[FoRK] When Does A Physical System Compute?

Eugen Leitl eugen at leitl.org
Wed Oct 9 01:09:19 PDT 2013


https://medium.com/the-physics-arxiv-blog/7e8b1b0212fe

When Does The Universe Compute?

The idea that every physical event is a computation has spread like wildfire
through science. That may need to change now that physicists have worked out
how to distinguish between systems that compute and those that don’t 

One of the hot topics in computer science is unconventional computing. This
is the exploitation of unusual or exotic systems to perform computations.

Examples are numerous. Perhaps the most advanced is quantum computation which
exploits the strange laws of quantum mechanics to perform computation. But
there are other more exotic approaches such as using DNA to perform millions
of simple calculations in parallel or even using slime mould to solve mazes.

Indeed some scientists claim that every physical event is a computation.
Others disagree saying that this simply redefines the notion of a physical
process and is either wrong or trivial.

That leads to an interesting and important question. What does it mean for a
physical system to compute? Can researchers decide objectively whether a
physical system is computing or not?

Today, we get an answer thanks to the work of Clare Horsman at the University
of Oxford in the UK and a few buddies. These guys say that a physical
computation is the use of a physical system to predict the outcome of an
abstract evolution. As such, it is closely related to, but crucially
different from, the notion of theory and experiment .

They go on to tease apart these ideas and to define the abstract and real
components that a computation must have. This allows them to explain why many
examples of unconventional computation are actually experiments rather than
computations.

The crucial ingredient in Horsman and co’s ideas is the relationship between
the abstract notion of a computation and its physical set up. This is where
the link with experiment and physics comes in.

In physics, scientist use an abstract theory to predict the outcome of
physical evolution. However, there is always some difference between the
theory and the experimental result. The goal of physics is to make this
difference as small as possible. When that happens, the result is a law of
physics.

Computation is essentially the same process. A computation uses the evolution
of a physical system to model an abstract theory.

But this only works when the link between the real and abstract worlds is
clear and well understood. In other words, a computation relies on the laws
of physics, on a physical system where the difference between theory and
reality is known to be small..

The problem with unconventional computing is that it often relies on
processes that are poorly understood. For example, the processes that slime
mould uses to solve a maze are largely unknown. For this reason it is not
computation.

Slime mould gives interesting answers in specific well-defined experiments
but cannot be used for general computation. Any attempt to use slime mould to
run a spreadsheet, for example, would produce an outcome that is unacceptably
different from the theoretical one.

That’s not to say that slime mould couldn’t run a spreadsheet in principle.
But this would require lots of adjustments over many iterations to build a
proper theoretical understanding of the system and how it works.

It is this process of engineering that turns a physical system into a
computer, argue Horsman and co. “Computers are highly engineered devices,”
they say.

And therein lies the key. “In general, … technology stands or falls on the
confidence in the underlying theory,” say Horsman and co. Without a good
understanding of the underlying physics and solid engineering to make the
device usable, a physical system simply evolves rather than computes.

This is why quantum computing is a good example of unconventional computing.
It is based on well understood physical theory and the devices themselves
have undergone many iterations of engineering development. In many cases,
they can truly be described as computers.

An important part of this engineering process is to develop a way for the
system to encode and decode information. “Without the encode and decode
steps, there is no computation; there is simply a physical system undergoing
evolution,” say Horsman and co.

That’s one of the problems with many unconventional computing systems—there
is no general way to encode or decode information. Taking slime mould again,
it’s simple to imagine the mould growing inside a maze but much harder to
imagine how this might extend to playing Minecraft or sorting a database.
There is no general.way to encode or decode information

“This is how we can escape from falling into the trap of ‘everything is
information’ or ‘the universe is a computer’: a system may potentially be a
computer, but without an encode and a decode step it is just a physical
system,” say Horsman and co.

These ideas are likely to generate some lively debate in the computing
community, where much funding is available for unconventional computing. The
momentum behind much of this work comes from the observation that some poorly
understood systems seem to compute much more quickly and efficiently than
silicon computers.

For example, when it comes to pattern recognition or walking up stairs, the
human brain leaves silicon supercomputers in the blocks.

Nobody is quite sure how the brain does this amazing trick powered by little
more than a bowl of porridge. But there is general agreement that if we could
somehow copy this ability, we would all be better off.

The worry is that if the work on unconventional systems is not considered
computation until we can exploit it systematically and generally, then some
lines of funding might dry up.

This would be absurd short-sightedness. Computer scientists must be able to
to rethink some of what they call computation if they need to.

There are implications for other scientists too. Horsman and co present a new
definition of computation. But in the process, they also sharpen the
definitions of physics and engineering in ways could be equally
transformational.

Either way, there are exciting times ahead for computers and those who build
them.

Ref:arxiv.org/abs/1309.7979: When Does A Physical System Compute?


More information about the FoRK mailing list