Anatomy
First off, what can we learn by comparing the way brains and computer networks are made?
The Brain
Most of us have a basic grasp of brain anatomy. We
know the human brain is made of two hemispheres (sides) connected by
a kind of "parallel cable" called the corpus
callosum. Each hemisphere is divided into four regions called
lobes, which neuroscientists suggest play differing roles in
different kinds of behavior. So, for example, it's known that the
temporal (side) lobe plays an important part in memory, the
frontal lobe is involved with planning our actions, and the
parietal and occipital lobes (at the middle and back of
the brain) are where we process information from our eyes and other
senses.
It's (at best) an oversimplification to suggest that the
different functions of the brain are localized in certain precise
brain regions (as the phrenologists famously tried to do) and equally
misleading to claim that the two hemispheres are like a dysfunctional
married couple trapped in a small house ("the right brain is more
creative; the left brain is more logical"). For the purposes of our
present comparison between the brain and Internet, what's of most
interest is the way the cerebral cortex of the brain (the convoluted top part
where all the interesting, human stuff happens) is made from billions
of neurons (brain cells) that "fire together, wire
together" in elaborate patterns, enabling everything from simple
pattern recognition to long-term memory. [4]
Artwork: If you were to cut vertically through the middle of your head, from top to bottom, through
a line that passes through the center of your nose, you'd see a cross-section of the brain's main functional areas
like this. Image courtesy of National Institute on Alcohol Abuse and Alcoholism (NIAAA).
The Internet
When we talk about the Internet, it's important
to be clear that we really mean the Internet (the global
network of interconnected computers) and not the World Wide Web (the
multimedia library of text, graphics, videos, and other stuff that we
access over the Net, for example, when we Google or use Wikipedia).
The Web is only one of many applications that use the Internet,
including email, VoIP (Internet telephony, such as Skype),
IPTV (television), and P2P file-sharing applications. There's actually a
value in comparing the brain to both the Internet and the
World Wide Web, but it's important to distinguish between the
two and be clear. The Internet is hundreds of millions of computers
(including computing devices such as cellphones and all kinds of
automated machines that communicate over the Net). The Web is
hundreds of millions of websites, most of which contain links
(once known as hypertext links—though that's now starting to sound
impossibly old-fashioned) to other pages on the same site and other
websites entirely.
If you send an email or browse a webpage, packets
of data travel over something like six to ten links between your
computer (a browser or client) and the distant computer
you're communicating with at the other end (a server). [5]
That gives you some idea of how many "layers" of networks are
involved in linking any two points on the Internet; there isn't
simply a one-to-one connection (at least, not in most cases).
Comparisons
It's easy to make crude anatomical comparisons between the
brain (on one hand) and the Web and the Net (on the other). Where the
brain has cells firing across synapses, the Net links
computers over ethernet cables,
fiber-optic cables, or
satellite links, and the Web uses hypertext links to connect one page to
others. Now all computers on the Net are not equally important. There
are major links between continents, for example, carried in a
relatively small number of hugely important undersea cables; [6] if one
of those fails, vast amounts of Internet traffic are slowed down as
they reroute across the world via longer and less direct routes. [7]
In the same way, some web pages are more important (and authoritative)
than others. Google built its success on using an algorithm called
PageRank to rate more highly pages that were more densely interlinked, supposing that they were more important than less-linked
pages. [8]
So its own homepage, Google.com scores highest with a
PageRank of 10, while an average, good quality page might score 3–4.
The sudden loss of Google would (initially) be a blow to the
Web, though we'd quickly find alternatives; the loss of a
PageRank 3–4 site or page wouldn't bother us at all.
