Templeton: “Open source ape” may become first AI
By Christine Peterson, on January 11th, 2002
from the unnerving-thoughts dept.
Senior Associate Brad Templeton, also chairman of Electronic Frontier Foundation, has been thinking about AI through uploading: "However, the uploading scenario presents a rather disturbing conclusion. The first super-beings may not be based on humans at all, but instead may be apes. In the course of modern science, it is always the case that we experiment with animals first, years before we attempt anything on people. It's the ethical way, and in many cases the only legal way. As such, as we develop the technology to scan or convert an existing brain into an artificial form, we'll try this first on animals. We'll start with lower ones, and then work up to our closest relatives, the chimpanzee and bonobo…Indeed, the software of this chimp brain might be made available for free distribution. An "open source" ape, for all to experiment on." He makes a plausible case; worth reading.
Re:apes don't fit your "human / animal" dichotomy
"Orang-utans, by contrast, are very benign people."
From
"Orangutans are isolated creatures that live most of the lives in the treetops of Borneo and Sumatra. There are two types of male orangutan, large and small, and the female seems to prefer the larger of the two types. But, "prefer" is a relative description, as all accounts of observations in the wild point to the undeniable fact that male orangutans rape the females in order to breed, and that the female seems to struggle less with the larger male then with the smaller male only indicate overwhelming force is being applied. Demonic Males, 1996."
The idea that apes are inherently better than humans is bogus. Bonobos may be "better" than chimps because they have less resource pressure–they don't share their range with gorillas. But bonobo females sometimes maim males.
As for child-rearing skills, I've read that dominant chimps often kill the infant offspring of deposed males.
Chris
build one Earth-wide AI and help it take over?
Also there is this 'three layers of revolution' Gaia Brain stuff. Quite interesting if you're into this 'build one AI and help it take over' plan.
apes don't fit your "human / animal" dichotomy
If "we succeed in uploading an ape mind" then who is "we", how does one judge "succeed" and what is a "mind" in the ape context?
"I [Brad] don't suggest that apes are anywhere near our capabilities." I would say that the Great Apes you mention, bonobos in particular, exceed average human capabilities in child raising and dispute resolution, at least within their wild culture. The idea that Great Apes are inferior in any way to Homo Sapiens seems to be rooted in a genuine reverence for violence, rape, lying, and building huge pyramid-or-obelisk-like structures to put dead bodies or nearly-dead bodies in suits in. So far as we know Great Apes don't do those well. Which is smart, because those are senseless activities.
If the augmented apes are as morally superior as the wild apes, it seems likely that Planet of the Apes would indeed be the optimal outcome… four species getting along symbiotically, none of which had any overly swelled heads about superiority to each other. Doctor Zaius was right: man is evil.
Orang-utans, by contrast, are very benign people.
And they *are* people: at least in some legal and moral sense. The "human versus animal" dichotomy falls apart under examination, e.g. this little unscientific poll – look how little it matters that the human genome is part of some of the creatures you can 'save'.
Apes don't fit your "human / animal" dichotomy – in fact most ape research has been banned in the UK which is otherwise a quite similar culture to the US or Canada. That hasn't happened with rats.
augmenting our moral superiors – the Great Apes
The upload idea is crap, I agree. But the idea that apes will be genetically or otherwise augmented *BEFORE HUMANS* is clearly quite sound.
It's extremely significant that Brad notes chimps and bonobos as likely first subjects. These are the most intelligent of our near cousins, and have sophisticated social customs and language of their own. Augmented bonobos or chimps are actually very formidable competitors to humans, very likely morally superior in aspects like child-raising or dispute resolution (bonobos in particular seem to have mastered the art of resolving disputes via sex not violence).
