over My Shoulder. Got To Go.
We find it easier to fit new information into existing structures
within our brains rather than memorizing by rote. I’ve used quite a
bit of modern Internet slang here. You’ll find young people recall this
information better than older people for whom GTG and POMS are
nonsense.
If you want to memorize something, experts recommend you
imagine bizarre images and relate them to a story pictured in the mind’s
eye. Try it and you may very well find you can still remember my sentence
in ten years time!
Let’s try something else. The following sentences are a little different,
yet the recall scores for information in the two are dramatically different:
1. I met an old tramp on 42 nd Street wearing a dirty grey rain coat.
2. New York on a cold damp November day; as I cross the street
I bump into an old man wearing a dirty grey Macintosh. His
shuffling gait suggests some sordid intent. I think nothing of it,
but this brief meeting was to change my life.
92 Are the Androids Dreaming Yet?
The addition of contextual cues allows you to form a mental picture.
By withholding some information at the end I have used a dramatic trick
to cause your brain to free wheel and imagine what happens next. You
are involved in the story. Notice the longer story, with more data in it, is
paradoxically more comprehensible and memorable.
Ed Tufte makes the point about our ability to process information
very forcefully. He believes presentation experts are wrong when they
recommend you keep your slides to a few words! He points out the
common advice to use only six bullets per slide and six words per bullet
comes from a misconception that has blighted a generation of presenters.
Studies performed on memory in the 1960s measured unrelated word
recall. Six words are all you can remember if the words are meaningless.
But if the words have meaning we can comprehend and absorb many
pages of data. Hundreds of millions of people throughout the world read
a newspaper every morning and can recall the stories throughout the
day; the poems, songs and plays we memorize when young are usually
long, comprising thousands of words, yet we are able to remember them
verbatim for the rest of our lives.
When we tell a story, we are trying to draw the reader in so they
can to experience our imaginary world and be ‘in’ the story. When I read
a story – perhaps Harry Potter – I don’t think about the grammar and
punctuation, or even the accuracy of character portrayal. I’m transported
to a different place. I experience a piece of the reality or ‘imaginality’
the storyteller has created. I can describe the characters, the scene, the
sounds and the smells. A good author forms a complete world in our
heads corresponding with the world they have in their heads. With more
abstract information, comprehension and retention is harder. Often if
the information does not hang together in a linear narrative it can be
impossible to take in at a single sitting. However, if it forms a story and
is well told so you ‘get it’, you do not need it repeated. We experience
something of this effect when we watch a good movie. “I’ve already seen
that one,” means you have absorbed the whole story in a single sitting.
You don’t need to watch it over again to comprehend it.
Comedy
Finally, when you mix all the elements up, emotional understanding,
body language, in-person communication and empathy; you get comedy.
Humans ‘do’ comedy from a very young age and it’s vitally important to
the fabric of our lives. What purpose comedy serves in communication
Body Language & Banter
93
My XBox is Broken
The One Ronnie
Dead Parrot Sketch
Monty Python
Gerald the Gorilla
Not the 9 O’Clock News
Fork Handles
The Two Ronnies
Andre Previn
Morecambe and Wise
Self Defense Against Fruit
Monty Python
is not clear. In life, telling a joke will make another person smile. This
causes people to be happy and happy people release chemicals into their
bloodstream which make them healthier. Happy people then tell jokes
to others. This circular process improves the well-being of communities
and helps bond people together. But why on Earth did comedy evolve to
be the mechanism that does this?
Comedy may be an important way to avoid an argument when
context is unclear. Much of what we say can be taken the wrong way.
Simple communication of fact can sound like criticism or challenge, and
94 Are the Androids Dreaming Yet?
humans are naturally hierarchical – not unlike packs of dogs or beached
walruses. Humor allows us to test the response of others to statements,
which might otherwise be taken the wrong way. Something said in a
‘jokey’ tone of voice may not generate a negative response, even though
the raw content might be quite provocative. “Ah, late again I see…”
It is worth taking a look at some great comedy sketches because
they bring home the richness of human interaction. Here are some of my
favorite links as an antidote to the heavy-duty mathematics I am about
to inflict on you.
The World’s Funniest Joke
Two hunters are out in the woods when one of them collapses. He
doesn’t seem to be breathing and his eyes are glazed. The other guy
whips out his phone and calls the emergency services. He gasps, “My
friend is dead! What can I do?” The operator says, “Calm down. I
can help. First, let’s make sure he’s dead.” There is a silence, then a
gunshot is heard. Back on the phone, the guy says, “OK, now what?”
Spike Milligan, from The Goon Show
I think comedy is a fitness display. It demonstrates to those around
us – particularly of the opposite sex – that we can be creative and use
non- computable thought processes, just as dancing is a fitness display of
our agility and coordination. When we tell a joke we are showing others
we can ‘think outside the box’, a valuable survival skill.
At a simple level it has been proven that animals with the ability to
behave randomly escape being eaten more often than animals that follow
a pattern. Non-computability is the ultimate behavioral randomizer
since it is not an algorithm and cannot be copied. The ability to take
non-computable thinking to its logical conclusion to create and invent
has clearly taken off for humans.
