I have a possible paradox for you, and a theory that should solve it.

There are a thousand trillion trillion (1e+28) molecules of air in a certain room. It's a constant 70 degrees Fahrenheit, so these air molecules are of course bouncing around like crazy.

Suppose you somehow record a moment of this chaos, and you try to interpret the activity of the molecules sort of like they're a newbielink:http://en.wikipedia.org/wiki/Billiard-ball_computer [nonactive].

In addition to the sheer quantity of activity going on, there's no single "right" way to translate between molecule activity and computation-related data, and depending on how you choose to do this translation, you could probably say the molecules are performing whatever computation you like. Interpret the activity of molecules one way, and some are adding five and three. Interpret another way, and some are calculating the square of 65. Interpreted yet another way, some are simulating a game of newbielink:http://en.wikipedia.org/wiki/Tic-tac-toe [nonactive].

If computation requires (or is) work, how can it be performed without any increase of entropy? That is the paradox.

Add to that the possibility that an infinite amount of computation is occurring in the room every second due to the limitless number of ways to interpret the molecule activity.

I propose that computation doesn't require any work at all, but getting a desired computation does. A free Tic-tac-toe simulation will occur in the room, but to make use of it, you would need to know where it will occur and how to interpret it, and knowing these would be far from free. For now, you should keep using normal computers.

It may seem like a trivial distinction that computation doesn't require work, but it could make a big difference for some hypothetical scenarios like newbielink:http://en.wikipedia.org/wiki/Boltzmann_brain [nonactive] and newbielink:http://en.wikipedia.org/wiki/Simulism [nonactive].

A Boltzmann brain is an entity that arises due to random fluctuations in chaos. If computation is free, things like Boltzmann brains might be more common.

And if computation is free and all around us, naturally occurring simulations could be much more common, increasing the probability that our reality is itself a simulation.