Planet of the Apes may not be so far off… or so undesirable. A society run by creatures who raise their children lovingly and don't even wean them for seven to ten years, who don't invent dangerous technologies or bizarre ideologies, may well be a lot more stable. Certainly such creatures should be allowed self-determination in their own forests and native habitat. There are already efforts to fully recognize them as persons, e.g. G.R.A.S.P. = Great Ape Status of Personhood
More radically, go to Greenpeace's cybercentre and run a Search (top right corner) with words like "bonobo", "chimp", "gorilla", "orang-utan", "Great Ape", "intelligence", and such. You'll see a whole discourse about this that transcends anything in the scientific community. Well worth checking out.
Recycle of the human brain please recycle!
I've been meditating on this so called mind uploading for a couple of years now. I'm surprised(and delited) to see the amount of people who seem to believe in it more and more. At the time I believed (and still do) that human or none-human mind duplication will not be done by pure and simple "brain scan" and straight forwardly recycled to some other brain mimicing support. What are we supposed to start with to study techniques of "mind" duplication? Is there even such a thing as a "mind"? Might we not just be a heavy mixture of electric circuits juiced up on hormones and neuro transmiters that output such a copmlex and very undefinitive behaviour? I don't want to go upon discussing endless debates…because these are endless. Mind uploading, I believe is not totally unaprochable though, a much wider detour could be tempted rather than straight "technical" mind scan. Maybe the development of artificial intelligent systems that interact with an individual on a "techno-psychological" level could do the trick. What I mean here is that a system with not necessarily human-like intelligence, but a certain type of interactive entity, could directly connect to our body and "learn" our behaviour, our personality, our humor, our psychology, and more generally our motivations. This process would inevitably last an important amount of time, maybe not even reaching perfection in time (before our death), but such a system would most likely be capable of replacing your personality on, let's say , the web, and maybe continue to undertake our former projects. This might be just a dream, but could also turn out to be wonderfully true. "Mind duplication" is a goal some forms of AI's are the tools to reach it.
Re:Fluffy
Fluffy is actually a pretty good name for an ape.
Orangutangs, especially, have lovely long fur that's quite fluffy when clean. They even like being shampooed.
Don't use banana scented shampoo, though. Trust me, you won't make that mistake more than once!
AI rights
Two things.
First, AI that is human equivilant is not the first thing we are going to do. By the time AIs get to human level intelligence expect to see all sorts of other things happening that will make challanging the rights of AIs silly (since AIs will quickly *pass* human level intelligence).
Second, most people won't give rights to AI anyway, so it will be a decision for the researcher to make. Hopefully, the researchers will be smart enough to kill the AI if it starts to go wrong.
Emulation
This is actually one of the oldest debates, and you'll find much written on it in a search. Many people answer that it's meaningless to talk about "emulation" in a pejorative sense, ie. what is the difference between an intelligence and an "emulation" of an intelligence, and why does it make a difference to you?
Can the emulation write a poem that moves you? Can you fall in love with it, or it with you? Is it original and creative? Does it have an inner dialogue? These are all things human "unemulated" intelligence has, so we presume a good emulation could do them too, and many others.
Anyway, there's plenty more on that subject out in the literature, and it wasn't my goal to restart that voluminous debate.
To address a point that relates to my article, however, it's far from certain just what rights will come to AIs and how quickly they will come. If you realize that South Africa didn't give rights to black people until the 1990s, you will see how long it can take for societies not to protect the fundemental rights. One of the hardest questions I have to ask is what franchise
means in a world where you can make another
intelligent being just by copying.
Even if humans stop feeling it's ethical to modify our AI children, they will be able to modify themselves. And they will probably be able to breed like we do, mixing components from parents to form a new hybrid child.
But probably far more than that. We don't have enough info to judge.
Re:Fluffy
I pale when people talk about "A.I."
Why?
First, because IF we do create 'artificial intelligence' it will only be by accident, and second, because, again, IF we do create 'artificial' intelligence we will inevitably create a delta between Human intelligence and AI, which will only create prejudice and misunderstanding. I.E. The moment we create 'true' AI, by our own definition, such an "intelligence" MUST be given the same 'Human rights' that all people have now. We will no longer be ethically able to 'harm', as an means to any end in the name of Science, such an intelligence.