Of course, another explanation might be that making people happy
is fun. People like to be around other fun people so humor encourages
crowds to form. If a saber-toothed tiger attacks you, and you are in a
crowd, you’re more likely to survive. You only have to outrun one
member of the crowd!
Chapter 4
THE BRAIN
Baby EEG
“The brain is a wonderful organ;
it starts working the moment
you get up in the morning and
does not stop until you get into
the office.”
Robert Frost
“The brain looks like nothing
more than a bowl of cold
porridge.”
Alan Turing
Physically the human brain is very boring. Alan Turing described it
as looking like a bowl of cold porridge. To get to the porridge you
must first cut through the skull, a two-millimeter thick protective
layer of bone. The adult human skull has almost no gaps in it, and the
only ways into the brain without a bone saw are through the eye sockets
or the soft area of bone at the back of the nose. Egyptian mummies had
their brains removed through the nose and preserved in a jar for the
afterlife!
Thinking with Porridge
Protecting the brain is very important and the skull does a good job
by being a tough, impenetrable barrier. But sometimes this toughness
backfires. In 2009, Richard Hammond, one of the presenters of the TV
motoring series Top Gear, suffered a crash while testing a land speed
record-breaking car. Although he was in a multipoint harness, the
crash, at over 200 miles per hour, bounced his helmeted head around
the inside of the cockpit and his brain was badly bruised. As you know
from experience, when you bruise you get swelling, and the brain is
no exception. However, the brain is encased in bone, so this swelling
has nowhere to escape. The resulting buildup of pressure is dangerous,
causing an interruption of blood supply to the un-bruised parts. Brain
damage in such accidents is often fatal; Richard Hammond was very
lucky to live through the experience.
Surgeons often need to cut into the skull to relieve pressure on
the brain, or to gain access to remove tumors. Going through the scalp
involves a great deal of blood, but once you have a clean hole in the skull
you can peel back the thin membranes, called the meninges, to reveal a
wrinkly folded whitish thing that looks a bit like a cauliflower. This is the
outer surface of the brain where much of our thinking is done. Unfolded,
this surface layer would cover the area of a football field and this intense
folding distinguishes the human brain from the brains of simpler animals.
Some animals, such as elephants and dolphins, have larger brains than
ours, but the area of their folded surface is considerably smaller. It is
thought that this efficient folding is key to giving us the ability to think
complex thoughts.
Analysis of Einstein’s brain held at Princeton University shows it
is not particularly massive, but it is strikingly more folded than average,
and has a shorter lateral sulcus – the fissure between the front and back
98 Are the Androids Dreaming Yet?
Einstein’s Brain
of the brain. Whether this is related to his highly creative thinking or just
random chance is unknown, but it’s an interesting data point in our quest
to understand creativity and intelligence.
Looking through a microscope, the wrinkly grey matter is composed
of 30 trillion neurons; small whitish cells sprouting filaments that wrap
around each other like the tentacles of an octopus. The tentacles, and
there can be as many as 10,000 per cell, are known as dendrites and spread
out to nearly touch other neurons. At the other end of the neuron is a
single axon. The gaps between the end of an axon and the next neuron’s
dendrites are called synapses, about one-tenth of the width of a human
hair and varied in structure. When a nerve ‘fires’, an electrical pulse
spreads out along the axon to the end and crosses the synapses to other
brain cells. This electrical pulse is not like the flow of current in a wire:
neurons don’t conduct electricity. It is more akin to dominoes falling in
a line. Ion gates in the walls of the neuron open, letting potassium ions
flow out. As the gates open in one section, the next section is triggered
and so on. Thus, electrical signals ripple out along the axon. As the
electrical signals cross the synapses they either excite or inhibit the firing
of adjacent neurons. There is a lot more structure to a neuron than was
once thought. The textbook model is of a sequence of ion sacks stacked
end to end rather like plant cells, but neurons have a far more complex
structure. Bundles of actin and tubulin form a skeleton in the neuron and
the neuron metabolizes ATP to recharge its firing mechanism. Neurons
behave far more like small animals than inanimate plant cells.
The Brain
99
The wiring of our brain looks a bit like the logic circuits of a
computer, and our best guess is the cells in our brain form some kind of
computer. The brain cells – a specialized form of nerve cell – connect to
the rest of the body via the nerve cells that largely run down our spine.
Thoughts trigger action and, in reverse, the nerves in our extremities
sense things in the environment and relay information back to the
brain. If I think, ‘move my finger’ my finger will move, and if it touches
something I will feel the sensation. Interestingly if my finger touches
something hot a reflex will kick in. Reflexes work without involving the
brain. We don’t have to think, “that hurts.” Instead, our finger reflexively
pulls away. We may say ouch, but by the time we do, our fingers already
moved away from the heat.
Nerve cells are much slower than the electronic systems we build
with copper and silicon. This speed is quite noticeable and limits the
rate we can do certain things. It takes around 0.08 seconds for a nerve
impulse to run down to the tips of our fingers, initiate an action and
return to give us the sensation of the action. This may sound fast but if
you’re a tennis player in a rally or a pianist faced with a fast passage, the
nerves don’t have time to make a full round trip signal before the next
action must be initiated. In these instances we need to run on autopilot
and there are parts of the body where the nervous system takes action
without the brain getting involved. This is particularly the case with
things like walking and balance, which must respond fast to changes in
ground conditions. The signals just don’t have time – and don’t need – to
go all the way up to the top of the body for instructions. Rather like the
heat reflex above, the peripheral nervous system can process information
locally. After all, brain cells and nerve cells are really all one type of cell.
