Here’s an idea I have: A computer made entirely out of wind. I thought of this
by contemplating how to build machinery inside a plasma. Every existing
machinery depends on it being possible to make a stable structure using solids,
but that doesn’t work in these extreme conditions. So the alternative
possibility I see is using wind currents.
I actually think something like this could work. The design I’m thinking of
would look something like this: There is one big vortex in the middle. Orbitting
it are many smaller vortex loops. The ones closer to the center orbit faster
than there the ones farther out, and they interact with each other in many
complicated ways. All of these vortices eventually die out, but hopefully they
have the chance to orbit many times beforehand. Using some cleverness (i.e. I
have no idea how this part is going to work) one can make a pattern like this
self-correcting. Once a design is complicated and stable, it’s not far out from
being Turing-complete. What you need to do is add components which have two
different stable states, and then make a system where a component changes it’s
state based on other components. While rough details like this look reasonable,
again, I don’t really have any idea how this is going to work.
One piece of evidence indicates that this is difficult or impossible: Namely,
that it could disprove the conjecture of Navier-Stokes existence of solutions.
The conjecture says that if you start with a complicated combination of wind
currents, you can always extrapolate what they would do in the future without
something ridiculous like an infinite amount of air in the same spot.
The reason it would fail is that in fluid dynamics, a smaller version of any state can always
be made which is both quicker and uses less energy. This property is called
supercriticality. This means that a wind-based computer can make a smaller
version of itself. The smaller version would make an even smaller version in
less time, and eventually an infinite number of computers would be made in a
finite amount of time. Now this conjecture is one of the Clay Millenium
Problems, so you’d expect someone to have tried this approach. The fact that
they failed shows that filling in those “I have no idea how this works” is tough