Does the brain operate the same way? It's
certainly true that aspects of our behavior are controlled by
circuits of activity within the brain. So, for example,
neuroscientist Joseph Le Doux has devoted much of his career to
fathoming out the complex circuits involved in emotions such as fear. [9]
Cognitive psychologist Alan Baddeley has advanced a successful
theory that our "working memories" (short-term memories where we
process immediate sensory impressions and things we're thinking
about) are made up of various discrete components (such as storage
areas for recording impressions from our eyes and ears), also linked in
essentially sequential circuits, though it remains uncertain exactly how they map to particular regions within the
brain. [10]
Cognitive neuropsychology is the fruitful field
that marries cognitive psychology (the computational theory of the
mind) with the actual anatomy of the brain, and it uses two main approaches. One is to
observe patients with particular patterns of brain damage (often
small-scale lesions damaging relatively specific brain areas) to see
what kind of cognitive problems they have, and then use those to refine
cognitive theories such as memory, perception, and attention. The other
is to look at a cognitive model of something like memory or
perception, speculate what might happen if part of it becomes
damaged, and try to find a real-world patient with exactly those
symptoms.
Summing up, what do we see from a crude anatomical comparison?
The brain is modular, with some degree of specialization between
different regions, and works in parallel. The Web and the Net are
similar: you can damage relatively large parts of either without
affecting the operation of the whole thing—and (Web or Net) it's
likely that it will rapidly recover from damage by relocating the things
it used to do in the damaged areas to other, intact places. Our brains have
the same kind of plasticity, especially when they're young (somewhat
less so when we're older).
Communication
Send an email to a friend and it's very different
from mailing them a letter. While you're letter travels in one piece
along one particular (though not necessarily predictable) route, your
email will be broken up into many separate packets, each with its
target address attached, and each traveling over whatever route
seems best at the time. Packet switching, as this is known,
has proved to be a hugely efficient way to send huge amounts of data
around the world. One of its main advantages is that it can usually (but
not always) route traffic around damage or congestion on the Net.
Picture: Moving house by packet switching: you'd dismantle your home and mail the bricks separately.
After they traveled in parallel by (potentially) separate routes, you'd reassemble them once they reached their destination.
Read more in our article on how the Internet works.
MIT's Daniel Graham and Daniel Rockmore have
recently suggested that it's profitable to consider whether the brain
(specifically the cortex) communicates in similar ways. At the
very least, they argue, that shifts the focus away from the
computational emphasis of cognitive psychology (what happens inside the
flowchart boxes of the mind) to greater consideration of how
different parts of the brain actually communicate (how the
boxes link together). More radically, they suggest there might be
something to be gained by exploring whether the brain shuffles
information in a similar way to how the Net switches packets between
discrete domains. Wisely, though, they end their comparison with a caution: "As
grand descriptions, analogies to technology all ultimately fail to
account for major aspects of brains. No single mechanistic
description has achieved more than a rudimentary description of
perceptual or cognitive systems." [11]
(But if they're heuristics, of course, that doesn't matter.)
The brain and the Net are essentially linear. You
can't send an email from London, England to someone in Tokyo without
it traveling through various intermediate domains (though it's
important to remember that Net traffic doesn't flow as the crow
flies, much less how the boat sails or even how the telephone call is
routed). In the same way, if you see something,
information has to pass in a certain sequence from the retinas in your eyes
to the visual cortex in your
brain (in the occipital lobe of whichever eye is seeing at the time). The same is not true of the
Web. Any Web page can link to any other without passing through any
intermediate page or pages, though there is usually some meaningful
connection: links are (generally) not random.
This spontaneous connectivity is one of the things that's made Twitter so popular and successful. With a
single click of your mouse, you can immediately "follow" your
favorite celebrity—Stephen Fry, Ashton Kutcher, or whoever—and be
one of thousands tweeting them at any given moment. And nothing much
changes in the world. But if your tweet catches their attention and
they tweet you in return, follow you, or ask their followers to follow you,
you can find yourself suddenly followed by hundreds or thousands of people who
never previously knew of your existence. Through one or two simple
"links" (the tweet you sent and the one you received in return),
the Web has slightly but significantly rewired itself. A direct
new link has been made from one side of cyberspace (Stephen Fry) to
the other (you) without passing through anything or anyone in
between. The spontaneous creation of links between previously unrelated
parts of the Web is one of the most creative and exciting possibilities
of cyberspace—and one that remains largely unexplored.