If you have a group of people, you can conduct a fun experiment
to show the speed of nerves. Hold hands in a big circle and squeeze the
hand of the person next to you. When they feel you squeeze, they should
squeeze the next person’s hand and so on. The rate at which people
squeeze hands around the circle is limited by the speed at which nerves
conduct the signals across our bodies.
Imaging the Brain
There are several ways to look inside the brain without recourse to a bone
saw. The methods are fascinating in their own right, even before we start
looking at the results. Each image is generated using a different physical
principle.
100 Are the Androids Dreaming Yet?
X-rays
The first Nobel Prize in Physics was awarded to Wilhelm Röntgen in
1901. He had discovered ‘X’ rays; so called because he had no better
name for them. X-rays, as they became known, are just light of a very
high frequency.
Light comes in a variety of colors; at the low end of the frequency
scale we see red, higher up blue and, at the top, violet. At this point
human eyes give up and cannot see anything higher, so ultraviolet
light is invisible to us. Bees, on the other hand, can see a long way into
the ultraviolet spectrum and some flowers have beautiful ultraviolet
markings that attract bees for pollination. Daylight contains a great
deal of ultraviolet light which is wasted on us – other than to tan our
skin. But all is not lost. Clever manufacturers put fluorescent dyes
into their washing powders which stick to our clothes and convert
ultraviolet into visible light, making our T-shirts look brighter as they
reflect more visible light than fell on them. You can see this effect most
easily in a disco when ultraviolet lights are shone on the dance floor
and anyone wearing a newly washed T-shirt will glow bright white. The
other common substance that fluoresces strongly on a dance floor is
tonic water. Quinine, the active ingredient in tonic water, is a strongly
Flowers in Ultraviolet Light
The Brain
101
Pit Viper
fluorescent substance which converts ultraviolet light down into the
visible spectrum. Photoactive dyes have recently become controversial
as suggestions have been made that they are unsafe and irritate the skin.
Going to discos might not be quite as fun in the future!
Thermal Imaging
102 Are the Androids Dreaming Yet?
At the bottom end of the spectrum is infrared light. Pit vipers have
evolved special organs on the sides of their heads to ‘see’ in this spectrum
and they use this sense to hunt prey in the dark. I use the word see with
some caution. We have no idea what their sensation of ‘heat-sight’
involves, but their organs are very precise, able to detect things only 0.2
degrees warmer than the background.
Infrared cues help several species of snakes, bats and insects locate
things in the dark, but the animal that excels at the task, albeit using
technology, is mankind. Special cameras allow us to use infrared to see
in the dark or detect where our houses lose heat.
X-rays are much higher in frequency – about one hundred times
that of the ultraviolet light that affects our T-shirts. The high frequency
corresponds to a small wavelength that allows the rays to pass through
our bodies. Later on in the book we will understand that frequency is not
a proper explanation for light, as it is not a wave but rather a particle that
obeys the laws of a wave. But for now we will ignore this detail.
The first use of X-ray images was to see broken bones. Bones block
the rays as they are dense, but the soft parts of our bodies are almost
completely transparent to X-rays. We can see the soft tissues if we turn
the contrast up, but there are problems when using X-rays to view the
brain. Our skull completely encases the brain and however much we turn
the contrast up, all we see is bone. The solution to this problem is to
perform sophisticated mathematical tricks using a computer to enhance
the contrast ratio and make image ‘slices’ through the living head.
The slicing technique was invented independently in the 1970s
by Sir Godfrey Hounsfield, working for EMI in England, and Allan
Cormack, of Tufts University in America, and they shared the 1979
Nobel Prize for Medicine for their work. Legend has it that EMI was
making so much money from The Beatles they could fund the enormous
development cost of the CAT scanner from the profits; true or not, it’s a
great invention.
The best way to understand the mathematics is to picture yourself
in an episode of ‘CSI’, the American television crime drama. An intruder
has attacked someone with a knife and there are blood spatters all over
the walls of the room. Enter the brilliant pathologist who reconstructs
the scene of the crime from the pattern of blood on the wall. She can map
the trajectory of the blood spatters and back-calculate that the attacker
must have been 5’ 4”, left-handed and wielding a 6” blade. In a CAT scan,
our head is hit with billions of rays that bounce and scatter over the walls
of the machine. Sensors detect the rays and a mathematical algorithm
calculates an image of the body that would produce such a pattern. To
The Brain
103
X-ray of Roentgen’s Wife’s Hand
simplify things we shine the X-rays onto the head as a narrow slit of light
so we only have to do the back calculation in two dimensions. Then we
stitch successive slices together in the computer to form a 3D virtual
image. Thus, doctors can ‘fly’ through the brain looking at structures
such as tumors from all angles.
104 Are the Androids Dreaming Yet?
There are two problems with X-ray imaging. Even with clever
mathematics, the dense bone in the skull blocks the rays so you don’t get
much contrast, making it hard to distinguish normal brain matter from
something like a tumor. But the bigger concern is X-rays are a form of
ionizing radiation, and ionizing radiation causes cancer.