Basically, once we create it, we can't change it, or we're violating our own ethics code. Whether or not we can change what *makes* up the AI, the fundamental pieces, and 'compile' a new version that yields the changes is up in the air, and personally, I doubt we'll ever be able to.
Personally, the closest we'll ever get in my opinion is a machine that EMULATES intelligence, not IS intelligent. That is, unless our own intelligence is nothing more than emulation, in which case "self advancing" AI is similarily absurd.
Re:Fluffy
I don't suggest that apes are anywhere near our capabilities. What I do think is that once we succeed in uploading an ape mind, we will quickly experiment in ways to enhance it, giving it human, and eventually superhuman enhancements.
We know a fair bit about the anotomical differences today, and we'll learn more as we come to understand genomics better. By the time we upload the ape, we may have a very good sense of the "delta" between ape and human that's come in the 7 million years since we split off from our common ancestor.
Thus people will take the uploaded ape mind and start applying bits of the delta to see what they do, plus other enhancements that are not derived from humanity.
Imagine in fact we have scanned dead humans so we understand this delta close to perfectly. It might not be legal or moral to experiment on the human, but it might be fully legal to apply "patches" derived from human scans to the ape, especially if the patches are abilities and not memories of real individuals.
This is the being that might become aware and superhuman first. Not something that is an ape, but something derived from it. It would, however, still contain much of the ape. We contain much of the ape ourselves, after all!
As to whether AI or uploading comes first, that's a well established and unsettled debate. None of us knows the answer. That I favour uploading is just a speculation. Clearly if AI comes first, the AIs may be able to help us move directly to human uploading.
I also believe that when we do learn how to upload, if that comes first, there will be partial AI around nonetheless. And that partial AI will be integrated into the uploaded apes in an attempt to marry the two and give consciousness and self awareness to an AI that doesn't have such qualities yet.
Fluffy
This article is pretty fluffy. Even if you were hooked all the way through, that last paragraph is just plain sillyness. Mainly, humans are only a little bit more rational in a natural state than chimps. I mean, by the rationality the articles used, we should be uploading ape minds since they're a lot nicer than chimps or humans.
My problem with uploading is two fold. For one, despite what the author thinks, the development of uploading will be a slow process if humans are doing the work. It will take a long time (viz. decades and decades) to, via trial and error, come up with a good enough understanding of the brain to copy a mind off of one and into some kind of digital algorithm (just how this is going to be done is rather complex, because it involves figuring out how to get a mind, but not necessarily a brain, into the computer). Even then, we've got a human mind that evolved to work in an environment that never anticipated anything like the post human state. This isn't the kind of change human minds were designed to adapt to, or any minds for that matter, so expect lots of failed uploaded minds if we try to do it. Eventually, after maybe 75 years, they'll get it.
AI, on the other hand, requires BIG, BIG, BIG advances in computer science. But, these advances come quickly. Consider that computer science has basically grown up from nothing to a whole field in the span of about 70 years or so. That's remarkable, considering we're just getting to something along the lines of a ToE in Physics after thousands of years. Computer science moves very quickly. It wouldn't surprise me if in 10 years we were able to build strong AI (`real AI') and just a few years after that to get good AI that creates a Singularity. Then that AI can upload us (or not if, being a lot smarter than us, it turns out that that's a bad idea of the universe, though from where we stand now the chances of this happening are slim, but a possibility should never be completely discounted) in now time. So, 10-20 years as opposed to close to 75. AI looks like the route to me.
Of course, if the magical upload machine shows up, let's use it, though I think that the vast majority of folks would probably not make the best first upload. Right now, I don't even think I'd make a good first upload, though I'm getting wiser all the time and becoming a better choice, though I'm sure someone else out there is much wiser and better mentally equiped to responsibly handle the post human experience and not waste the universe.