Re: [UPLOADING] Is an exact duplicate "me"?

Hal Finney (hal@rain.org)
Sun, 12 Jul 1998 19:40:12 -0700

harv@gate.net (Harvey Newstrom) writes:
> Hal Finney <hal@rain.org> writes:
> > Suppose you are an upload. Someone proposes to pause your program for
> > a moment, then resume it. Will this kill you?
>
> Yes. You would be killing me and then unkilling me. Suspending my
> program by pausing it is similar to the idea behind cryonics. When I
> die, deanimate, or cease to function, I will be suspended. Hopefully,
> someone will later reanimate or unsuspend me. That is what you are
> doing here: You are suspending and then unsuspending the program.
>
> As with cryonics, if I am unsuspended at some point in the future, I
> would consider myself alive. If I am never unsuspended, and my remains
> deteriorate, then I would consider myself dead.
>
> > (Note that such behavior
> > is an inherent property of timesharing computer systems, and that in fact
> > the discrete nature of computer simulations implies that there can be
> > said to be a "pause" between each clock tick.)
>
> I don't mind these theoretical pauses between ticks. What I mind is
> having myself stopped while the rest of the universe goes on without me.
> If I am stopping/starting at the same rate as the rest of the universe,
> then I will continue to experience life in real-time with no gaps or
> loss of experience.

Suppose we take some intermediate case, such as being an upload on a relatively slow computer. Are you dead then? Is there some threshold of speed such that you are literally dead if the computer runs slower than that, because the rest of the universe is going on without you, while you are still alive if the computer runs faster?

Can you imagine discovering that you are running on a timesharing computer and there are occasional delays of up to 1 millisecond during which you are suspended? Would you then view yourself as constantly dying and being reborn? Does this frighten you and activate your instincts which fear death?

> > Someone proposes to pause your program and then resume it on another
> > machine. Will this kill you?
>
> I don't know. This is the same question we are considering now. You
> really can't pause and unpause a program to a different processor. You
> must kill the first program, load a copy into the second program, start
> the second program and then kill the original. What you are asking is
> the same as the original question.
>
> By using uploaded programs in your example, you are removing normal
> human life from the process. It is pretty obvious when a human is and
> isn't functioning. Programs start and stop all the time. Copies of
> them are run here and there. Are they the same program, different
> instances of the same program, or different programs? These are
> semantic questions as to what is "sameness". I would not be willing to
> let a program running my consciousness be terminated, even if a copy of
> it were to be started somewhere else. The choice to kill my program is
> based on my program alone. I don't care what other programs are also
> running, my program does not want to be killed.

I want to emphasize here that we are not arguing any semantic issues about sameness or anything else. All I want to know is whether the scenarios frighten you and make you fear that you will die. Would you be afraid to run on a multiprocessor system because you know that it would switch you among its internal processors, even though the switching would be undetectable to you? If you found that you were running on such a system, would you conclude that you were constantly dying and being reborn, and feel alarmed and upset as a result?

What if it were discovered that the human brain occasionally switches among its neurons, letting some neurons rest and using other nearby neurons while those first ones recover. Would you conclude that you are constantly dying, because the site of processing in your brain is constantly shifting around and not staying with the same set of neurons?

> > Suppose the computer has redundancy internally so that everything is
> > duplicated, two copies of each processor side by side, likewise for memory
> > elements, communication circuits, etc. You are run on such a computer.
> > Someone proposes to remove some element of the redundancy so that
> > there will no longer be this internal duplication. Will this kill you?
>
> Semantics again. There are two copies of me running. If you kill one,
> it will die and there will only be one copy of me left running. Ask the
> identical copies which process should be killed, and I think they will
> give the identical answer: Kill the other one!

What if the redundancy is obtained by using double-thick wires and double-sized transistors in the circuit design? Now someone proposes to save costs by shaving away half of this circuitry, while leaving the logic alone. Would this bother you?

Suppose this thick circuitry has a line dividing it which can be switched between electronically insulating and conducting states. In one state it is conducting and we have the thick-wire case. In the other state it is insulating and we have the divided-processor case. In each state the electric flow on both sides of the line is exactly the same, so no actual electrons ever need to cross the line (although they may drift across due to thermal noise).

Does the conductivity state of this divider change your answer when someone asks whether they can cut away the left half of each wire?

I appreciate you taking the time to attempt to answer these rather exotic thought experiments. My instinct is that the nature of electronic circuits is such that the notion of "how many circuits" there are is not well defined, yet it appears to be central to your idea of identity. That is why I am trying to suggest these various examples where it becomes questionable how many instances of a given computational process there are.

Hal