We are told to wear sun block to protect our skin from ultraviolet
light; X-rays are 100 times more potent and can do a great deal of damage.
Fortunately, the body repairs itself quite well in the presence of low levels
of radiation. The double part of the double helix in our DNA allows a
set of proteins in our cells to go around correcting errors when they
detect a mismatch between the two strands. But, now and again an X-ray
might make an irreparable fault in both copies. If enough of these faults
accumulate, they can lead to cancer or, if the errors are in reproductive
organs, birth defects. Doctors try to minimize the radiation we receive
and give us as few CAT scans as possible during our lifetime, especially
when we are young and have not yet had children.
MRI
X-rays dominated our ability to see into the human body until the
mid-1970s when Raymond Damadian came up with the idea of using
magnetism. Magnetic fields are not absorbed by bone and present
no danger as they do not damage DNA. Ironically, the technique was
originally known as Nuclear Magnetic Resonance, ‘NMR’, which patients
thought must be dangerous because of the word nuclear. The name was
Functional MRI: Working Memory
The Brain
105
Diffusion Tensor Image
changed to the one we use today: Magnetic Resonance Imaging, ‘MRI’.
The system works by applying a strong magnetic field to your body to
excite the hydrogen atoms. Since we are mostly H2O there are plenty of
these.
Three magnetic fields are used. First, an extremely strong field is
applied to the whole body. This causes all the hydrogen atoms in the
water and fat to spin in line with the field of the machine. Next a gradient
field is applied to the top of your head so it is slightly more magnetized
than the bottom of your feet and, finally, a pulse of magnetism is applied
to the top of your head. The spinning hydrogen atoms line up a little
more when this pulse is applied and then randomize again when it is
switched off. As they randomize, they give off energy. The clever part is
the gradient field which causes the atoms to give off energy at slightly
different times – the top of your head first, your neck a fraction of a
106 Are the Androids Dreaming Yet?
fMRI
second later, and so on down to your feet. What you see at any one time
is a slice through a specific section of the body. You can then build up 3D
images from these slices and look at the soft watery tissue rather than the
hard bone you can see with an X-ray.
MRI scans give detailed images but today there are many more
imaging tricks you can play. Give the patient gadolinium to eat – a type
of paramagnetic material – and this contrast agent will highlight active
parts of the brain. You can ‘see’ which parts are active: the location of
emotions such as love, joy and even the effect of smells as the brain
experiences things. This is still coarse grained information; it shows only
the general area of excitation and it does not tell us what is going on at
the nerve level, but the images are fascinating.
Another recent development in imaging is the diffusion MRI. If you
remember your school physics, molecules travel with a random walk:
they diffuse along pathways just as people wander along a corridor. If the
corridor is full of people, they are jostled around and make little progress.
If the corridor is empty, they move in straight lines. This difference in
jostling affects the reading in an MRI and allows you to color code the
image according to the rate of motion of water along the pathways. You
can therefore ‘see’ the rate at which signals flow in the brain and not only
locate thoughts, but also see the links between them.
The Brain
107
Functional PET
PET
The last scan we will look at is functional positron emission tomography,
or f-PET. In this machine the scanner detects positrons given off by
excited oxygen atoms.
As you think, you burn glucose by combining it with oxygen.
The parts of the brain that are thinking hard use a great deal of oxygen
and this shows up in scans. Again the consecutive slice trick is used to
generate a 3D image that allows you to fly through the brain as it works
on a problem.
There is one problem common to all these methods. X-rays, MRI
and PET scans only show us the location of thoughts with an accuracy
of a few millimeters. Each pixel in the image contains around 10 million
neurons, so we can’t see the details of thought. For a scale comparison it
108 Are the Androids Dreaming Yet?
is like looking at a car factory from space. You can see cars and people
going into the factory but you can’t read the owner’s manual. We need
to be able to see at least 10 million times more detail than our current
technologies allow to see a thought.
A Quick Tour
Now that we understand how to look inside the brain, let’s take a tour
around it. The brain is a highly distributed thinking machine. Some
things, such as hearing, are located in specific places while others, like
the enjoyment of music, are spread out.
Our eyes work as an extension of the brain and use a specialized
type of nerve cell. Light falls on the retina and stimulates these cells,
causing nerve impulses to run along the optic nerve into the center of the
The Brain
The Brain
109
Visual Processing System
brain. The impulses split and form two distinct paths, one through the
cerebral cortex, which gives us the sensation of conscious vision, and the
other into the lower brain which provides us with instinctive reactions.
The right hand side of your body is connected to the left hemisphere
of the brain and vice versa. This means each hand is controlled by the
opposite side of the brain. But, your eyes see both your hands. To resolve
this conundrum a very complex thing has to happen to the optic nerve in
the center of the brain. The optic nerve from each eye splits and crosses
over in the middle, so the left side of the left eye and the left side of the
right eye goes to the right hand side of the brain and vice versa. This
keeps the brain focused on the correct hand.
110 Are the Androids Dreaming Yet?
Frogs Eyes are Very Sensitive
The processing power of the eye is staggering. The human retina has
about 120 million rods and 7 million cones, giving it an average resolution
of 10,000 by 10,000 pixels. Each rod is sensitive to individual photons
but we register light consciously only if we see around 5-7 photons. It
is thought frogs can react to single photons because of the chemistry of
their eyes and the fact they are cold-blooded, but this is not proven.