Functions
It's easy to get carried away with fanciful comparisons
between brain, Web, and Net—unless you remember that they're
designed to do completely different jobs. Although we might have
high-flying notions about philosophy and poetry, the blunt truth is
that our brains are designed to run our bodies, nothing more and
nothing less.
[21]
That's why behaviorism proved to be relatively
successful: we can understand a certain amount of human behavior as
mere "animal behavior," at its crudest, almost knee-jerk reflexes
from sensory stimuli to motor (muscular) responses. The Internet and
the Web obviously have no body to control: the Internet's purpose is
to carry information from one computer to another, while the Web is
a highly dynamic repository of human knowledge. Push behaviorism aside, and
start to ask how exactly the brain controls the body, and it's
immediately clear that brains have internal functions that resemble
those of both the Net and the Web: they carry information (like the
Net) and they're active repositories of knowledge (like the Web).
Now it's easy to see obvious parallels between, say, human memory
and computer memory;
but, instead, let's explore the comparison further by considering something less obvious,
perception and pattern recognition, which is one of the human brain's most important functions.
Processing sensory information—seeing, hearing, smelling, tasting, or touching something—is largely about recognizing patterns, which is why it's fair to say
that much of what we see happens in the brain rather than the eye.
From recognizing faces to reading written language, visual perception
is substantially a problem of pattern recognition. From the mid 1980s
or so, psychologists and computer scientists joined forces to
develop neural networks, which are
computer models of pattern
recognition based on layers of connected "units" roughly
analogous to brain cells. You can train a neural network to recognize
things by presenting it with many different examples. So if you show
it a hundred pictures of dogs and a hundred pictures of cats, and
explain which is which, it should be able to tell you whether a
picture of an unknown animal is either a cat or a dog. Neural
networks are built from layers of discrete units connected together
by links of different "weight" and they learn links by a feedback
process (typically one called back-propagation) that adjusts the weights of the links
each time new information passes through them.
Photo: A neural network recognizes patterns using interconnected layers of input units (red), hidden units (blue), and output units (yellow). The weights of the connections between the units represent, in distributed form, the things the network learns.
Can the Net or the Web recognize patterns in
similar ways to the brain, using something like a neural network?
It's hard to see any comparison with the Net. The domains between
which packets of information are switched are not linked by
connections whose weight changes. The Net doesn't change the way
it carries information according to the meaning of the
information it carries. Indeed, the whole reason the Net has been
such a successful design is that it takes no account whatsoever of
the type of information it carries. That means a network that was
originally conceived to carry simple messages between computers (such as
emails) has also been able to carry Web pages, VoIP telephone calls,
TV pictures, and much more besides. (Technically, this is known as
the end-to-end principle, which means that
the inner structure of the network hasn't been designed according to what the network
itself is carrying, and it's related to the similar concept of Net neutrality,
where all Internet traffic is treated the same way.)
But what about the Web? Does that function as a
neural network? Does it recognize patterns? It's certainly true that
the Web consists of discrete points (websites) connected to other
discrete points (other domains) by weighted links. Although, in
principle, every link on the Web is the same as every other link,
some links clearly carry more weight than others: as we've already
seen, that's why Google gives much more credit to a link to your
website from NASA, the BBC, or the White House than from Acme
Dishwashers or Billy's Elvis Presley Fan Site. So the Web has some of
the structure of a neural network, but can it function the same way?
There's a key difference between a neural network and the Web that
we've not yet considered. While all the "units" in the Web
(websites) are, in principle, equivalent, the units in a neural
network fall into three different types: input units (through which
new information is fed in), output units (where results appear), and
hidden units (in between the input and output units where the actual
processing is done). In the diagram up above, information flows
through the network as it does in a computer, from input through processing
(hidden units) to output (as drawn, from left to right). The Web corresponds to the hidden units in the
middle of a neural network: it has no obvious input or output. Not only that,
but it's not arranged to process information in a linear fashion, like a neural network.
It's more like a sphere built entirely from hidden units, where any hidden
unit can, in theory, connect with any other (and not just the units either side,
as in a simple neural network).
