RE: Minds, Machines, and the Multiverse

From: hal@finney.org
Date: Sat May 06 2000 - 18:51:34 MDT


Billy writes:
> A VR can never be the environment of anything in a thermodynamic sense. It
> can't even be the environment of a purely virtual device. All the bits in
> your robot are represented by real bits in a physical computer. So are all
> the bits representing the current state of the VR, and its past and future
> states. When you try to reverse the virtual operations of your simulated
> robot, it is this mass of physical bits that must be manipulated. In order
> to return the VR world to a previous state, you have to return some pattern
> of bits in the physical machine back to a previous state. At that point you
> run into exactly the same constraints you would have if you weren't using a
> VR.

No, you don't run into the same constraints. The reason is that in
the case of the VR, the bits which must be reversed are all bits in
your computer. You had complete control over the design of this system
and you can arrange for the system to be reversible.

In the case of a robot interacting with the real world of grass and trees
and rocks, those systems are not reversible. They contain dynamics which
increase entropy and prevent reversibility.

You can reverse simulated grass, simulated trees, simulated rocks,
and their interactions with a simulated robot in a virtual reality.
This is a special case of creating a reversible computer. But you cannot
reverse real world systems. That is the fundamental difference.

> This is just a special case of a much more general (and obvious) rule: no
> form of software can ever perform a computation that could not, in
> principle, be done by the hardware it runs on.

That's not what we're talking about doing. We are asking whether a robot
in a virtual reality which interacts with a simulated environment Could
use less energy than a robot could which interacts with physical reality.
I refer you back to our earlier exchange:

: > > For example if you want to minimize energy costs in running your brain,
: > > you have two choices: live in the real world with an artificial brain,
: > > or live in a VR. The latter gives you more opportunities to save energy,
: > > because for example you can reverse time for the entire world in order
: > > to undo some reversible calculations.
: >
: > Now we're going back to the same idea you ruled out in your first paragraph.
: > If a mind is implemented in software you already have complete control over
: > its 'reality'. There is nothing you can do by putting it in a VR that you
: > couldn't just do directly. There may be advantages of convenience or
: > flexibility in the VR approach, but a more direct implementation of any
: > given trick will always run faster and cost less energy.
:
: If you're living in the real world, you're interacting with the real world.
: You can't rewind the real world.
:
: If you're living in a virtual world, you're interacting with the virtual
: world. You can rewind the virtual world and this gives you opportunities
: to save energy. That's the difference.

Hal



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:28:28 MST