Some animals, including some frogs and my cat, have a tapetum
lucidum. This is a reflective backing to the eye that allows each photon
two chances to react with a rod, once on the way in and, if that fails,
once on the way out. This is why you can see the eyes of some animals if
you shine a light into the forest on a dark night. Cones are less sensitive
than rods but give us color perception. In the human eye, there are three
types of cone: a red, a green and a blue, giving us trichromatic vision.
We see colors because light stimulates more than one types of cell and
we infer the color in between. A fourth type of cone is present in some
species such as birds, reptiles, and fish. This gives them tetra-chromic
vision, allowing them to see into the ultraviolet range. It is speculated
some humans might have this ability but so far none has come forward.
Some animals lack the ability to see certain colors. Most dogs can’t see
red. This gives cats a big advantage!
Many people wonder if we all see the same color as each other. Is
your red the same as mine? The brain’s perception of color is complex.
Although the color red is absolute and can be detected by a calibrated
sensor, our perception of color is relative. We perceive them in the context
The Brain
111
Color is Relative
of other colors – not in isolation. The two panels above contain identical
blocks of color but they look very different against the background. Check
out the website if you have a black and white book. It is an irrelevant
question to ask if my red is the same as yours, since my red against one
background is not even the same as my red against another.
People generally agree on naming colors but not all languages have
the eleven specifically named colors of modern English: black, blue,
brown, gray, green, orange, pink, purple, red, white, yellow, if you are
interested. Ancient Celtic languages, so called ‘gru’ languages, recognized
only four colors and other languages don’t distinguish purple from blue.
Color, or at least the naming of color, is a cultural thing.
Impressionist Painting, Monet Haystack
112 Are the Androids Dreaming Yet?
The resolution of the eye is not the same across the image. High
resolution is concentrated in the center, while lower resolution black
and white vision dominates the edge. This peripheral vision helps us
detect predators or play football but it is not the focus of our attention.
When we focus our attention on something, we turn our eyes to look at
it directly. The central part of our eye is called the fovea centralis and is
composed of cones. About half our cones are concentrated in this very
small section and this gives us immense visual acuity. For a computer
display to outperform this section of the eye it would need one billion by
Scintillating Dots Optical Illusion
one billion pixels. The fovea centralis is tiny, only two degrees across, so
our eyes must dart around the image to take in all the detail. Once the
basic information is encoded in our retina and sent down the optic nerve,
it goes into a production line process in the visual cortex where all the
elements are analyzed. Our brains extract information from the image
such as texture, edges and depth perception in specialized portions of
the brain. Because of this specialization it is possible to play tricks on the
brain with images that are not easy to process. Some we find pleasurable,
while others can be a little disturbing.
The Brain
113
Penrose Steps
Optical Illusions
This picture is an illusion that plays with your stereoscopic synthesis. The
dots appears to flip between black and white. Other illusions play with
depth perception. The Penrose Steps are a type of illusion that tries to
build an impossible physical model in our cerebral cortex. The brain sees
perspective and depth perception cues, but the resulting shape could
never exist.
Hearing
Unlike sight, hearing is an absolute sense. Our ears capture and focus
sound down to the eardrum where a set of small hairs called cilia convert
it into electrical impulses. The impulses stimulate cells corresponding to
specific pitches.
We are born with perfect pitch, yet most of us lose it early on. When
I hear Maria Carey sing a top B flat a specific set of neurons located near
the ear fires, and if she sings a top ‘A’ then a different clump of neurons
are stimulated. By the time most children come to learn music they have
edited out this absolute pitch information. One group of children who
do not lose the ability are Chinese pianists. Because Chinese is a tonal
language – where the pitch of words affects their meaning – and because
114 Are the Androids Dreaming Yet?
McGurk Effect; Go to the Website and Watch the Linked Video
Chinese children tend to learn the piano very young, they don’t lose the
absolute part of pitch. An astonishing 93% of these children develop and
retain perfect pitch throughout their lives.
There are many cross connections between the audio and video
processing systems. At parties you often can’t hear speakers clearly because
of the background noise. Watching their lips will help comprehension,
but which sense wins if there is conflict between the two? The McGurk
effect shows this.
To test the effect, go to the website, watch the video and see if you
can distinguish when a speaker talking normally and when he is making
the mouth movement of another sound. There is a winner. Try it for
yourself; check out the link on my website.
Once upon a time people imagined the brain was like a camera
forming an image of the world, but if this were the case there would be
a paradox. Who is looking at the image in our brain to make sense of it?
Modern research shows we don’t take a complete picture of the world
like a camera but rather parse the image into its constituent parts on the
fly.
If someone asks, “Which side of the house is the tree on?” your
brain parses the question and compares it with the image map in your
mind’s eye. What is the image composed of: trees, houses, sky, grass?
Your brain manipulates the linguistic question about the relationship
of elements and matches it with the visio-spatial understanding of the
image, allowing you to answer the question. You might not have to answer
The Brain
115
Humans’ Ability to Concentrate
the question verbally. If you hit a baseball, no language is involved; you
distinguish the ball from the background and perform quite a feat of
tracking and calculation to connect it with your bat.