But can we stretch the analogy just a little more? We might consider whether certain units of the
Web can act as inputs and outputs. Blogs, for example, often pick up
exciting topical developments from the real world, which are then
discussed and disseminated by other blogs and other websites before,
occasionally, prompting dramatic real-world events of their own. Back
in 2002, for example, in what is often cited as the first
big demonstration of blog power, US politician Trent Lott famously resigned as Senate Republican Leader
after bloggers seized on a careless remark
he made that they considered racist, but which the mainstream media
had chosen to ignore. After much heated online debate, the now-amplified
story was eventually picked up again by mainstream journalists and so
much attention was focused on Lott that he decided to stand down. [12]
Was this an example of the Web acting like a neural network? Attentive
websites had served as the inputs, the interlinked network of blogs,
websites, and social networkers discussed and disseminated those inputs,
hyping them up or playing them down, and finally, the network as a
whole somehow arrived at an aggregated conclusion (an output) that prompted very
real action? The interesting thing about this analogy (and it is only
an analogy) is that any website (or Facebook or Twitter account)
might act as an "active" input or output or a more "passive"
hidden unit, merely playing a small part in the collective
pattern-recognition and decision-making process. So if the Web is,
in any sense, a neural network, it's a very special kind of neural network
where the input, hidden, and output units are in a constant state of flux.
Abnormalities
Brains are complex and wobble about, dangerously
exposed, on the tops of our heads. If they're lucky enough not to
suffer physical damage through something like a head injury, they
have to survive mental illness (which affects a third of us at some point
of our lives), and gradual deterioration as we get older. [13]
The field of abnormal psychology considers behavioral changes caused by
psychiatric disorders such as schizophrenia and depression, while
neurology and neuropsychology probe the various cognitive disorders
caused by brain damage such as following a stroke. It's intriguing to
consider whether the Net and the Web could suffer analogous problems.
People suffer from psychiatric disorders for a
whole variety of different reasons, from "life events" (a failed
marriage could plunge you into depression) to imbalances in brain
chemicals (now widely supposed to cause crippling disorders such as
schizophrenia). It seems to stretch the analogy too far to consider
the Net or the Web developing similar problems. As we've already
considered, there's more mileage in thinking about whether the Net or
the Web can be degraded by localized damage—and how they might react
and respond. We'd only expect the Web (say) to show a brain-like
response to "lesions" if it were arranged in a modular fashion.
To a certain extent, that's true: websites, for example, are
specialized and dedicated to particular topics. If the most
authoritative website about dogs (say a site that most dog owners
relied on for all their information) disappeared overnight, we might
expect the Web to experience something akin to highly selective
amnesia. But it would still have perfect knowledge about cats and
other animals, some knowledge of dogs would survive dispersed across
other websites, and, sooner or later, someone else would reconstruct
the original dog knowledge on a different website, maybe even better
than the original.
Emotions
Libraries are sober zones where even talking may
be forbidden; emotions are cooled, calmed, dispersed, and dissipated. Individual
books might be ranting polemics, but a library as a whole has no
overriding opinion on any subject. Encyclopedias such as
Wikipedia embody a similar spirit with a policy known as
neutral
point of view (NPOV): articles have to show balance without obviously
favoring one argument or another. The Internet is a neutral place
too: thanks to
network neutrality, the traffic that travels from A to B speeds or chugs from domain
to domain irrespective of whether it's the President's State of the
Union address or a vile example of racist hate speech; in a democracy
with a right to free speech, that's exactly what we might expect.
The Web, on the other hand, is quite different.
Happy, angry, elated, or sad, websites can certainly convey an
emotional tone, but it's static—and not quite the same thing
as the ever-changing emotions that flood through our own brains. Blog
posts can obviously be emotional too, and here the tone can
fluctuate from day to day or even minute to minute. Still, although
the emotions are variable, the emotions expressed in any one post are essentially frozen
in time: whether it's a happy post or an angry post, that's what it will always be.
