I'm going to try to abstract some common
principles

among the different self-organizing
systems we've looked at so far.

The way I'm going to try to do that is by
looking at the unifying framework

of information processing and asking about
these systems

as information processing systems.

The first question to ask is

"do self organizing systems – particularly
in biology – process information?"

Well, in the last few decades, it's become
more and more common

to view such systems as information
processors.

We can see that by taking a quick tour of
some samples of the literature.

In this literature, we see books called
"Information Processing in Social Insects",

papers in journals such as Nature:
"Getting the Behavior of Social Insects

to Compute".

As far as the brain is concerned, we see
books about neural information processing

systems, the biophysics of computation.

Bacteria play a role, too. People are now
talking about information processing

in bacteria, particularly with respect to
the phenomenon called "quorum sensing",

which we didn't cover here.

The immune system is another example of a
system people are looking at as doing

information processing.

Genetic circuits, cells and tissues.
They're all now seen as doing

computation of some kind.

Even plants are now seen as being a venue
for studying

emergent distributed computation.

Even more far out, perhaps,

is looking at slime molds as a kind of
computer, as is done in this recent book.

For me, as a computer scientist, it's very
useful to think about these issues by

comparing information processing in
computer science and in biology.

So I'm gonna take a few minutes to do
that. To start with, we can ask what plays

the role of information in each system. In
computer science, that is, the kinds of

computers that sit on your desk,
information is digital. It's static;

it sits there in memory.
It's passive in a sense.

The central processing unit reads zeros
and ones on some kind of memory storage.

However, in biology, information is no
longer digital or static; it's active.

And it's analogue in the sense that it's
not necessarily made up of ones and zeros.

It's distributed in space and time over
the system's components. Information

in biology is represented via patterns
of individuals and their products.

For example, you can think of the ants and
their trails as representing information

about the sources of food and their
locations, or the flocking of birds or

the schooling of fish as being the kinds
of patterns that represent information

in the system.

Information is gathered via local
statistical sampling from these patterns.

So for instance, an ant cannot see the
whole information that's contained

in the system, which might consist of all
of the pheromone trails

that have been laid by the ant colony.
But they can gather some statistics by

sampling the local concentrations of
pheromone right where they are.

Or when ants are trying to decide what
task to adopt, they gather statistics

about what other ants have been doing.
How is information processed?

In computer science, we know that
information is typically processed

via deterministic programs that are serial
(that is, one step at a time), error free

(programmers make a lot of effort to debug
their programs), and centralized rules

(that is, in a central processing unit)
for reading, moving, and writing

information. Whereas biology does
information processing in a very different

way. It's via the decentralized, local,
and very fine-grained, and stochastic

(that is, probabilistic) actions, actions
that involve randomness. In these

self-organizing systems we saw an
interplay of positive and negative feedback.

Positive feedback involved mechanisms

such as recruitment (an ant laying a
pheromone recruits other ants to follow

the same trail) , reinforcement, where
ants reinforce the trails that they see

or fireflies reinforce the flashes that
they see. There's also negative feedback

that controls the positive feedback from
getting out of control, such as

competition, density limitations. When
birds are flocking together, there's

positive feedback between birds that, by
having birds try to get closer to each

other but also negative feedback by
forcing them to separate when they get too

close. In the same way, ant foraging has
some negative feedback by having pheromone

trails evaporate if not reinforced. We
also see that randomness is ubiquitous,

unlike in computer science, where
randomness is stamped out. In nature,

in biology, randomness is ubiquitous, and
it's used by the system to its advantage.

Finally, in biology, more and more, people
studying information processing find that

the language of dynamical systems may be
more useful than the language of

computation.

Finally, an important question: 
how is it that this information

that's being processed in these systems
acquire some kind of function or purpose

or even meaning? Well, we know in computer
science that information processing is

done for our own purposes.

We run programs to compute things 
to help us do our jobs

or entertain us or whatever purpose we
have, whereas in biology, there's no

meaning or purpose from an external
source. Rather, it's the natural selection

for adaptive function that gives rise to
the meaning of the information.

So that really is the meaning of the
information, say, for an ant colony when

it's creating foraging trails. Natural
selection has shaped ants to have

the mechanisms that create that
information and that interpret that

information because it has adaptive
function for individuals and for the species.

Well, computer scientists are
getting more and more interested in

becoming inspired by self-organized
systems to create self-organized

computing systems. The desire is to have
life-like computing systems that have

emergent behavior from simple rules rather
than the rather non-life-like and brittle,

complicated computing systems that we have
today.

And we've seen some of these.

We've seen genetic algorithms, which 
was inspired by Darwinian evolution.

Ant foraging has inspired a set of
algorithms called "ant colony

optimization" that have been used for
things like telecommunications routing.

Firefly synchronization and other related
synchronization methods in biology have

inspired distributed synchronization in
computers and networks.

As I said, Darwinian evolution has 
inspired genetic algorithms.

Brains have inspired neural networks.
Immune systems have inspired algorithms

that perform computer security or network
security based on ideas of how the immune

system protects the body.

And, as we saw earlier in these slides,
slime molds have recently been inspiring

computer scientists to develop new kinds
of search algorithms based on the way

that slime molds self-organize from
individual cells into a unified whole.

As we understand more and more about how
self-organized systems work in biology,

I think that computer systems will take
more and more inspiration and become more

and more lifelike.

At the same time, as we apply ideas 
from computation and

information processing to understand
biological systems, more and more we'll

see a kind of unified framework under
which we can group all these systems in

terms of how it is that they self-organize 
by processing information.