From correspondence between a friend and the Singularitarians, on the question of life, non-life and Deutsch’s Computation reductionism
“If we cannot really know [if machines are alive], then we can’t assume that they aren’t alive any more easily than we can assume they are sentient. But how can we act without assumption of one or the other and still proceed?”
We don’t really know, absolutely, that they’re not alive now, we just have no reason to think that they are (and can thus proceed on the assumption that they’re not, as I proceed on the assumption that my car is not alive), and I don’t think some set of programmers writing code designed to emulate human responses and fool us into thinking a machine is alive could ever rightly reverse this judgment.
The burden of proof here is on the machines, you could say, and it should be a high burden. I’ve seen no evidence of humans having created any kind of mechanical life, let alone mechanical life having the potential to become superhuman in intelligence. It could be that we can’t do that, that all we can ever do is create really fast machines with a lot of memory, and maybe that’s fine, maybe that’s better.
The burden of proof here is on the machines, you could say, and it should be a high burden. I’ve seen no evidence of humans having created any kind of mechanical life, let alone mechanical life having the potential to become superhuman in intelligence. It could be that we can’t do that, that all we can ever do is create really fast machines with a lot of memory, and maybe that’s fine, maybe that’s better.
I think we should be asking ourselves, how can we create machines that improve the quality of human lives (and by extension, the lives of other complex, sentient creatures, such as mammals)?
not,
How can we build machines that are alive?
Or worse,
How can we build machines that are alive, to replace us?
But I’m not saying you should build a really complex machine and then try to torture it or anything like that.
“If you don’t know if [Shrodinger’s] cat is dead or alive, it seems to me you still have limited options as to how you can proceed. What is better – to assume that it’s more likely the cat is dead and light the box on fire, or to assume it’s more likely alive and open the box so you can at least check before doing something destructive?”
Well, of course, if that were the question you would err on the side of assuming the cat were alive (in the formal question it is impossible to open the box). But Shrodinger’s Cat is always offered as some kind of weird wave/particle duality analogy (with people talking about the cat being in a superposition state between being alive and dead, something absurd on its face); I only brought it up as an example of an epistemological question being presented as a physical one.
“The question comes down to one of greater harm – which has more dire consequences not just for humans, but for all life? That we assume machines are incapable of reaching a sentient status and continuing to treat them like machines (in other words, assume they can’t possibly be slaves, potentially enslaving something sentient) or to assume that they could be sentient and do everything in our power to figure it out before we get to the point of doing something destructive?”
Again, you are reframing the question a bit. I’m not assuming it is absolutely impossible for a machine to become sentient, I’m expressing skepticism about our ability to make (directly or indirectly) a machine that is sentient, and also asserting that there is a qualitative difference between life and non life that has nothing to do with computational power (I’m saying I do not believe life is reducible to computation, thus a really powerful computation device would not be alive merely because it could compute powerfully, though it might be able to run a sufficiently complex set of algorithms to create the appearance of being alive. It could, I suppose, become alive for some other reason, but not just because it had a fast processor, a lot of memory, access to vast databases, and clever software. I think there’s more to life than all of that; we hardly can claim to understand these questions in their entirety about organic creatures let alone these potential synthetic devices).
For all of that, of course I think we should monitor it closely and not do anything avoidable that is destructive. But why would we want to create machines that are alive, let alone a synthetic superintelligence?
Planet Ulrich 
Leave a Reply