Because the brain is editing the scene on the fly to keep within
its processing power, the eye only sees what it turns its attention to.
Magicians take advantage of this to play amazing tricks on us. Watch the
video on the web and then tell me what you see.
VISIT THE WEB AND VIEW THE VIDEO TO SEE WHAT HAPPENS
Tiger Woods Swing
You can see just how intensively the brain works on a given problem,
throwing away all unnecessary information.
The brain contains mirror neurons, a type of brain cell that
responds when we see another human do something. These neurons
fire as if we were performing the action ourselves even though we are
merely witnessing it. It is one of the ways we learn a skill. If I watch Tiger
Woods’s golf swing, my mirror neurons will fire as if I were practicing
his swing. Later when I practice the swing for real, my neurons will have
already been partially programmed. This effect is presumably the reason
we enjoy watching sports; our mirror neurons allow us to begin acquiring
a skill while sitting in an armchair! This is clearly a useful evolutionary
trait but you do also need to practice for real!
Mirror neurons also fire in response to witnessing emotions. When
we see an actor laugh or cry, we experience their emotion as if for real.
This helps us empathize with the person we are watching and is part of
the reason we enjoy movies and plays.
Neural Network
Thinking
“We cannot solve our problems
with the same thinking we used
when we created them.”
Albert Einstein
If you feel mentally exhausted reading this book, don’t worry. This
is normal. Mental work takes energy. Scientists estimate the brain
consumes 20% of our resting energy; around 12 watts. Physical fitness
is important for thinking. If you get out of breath running for a bus,
thinking is going to be harder for you. Studies are mixed about whether
the additional work involved in solving a difficult problem causes you to
use more energy. We certainly see an increase in the flow of glucose to
the appropriate part of the brain, but the overall energy use in the brain
is quite high in the first place, so it is hard to see the incremental effect.
Unlike muscles, which store energy locally as glycogen, brain cells
‘burn’ glucose and oxygen from the blood stream in real time. If scientists
detect glucose and oxygen flowing to a part of the brain they know it
must be working on a problem. As we know, there are several ways to
make glucose and oxygen show up in brain scanners. You can, therefore,
inject someone with the right chemical markers, wheel them into a brain
scanner, and watch them learn new skills. On a practical level, there
is limited space within a scanner and you can’t wield a golf club, for
example. Julien Doyon, a researcher at the University of Montreal, was
recounting this problem to a friend and she suggested knitting. Knitting
is a physical activity you learn just like a golf swing or a tennis stroke, with
all the initial fumbles and jerky activity, settling down to a fluid learned
skill. Most experienced knitters can engage in a full conversation while
knitting complex patterns, only needing to break off and concentrate
during a pattern change. Luckily, there are ceramic and bamboo knitting
118 Are the Androids Dreaming Yet?
needles which don’t interfere with MRI scanners, and they are small – no
golf swing problems here. Studies of knitters show that when they initially
learn a skill, several areas of their brain light up, but after a while, the
brain activity becomes concentrated in the sensorimotor striatal territory.
Glucose, the brain’s power source, is a sugar we get directly from
eating sweets or indirectly by digesting starch. Some studies show
children do slightly better at school if they eat starchy foods in the
morning for breakfast – a bowl of cereal or porridge. When you think
and work your brain consumes the glucose in your blood, and blood
glucose level drops. If there is a steady source of glucose from the starch
digesting in your gut, the glucose is constantly topped up and the level
will stay high. If there is no input of glucose from your gut, the body will
first get glucose from glycogen in your liver or generate it by converting
fat reserves. This takes more work so the body tends to avoid doing so
until it absolutely has to. You can function with slightly lower glucose
levels but the body will shut down a little. One thing that suffers as a
result is the brain’s ability to perform cognitive tasks. A quick and easy
way to fix this is to consume some raw glucose and most fridges have a
ready supply in the form of sugary drinks. Stories of kids running amok,
due to sugar highs brought on by too many sweets and sodas, appear to
be an urban legend. In tests, parents told their children have had a sugar
drink report them to be hyperactive even if they had been given a sugar
free drink. I’m not suggesting you drink lots of sugary drinks – it is bad
for your teeth and will make you fat – but the occasional soda is fine.
Memory
Scientists are just beginning to explore the mechanisms that lay down
memory in the brain. There are two main classes of theory. The first
believes memory is formed in the large scale wiring of the brain.
Neurons connect with other neurons and the number and strength of
these connections cause memory. When we learn, new connections are
formed. The electrical activity in a given part of the brain triggers the
formation of new dendrites. They grow, piloted by tubulin micro-tubes,
rather like vines growing in a slow motion nature clip. Once a microtube
guided filament is close enough to other, a synapse forms. This
gross-scale wiring growth is one method of memory formation. Another
gross-scale effect is myelination. Myelin is the insulation the body uses
on nerves cells, including nerve cells in the brain. It looks a bit like the
insulation we used in the 1930s. Before the invention of plastic, strips of
waxed canvas were wrapped around wires to provide insulation. Myelin
The Brain
119
Synapse
has a similar structure. It is a flat protein laid down as a spiral on the
outside of nerve cells. The theory is that cell firing causes myelination,
which permanently imprints the memory.
