Re: the ultimate refrigerator

Wei Dai (weidai@eskimo.com)
Fri, 19 Dec 1997 16:45:04 -0800


On Fri, Dec 19, 1997 at 11:57:33AM -0800, Robin Hanson wrote:
> In making my contribution to reversible computing
> (http://hanson.berkeley.edu/reverse.html) I learned enough say with great
> confidence that there is no particular advantage to erasing bits at lower
> temperatures. If it were otherwise you could make a perpetual motion machine:
> erase bits (= replace unknown bits with known bits) at low temps and then
> reverse the operation (replace known bits with unknown bits) at high temps.
> By the "costs less energy" intution this cycle would create available energy.

Isn't that exactly what a heat engine does, when you think of it in
information theoretic terms? Of course it is not a perpetual motion
machine since when it erases bits, it raises the temperature of the low
temp heat sink, and when it randomizes bits, it lowers the temperature of
the high temp heat source, so eventually the temperature difference
disappears. (Of course I'm excluding black holes here, since
when you "heat up" a black hole by dumping energy into it, it actually
gets cooler.)

I don't see how you can say there is no advantage to erasing bits at lower
temperatures, since your free energy is limitied, and at lower
temperatures the same amount of free energy allows you to erase more bits.
The trick is to obtain a low temperature without spending energy on active
cooling, and that's what a black hole allows you to do. Perhaps what
you're saying is that it wouldn't help to cool your computer to 0.1K if
the best heat sink you have access to has a temperature of 3K. I certainly
agree with this.

I think we're probably not really disagreeing, just emphasizing different
aspects of the same idea.