Hal writes:
>Robin writes:
>> On 5/5/2001, Hal Finney wrote:
>> >A reversible computer is as likely to take a step backwards as forwards.
>> >So even if it manages to complete a calculation from A to B, the process
>> >will be a random walk, moving forward and backwards many times over each
>> >portion of the path from A to B.
>>
>> I think you mean to say that it is *nearly* equally likely to take
>> steps in either direction. There needs to be some small bias to get
>> something predictable to happen.
>
>Yes, although the bias can in principle be as small as you like, depending
>on how long you want it to wait. I wonder what it would be like if you
>were implemented on a computer with no bias? I guess the question is how
>far it would get via random walk before the computer falls apart (sqrt(t)
>steps after time t). But using zero energy has enough advantages that
>maybe in some circumstances you would risk it.
Now Hal had earlier written
> The model of reversible computation I was using is due to Bennett
> and is what he calls "Brownian" computation. A good article by
> him is on Robert Bradbury's site at
> http://www.aeiveos.com/~bradbury/Authors/Computing/Bennett-CH/NotHoRC.html.
Okay, sure enough, this article discusses the "Brownian" type
reversible computer that Eugene and I had never heard of. And,
of course, as Hal had inferred, it's a perfectly good way to
do reversible computing. But it's not the only way. You all
have surely read about the billiard-ball model, which Bennett
also discusses here. This more familiar model proceeds in a
linear trajectory, and is exactly what I had in mind when I
first raised this topic, and evidently what Eugene and Robin
had in mind as well when they responded.
So for zero energy, Hal, and a possible resumption of our earlier
discussion, consider that by "reversible computing", most people
are talking about something like the billiard ball model.
Lee
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 10:00:03 MDT