The alternate class of theory proposes memory is encoded at a much
smaller scale. Neurons are quite complex structures in their own right.
Inside each neuron is a lattice of proteins, which forms a skeleton. Part
of that skeleton provides structural integrity to the neuron, while other
elements provide control and motility. It is this control part of the skeleton
that people believe might encode memory. A 2012 paper by Travis Craddock
and Jack Tuszynski of the University of Alberta, and anesthesiologist Stuart
Hameroff of the University of Arizona proposes a protein called CaMKII
binds to the cytoskeleton in 32 different configurations, providing a binary
data encoding. It is an elegant idea but it also relies on your believing their
model for quantum neuron processing which is still highly controversial.
If proven, they are my top Nobel Prize tip for this decade!
Photographic Memory
Until recently conventional wisdom held that true photographic memory
was a myth and the few people claiming to have it really used some sort
of mnemonic memory technique to selectively memorize things. The
120 Are the Androids Dreaming Yet?
most famous case was a Russian journalist known as ‘S’. He habitually
memorized things using association with places. In antiquity this was
taught as ‘the method of loci’. The unusual thing was his inability to turn
the effect off, and he found it as much a curse as a blessing. He was unable
to forget useless information and found it hard to interpret complex
images, tending to see areas of color and shade rather than objects such as
trees, houses and fields.
Very recently some people have come forward, six in America and
one in the UK, who appear to have genuine photographic memories It is
well worth watching the TV documentary The Boy Who Can’t Forget to
gain a sense of what this is like. These people appear to lack the ability to
forget, and this turns our understanding of memory on its head. It seems
memory might work the opposite way we thought. We had previously
thought we only remember what we pay attention to, but perhaps we must
actively forget, and this ability is missing in these subjects. Scientists are
studying these people to see if they can understand more about memory.
The Aging Brain
We can explode a myth and encourage older readers simultaneously.
Memory does not deteriorate with age, or at least not until we are very
old. Most studies looking at memory deterioration focus on the very
old and compare them with the very young. Even then, the differences
are small. When people are asked to attempt memory problems there
is a mild drop off with age but the results are quite similar. The most
likely reason older people don’t remember so well is they don’t believe
they can. Perhaps they don’t have as much incentive to remember new
information. Why learn someone’s name if you’re unlikely to meet them
again? Since IQ actually increases with age, don’t believe people when
they say you are going downhill from the age of 40. You are not!
Computer Brains
Computers are really quite simple compared with all the evolved
baggage we humans carry around. When a computer is presented with
instructions, for example, for a program like Excel and a file such as my
expenses, it will load everything into memory and ‘run’ it. The process of
running a program is simple. Each instruction is a number. The computer
reads the number, looks it up in a table, finds a corresponding number,
and writes that down. Essentially that’s all there is to it. From a simple
mechanism like this, we get the enormous complexity of a modern
The Brain
121
computer. The sophistication is achieved through reading and writing
many numbers in parallel, and chaining the steps together so that if you
read a particular number it triggers another read/write process, and so
on. I’m glossing over some details such as logical functions but, if you
know how a modern computer chip is constructed, my description is
not far off. Almost all logic today is implemented in tables to achieve the
speeds we expect from modern chips.
All modern computers are clocked. A small piece of quartz rock
has been polished, coated with metal, and wired up to a control circuit in
the computer. When you apply voltage to the rock it bends and absorbs
energy. When the voltage is taken away it bends back and gives out the
energy. This is effectively a pendulum and it can be used to make an
accurate clock. I used to design these for a living. Every logic gate in a
computer is connected to this clock, and each time the clock ticks the
logic gates in a computer compute.
Most modern computers are entirely synchronous. The clock rate
is set so that the gates in the computer fully recover by the time of the
next tick, and every gate is therefore ready in its standard position when
the next instruction arrives. The human brain does not have a central
clock. Each neuron acts independently – firing regardless of whether the
neurons it is adjacent to are ready or not. It is wrong to think of the brain
as digital. Each neuron does fire and recover, but it may be triggered again
before it fully recovers. This makes for a chaotic and essentially analogue
operation. If one neuron fires when a second has only half recovered, then
it gets half an effect. If the neuron is 80% recovered, an 80% effect. Neuron
recovery time is quite long, perhaps as much as 1/1000 th of a second, and
they are wired in three dimensions to as many as 10,000 other neurons.
It is perfectly possible for a set of neurons to run one ‘program’ when
they are rested and a completely different ‘program’ when they are 50%
recovered and yet another programs if triggered from different starting
locations. I have said ‘program,’ but arguing a brain runs a ‘program’ is
misleading. It is not organized like this.
Neural Networks
A neural network is our best attempt at a computer model for the human
brain. Each neuron is represented by an entry in a table. The entry records
all the connections to it, along with the strength of each connection –
these are called ‘weights’. In some models the connections can be both
122 Are the Androids Dreaming Yet?
inhibitors and activators like in real synapses. An individual neuron will
fire if the sum of all the connections multiplied by the weights reaches a
certain pre-determined threshold.
