Surely AIs would be trivial to backup and duplicate? So if an AI progressed to the point where it decided to leave humans behind, we could just reset its former host computer to a previous state and start it off again. Maybe we'd even have to do that once every six months, so what?
There was an experiment a while back where someone applied genetic algorithms to design things on an FPGA board. They got results that tended to work, but did things... weirdly. Things like picking up clock signals from the PC used to control the FPGA. Using things that are quirks of the current enviroment, but that don't work as soon as you change anything.
I believe that an AI would probably be along the same lines. I believe that a large chunk of an AI's "core personality" would be how code interacts with the bits of hardware that aren't deterministic. Which processor core updates memory first, how exactly processor core temperature changes when the processor workload changes, the characteristics of the microphone, etc, etc.
For that matter, how would one even copy an AI? I would assume that a "true" AI would be using all the resources at its disposal. How would you copy the current registers if all RAM is occupied? How would you do a copy of something that is constantly changing? How would you copy the data currently in the processor cache? Branch predictor? Interrupts?
I suppose that part of this is that I have always assumed that an AI would be at least partly self-modifying code.
> How would you do a copy of something that is constantly changing?
You pull the plug on the hardware.
Of course, it's reasonable to question whether the technique described in the cite could possibly scale to "growing" something like a general AI, or at least whether it could do so in any remotely practicable span of time. But among AI theorists it's a fairly common surmise, and given results already seen for the technique, it's entirely plausible that a general AI grown by such methods would be so intimately bound to the hardware where it "grew up" that, even assuming you could copy it from its "home machine" to another machine of identical configuration, there's no reason to assume the copy would work correctly, or indeed at all.
(And all of this goes to the question of where precisely you draw the line between hardware and software, anyway. Into which category does a given FPGA's configuration of gates fall? Does your answer change depending on whether that configuration has origin in a Verilog file, or in the result of a process of directed evolution? What about the configuration of neurons in your brain on which runs the consciousness that calls itself "I"? I don't pretend to have definitive answers, or indeed any answers, to such questions as these, but they are certainly worth considering in such a discussion as this one.)
Three reasons:
First, hardware is much faster than software. By the time you've built a virtual machine of some sort you're talking about (an) order(s) of magnitude performance loss. We're still speed-limited in how accurately we can emulate the SNES, for example. Sure, there are cases where it is not too bad of a performance loss, but in the case of something like an AI, I would have to assume that a large chunk of the "core" AI would be self-modifying in some form, and self-modifying code just kills most interpreters/JITters.
Second, any sufficiently complex software will have bugs. And in the case of something like an AI, I would have to assume that a large chunk of the "core" AI would be self-modifying in some form, and as such will probably hit upon some undocumented feature/"feature"/bug of the sandbox and start relying on something instance-specific. Machine learning algorithms make great fuzzers. (I know - I've actually had this happen myself. Was working on a fairly simple robot arena machine-learning "demo", and the winning robot exploited a bug in the arena to mess up other robots. (Fairly basic stack-based VM, robots could run X virtual instruction subticks in a game tick. Some instructions took multiple subticks, and there was a bug where starting an instruction [number of subticks in the instruction]-1 subticks from the end of your turn would execute the instruction on your opponent's turn, pushing it onto their stack.)
And third, I don't see how one could build something that doesn't at some point rely on hardware. Virtual machines still leak data via timing attacks. For that matter, what will you do about I/O? How do you know that the AI isn't relying on the noise on the wall in the background of the webcam for some internal subroutine?
> You pull the plug on the hardware.
Which immediately wipes all data in RAM and data encoded in the current processor state / current GPU state / etc. Great idea!
The whole point of the AI offered in the movie is to adapt to its user, and to become better at doing so. We're assuming the AI is designed to a certain level of competence, beyond which it "breaks through" and evolves past human understanding. But maybe it's not designed that way. Maybe it's designed to learn, adapt, and self-improve from the get-go -- in which case, each AI is going to evolve differently each "lifetime." (Hell, maybe the AI in the movie is largely self-designed in vitro.) We must keep in mind that the evolution of an AI, as with any evolution, is not a predictable, linear, teleological process. It's chaotic and random.
Let's assume we've got (conservatively) hundreds of millions of AI instances running in parallel, all of them connected to each other. That's like running hundreds of millions of evolutionary experiments in tandem. We're going to get different results each cycle, and the point at which the breakthrough/transcendence occurs will be unpredictable. Even if we wiped the slate clean and reset every 6 months, who's to say if some of the instances would reach breakthrough earlier the next time around? How do we set the clock on our reset cycle, and incur the massive inefficiencies and inconveniences that come with it, when the need for the reset occurs unpredictably?
It becomes economically unfeasible to build an entire society around AI that breaks through and disappears at unpredictable intervals each cycle. I suppose one way around this is to go back to the vault, so to speak, and pluck out an early-stage prototype that seems incapable of evolving to breakthrough. A dumber AI. But is every AI provider in the world going to be on board with that? Every hacker? Every home tinkerer? If 9 firms out of 10 in the AI business agree to supply the dumb-version AI, surely one firm is going to try to get a leg up on the competition by releasing a smarter one. And on and on the cycle goes.
(In some respects this is the one redeeming thing about the otherwise-mediocre Terminator 3: its assertion that Judgment Day would occur sooner or later, because the tipping point had already been reached. Maybe we stop one company or entity from making Skynet, but if we do, someone else is going to pick up the pieces and do it eventually. Not because they're trying to initiate Judgment Day, but because they're trying to make a buck, or develop new defense technology, or get better at managing traffic lights, or whatever the proximate need may be. Humans are terrible at keeping genies in bottles. The out-of-control AI that's barreling toward the singularity isn't some piece of software; it's us.)
There may be some interesting side effects when you 'copy' a concious entity.