Volume 2, Issue 1 
1st Quarter, 2007


The Role of AGI in Cybernetic Immortality

Ben Goertzel, Ph.D.

Page 3 of 7

Practicalities of Cyber-immortality

What about the practicalities of cyber-immortality.  Well, one approach to cyber-immortality is a topic that we've already gone over, somewhat.  You can scan the brain. 



Image 1: Approaches to Cyberimmortality

You can copy the data into a computer.  The copying process moves information bit by bit, giving some possibility for gradual uploading, providing some possibilities for a perceived continuity of consciousness. 

Another possibility is to gather information about an individual's properties and create a computer program displaying these same properties.  Take everything I ever wrote, everything I ever said that got recorded, movies of my behavior -- take all that data, put it all together in some program, based on a software program designed to feel like it’s self-aware, designed to feel like it’s Ben Goertzel. 

I think this is an interesting notion.  Maybe you could, with a sufficiently powerful AI, do this piecing together. In principle, maybe you could reconstitute Ben Goertzel from the traits, of everything that Ben did.  But it's really, really hard.  I'd rather not rely on it for my own immortality.

There’s also the possibility, maybe, of a kind of quantum physics approach to reconstituting people.  In principle, according to quantum physics, every macroscopic event is recorded in the universe itself.   In the little perturbations of particles scattered through the cosmos.  Quantum theory says that information is never actually destroyed – and you could compute the past from the present.  In principle, with a sufficiently powerful computer, you could roll back time and figure out every single thing about every one of us.   That would be a powerful way of doing uploading, but I’m not sure it would ever be feasible. 

What are the obstacles between cyber-immortality and where we are now? 



Image 2: Obstacles to Practical Cyberimmortality

Well -- basically everything.  We don't have a scanned-in brain in enough detail.  We don't have computer hardware that's good enough to receive a human-like intelligence; and we don't understand that much about self-awareness and other relevant phenomena to know if what we're doing will preserve what's important about ourselves.  So right now it's a very valid and important goal to have -- as something to guide our thinking and our research -- but we shouldn't delude ourselves that any of the component technologies are ready. 

Using AI to Explore Cyber-immortality Issues

Now one thing that's occurred to me is that some of these issues, these philosophical and conceptual issues related to cyber-immortality and uploading, can actually be explored using artificial intelligences. 

It's sort of funny to think about uploading an AI because an AI's already a digital thing.   But, imagine you have an AI that you can talk to and that says it feels like it’s self-aware.  It says to you: "Hey Ben, I’m your AI.  I’m conscious.  I'm aware.  How’re you doing?" 

Suppose you take that AI and then you copy it to a different medium -- say, a different kind of computer hardware.  Then when it becomes smarter, what does that uploaded, improved AI say about the other one?  Does it say, "Hey, you stole my hard drive"? Or does it feel like the same one?  How many changes can you make to the AI's intelligence levels, to the AI's implementation -- and have it still feel like it is the same mind, the same identity. 

It may well be possible to experiment with these ideas with AI programs with more flexibility than we can do with humans -- because with an AI, both we and it command a greater ability to inspect its own internals than exists with human beings.  Potentially, we could discover disturbing things.  We might discover that, in every case, if we double the intelligence of a system, it doesn't feel like its old self at all afterwards.  It feels like it's a totally different thing.  And if we did discover this, this would let us know that doubling our intelligence is basically equivalent to murdering our identity. 

On the other hand, we might find that if intelligence is ramped up gradually, then there is a feeling of continuity in the emergent pattern constituting our phenomenal self – that is, our identity is preserved.  And then we would know that we should upload ourselves more gradually, if we care about identity preservation. 

I think it may wind up that we can explore a lot of these issues at the boundary of cognitive science, uploading, and philosophy with AI minds, rather than by experimenting with them initially on ourselves.



Image 3: Uses of AI for Cyberimmortality

Of course, whether this is the way things happen depends on whether AI progresses rapidly compared to the infrastructural technologies for human uploading. 

All in all, I think there are two uses that AI technology may have to assist with cyber-immortality.  One of them is that if AI proceeds rapidly, artificially intelligent scientists can help us figure out and solve these very difficult problems with uploading.  They can help us figure how to scan brains, how to build better computer hardware.  And the other issue is the one I just mentioned: AI minds may potentially serve as subjects for experiments with the nature of consciousness itself and mind transformation. 

And there are ethical issues here.  Once you've trained an AI mind, you may have a system that’s a conscious intelligence just as much as we are.  You know, it's not very nice just to mutate its consciousness, against its will.  "I'm sorry; we are now going to make you mentally challenged.  You didn't like it, huh?  Too bad.  You're just an AI.  You have no rights."  You don't want to do that. 

On the other hand, it may well be the AI mind will willingly participate in appropriate experiments on its mind -- as I would personally, particularly, if as with an AI there would be opportunities to be rolled back to my prior state. 

So I think AI can have an important role in cyber-immortality in these two different ways. 

 

 

1 2 3 4 5 6 7 Next Page>