Is there any kind of analog for the fluctuating, responsive emotions in
the human mind? The aggregated emotions of what's called the
blogosphere and the Twittersphere come closer. [14]
You'll often read reports saying things like "The blogosphere went wild..."; indeed, if you
search for the exact phrase "the Twittersphere went", you'll get a
variety of emotional options, including "crazy," "nuts,"
"berserk," "to war," "into overdrive,"
"ballistic," "supermental," "into record-breaking mode,"
"wild with speculation"—and so on. Does this collective chit-chat
amount to anything that could be described as an emotional response? Maybe we should ask that question on Twitter?
Self-awareness
We could argue at length over a definition of
intelligence, but let's assume it's one of the things that most
distinguishes humans from "less-sophisticated" animals and (flying in the face of the Turing test)
machines like computers and robots. [15]
Let's say it's an ability to improvise (generalize or abstract a novel response) from previous
experiences to solve some problem you've never previously
encountered. The human brain is intelligent, by definition; we deem
it so just by having invented the concept of intelligence. But is
there any sense in which the Internet or the Web could be considered
intelligent? The question has no obvious meaning where the Net (which
is merely concerned with communication) is concerned. But what about
the Web? What about things like the blogosphere and the
Twittersphere? Do they have an intelligence beyond the intelligence
of their individual users? The very existence of those terms suggests
there's a meta level on which the Web now operates; that, in turn,
raises the possibility of meta-phenomena such as intelligence,
self-awareness, and consciousness (which may or may not be the same
thing).
Is the Web self-aware? Does a blog post about the
blogosphere or Twittersphere "going wild" amount to the early
stirrings of self-awareness and consciousness? Isn't that taking the
analogy too far? I think it probably is. One of the key aspects of
consciousness is surely the sense of being the "internal conductor"
of your own mental and physical orchestra (however specious
that may be [16]): you are actively conscious of being
or doing something yourself, but you can't be
conscious on behalf of someone else. Thus, I can feel conscious that
I am "staring out of my eyes" and typing these words now, just as
you can feel conscious that you're staring out of your eyes and
reading them. But I can't be conscious on your behalf that you're
reading the words; and, in the same way, I can't be conscious, as a
blogger, on behalf of something called the blogosphere; even if I'm
an active part of it, I might just as well be a passive commentator.
Blogging that "the blogosphere went wild" is entirely different
from saying "I scored that goal," not least because the first
statement has to be shared, externally with other people, while the
second can remain valid purely as an internal thought. A conscious
Web would be something entirely different. Consciousness would be
quintessentially meta: something above, beyond, and entirely apart
from the Web itself, something that we, as mere constituents, would
presumably have no more way to experience than an individual brain
cell could experience human consciousness.
Photo: Self-awareness: Is there anybody out there? It's easy to comprehend things smaller than you are, but harder to grasp that you're part of something bigger—like a galaxy of stars or even the entire universe. Would we, could we, be aware of the Internet's or Web's self-awareness, if it ever did occur? Photo of stellar swarm M80 (NGC 6093), a dense star cluster in the Milky Way galaxy, taken by the Hubble Space Telescope and courtesy of NASA on the Commons.
Is the possibility of a conscious Web good, bad,
or ugly? Commentators such as Jaron Lanier have already speculated that the rise of the "hive mind" (a popular term for the kind of
aggregated, anonymous actions, reactions, and emotions we see online)
comes at the expense of what is most unique and beautiful about individual
humans. [17]
Others are more optimistic. In an occasionally fanciful
book entitled World Wide Mind, science writer Michael Chorost
suggests how ingenious, technically plausible neural implants called
"rigs" could be used to interconnect humans so they could
directly experience one another's thoughts and emotions, heralding
the possibility of a kind of Wi-Fi global consciousness leading to a
"larger, more capacious, more empathetic species." [18]
Then again, we have to remember that this whole
exercise of comparing the brain, the Web, and the Net is simply an
analogy and a heuristic. So the questions we should really be asking
are not "Is the Web conscious?" and "Could it ever become
conscious?", but "What would be analogous to consciousness in the
Web?" and "What can we learn by wondering whether such a thing
could ever happen?" Those remain exciting and open questions.