A neural network does not run a program in the conventional
sense, and must be trained through experience rather like a human
brain. The training process allows the weights in the network table to
be adjusted to give the correct result. But, unlike the brain, you can read
the weights and even save them to a disk. The neural network tables
start with random settings. You show the network the letter ‘A’ and
adjust the weights in the tables until it gives a positive answer: ‘It’s an
A’. Repeat the process with the other letters until the network correctly
distinguishes them. As you do this a computer algorithm constantly
adjusts the weighting tables using a method called ‘back propagation’.
At the end of the training process you can show the network some
new input and see how it does. For example, a letter ‘A’ that is in a slightly
different font to anything in the training set. Trained neural networks
can perform complex tasks such as recognizing faces or making clinical
diagnoses, and they can be allowed to modify their weighting tables as
they work so they learn from experience in a similar way to a human
brain. Strong AI proponents believe making a thinking machine is just
a matter of building a really large, fast neural network and working out
how to train it efficiently.
Quantum Brains
Conventional wisdom says each brain cell is a single processing unit
making an on-off decision – fire, or don’t fire – depending on the state
of its neighbors. But, Stuart Hameroff, Professor of Anesthesiology
at the University of Arizona, thinks neurons are not the fundamental
information-processing unit in the brain. He suggests that this accolade
should go to tubulin. Tubulin is a small, versatile protein that selfassembles
into filaments rather like the way buckyballs – a magnetic
children’s toy – can be arranged. There are two types of tubulin molecule:
α and β. They slot together and wrap around to form a micro tube about
25nm in diameter.
Tubulin micro tubes do several important things in the body.
They form the skeleton of neurons and give them structure. They are
involved in guiding neurons as they grow towards each other to form
new connections, and they also operate in the nucleus of a cell to unzip
The Brain
123
Paramecium
DNA into its two complementary strands when a cell divides. In singlecelled
organisms, including paramecium, the ends of the tubes stick out
of the body and form the cilia that drive the organism along.
The presence of tubulin in complex, single-celled organisms
provides a clue that the smallest information processing unit might not
be the neuron. Some single cell organisms, such as paramecium, display
complex behavior: hunting for prey and escaping danger. This suggests
they can process small amounts of information without the need for a
matrix of neurons. Since we evolved from these organisms, why wouldn’t
our brain cells take advantage of this sub-cellular intelligence?
The structure of tubulin lends itself to digital processing as the
molecules forming the walls have two stable states and can flip between
them. We might recognize this as the basis of a binary computer, and
cells might have little computers within them. They would not need to
process many bits to be useful. Perhaps single-cell organisms developed
information processing capabilities in their micro tube structures that
allowed them to better survive and, as their nervous systems evolved, they
coupled these structures to form the brains we see today. This piece of
theory is not too controversial. After all, nerves have wiring within them
to carry information to the synapses and it’s likely this wiring is involved
in the thinking process. But Hameroff is not finished. He has teamed up
with Roger Penrose to bring quantum mechanics into the picture.
124 Are the Androids Dreaming Yet?
Their reasoning is straightforward but has generated a great deal
of controversy. Hameroff observes that anesthetics cause humans to
lose consciousness by binding to tubulin, but they do not halt all brain
function. He, therefore, concludes our conscious thinking is mediated
by tubulin, not the larger scale firing of the neurons. Penrose had been
looking for a mechanism in the brain that would explain how brains
solve non-computational problems. Together Penrose and Hameroff
propose tubulin micro tubes are quantum gravity computers that allow
us to think non-computationally and are the seat of consciousness. The
ideas are still being worked.
Penrose and Hameroff have a difficult task conveying their ideas
to the rest of the scientific community. Scientists don’t recognize a
need for something that can think non-computably, so they are highly
skeptical of a mechanism which performs that sort of thought. The
latest development on the Hameroff Penrose model comes in the work
of Travis Craddock, now of Nova Southeastern University, Florida, and
others. They have written a paper arguing signals propagate according
to quantum principles within microtubules through the excitation
of thiamin molecules along the length of the tube. They believe these
molecules are quantum, entangled in a similar manner to the mechanism
recently discovered in photosynthesis. The geometry of these molecules
is set out in a similar way to the active areas in chlorophyll and they
have a complementary problems to solve. Chlorophyll tries to maximize
energy conversion efficiency, while a microtubule tries to minimize the
use of energy while propagating signals along a nerve. You might wonder
Tubulin Protein
The Brain
125
Tubulin
where the light comes from since tubulin is housed deep within the
neurons inside our brains and shielded from light by our skull. It turns
out that the mitochondria which powers our bodies emit photons of UV
light as a waste product of their metabolism. The speculation is tubulin
harvests this waste energy.
Before we argue for this mechanism any further we still need to
establish that a non-computational mechanism is needed to allow human
thought. In the next chapters, we will look at the nature of knowledge
and, in particular, mathematical creativity and the Wiles Paradox.
Quantum Coupling of Tubulin in Microtubule

Chapter 5
KNOWLEDGE
Chimpanzee and Typewriter
“There’s an infinite number of
monkeys outside who want to
talk to us about this script for
Hamlet they’ve worked out.”
Douglas Adams
“I’m not young enough to know
everything.”
J.M. Barrie
“He has Van Gogh’s ear for music.”
Billy Wilder
Could an army of monkeys write Hamlet by bashing away randomly
on typewriters? Of course, we don’t mean this literally. We are
asking whether knowledge can be created without understanding.
Can a monkey, or perhaps some form of computerized random number