Thinking it again, I see that I was mistaken: it is possible that an evolving AI could come to see humans as a threat and proceed to exterminate us. This is just evolution, a lower species gets trumped by a stronger one. Biological or not is irrelevant. Perhaps this is how it’s mean to be. Let’s have fun while we can:” let’s kill the State!
Species don’t evolve in contrast to each other, but in contrast to the environment. This has to be the most important concept that needs to be addressed here, because it’s one of the most common errors to think that species evolve in terms of each other. Natural selection follows the same parameters of social selection: environment defines the fitness of a class of organisms.
In my imagination of this event, the machines would exterminate us like I would exterminate ants. Why? Because ants are a pest. I don’t care if the ants are good at certain things and I’m good at others, I kill ants whenever I see them in my house. If we were like ants to these machines, I imagine we’d get the same treatment, no more no less.
Do you have anything to justify this thought process?
Justify in what way? I don’t need justification for killing ants, do I?
No, justify the conclusion that AI would see a human as an ant. Humans don’t even see apes as ants. In fact, that’s why many scientists study apes closely as they’re more like cousins from the countryside than ants or bugs or some sort of pest.
You want me to speculate as to how advanced these AI are going to be?
Furthermore, why has joke thread become serious thread?
Ants have no concept of property rights (thus they enter your house), no concept of trade (thus, even if they’re better at some things, you cannot reap the benefit of that), and no extensive communicative ability. Most of the counter-examples I’ve seen for this hypothetical comparative advantage scenario revolve around robots viewing humans as we view animals, which I don’t think necessarily would be the case. Animals don’t have a full concept of trade or property rights.
Well, if a robot apocalypse comes along, you can try diplomacy.
Well lets be honest, does knowing more imply that one is intellectually different from say someone from two centuries from present? I take this question seriously because if we take your logic to its final extent, somehow the first humans wouldn’t be considered human to us because they couldn’t imagine planes, trains, and automobiles. But would that stick? Consider that an AI is really just a human made of silica. Such a man can learn faster, know more, and never need to rest like we do, but he’s still a man just made a bit sharper with better recall and indexing. There is nothing fundamental about the AI’s cognition that separates from humans or similar animals. In this case, it’s a matter of degree, not kind.
If that’s the kind of AI we’re talking about. But I think that would be far too dangerous. Knowing that an AI could become trillions of times more intelligent than humans, I would recommend keeping it to a program where humans control entirely what the AI is able to sense and manipulate.
The AI may be “ambitious” or “want to destroy pests” within that universe, but not in our own physical universe. For it would apparently be physically impossible for the AI program to know anything about our universe, unless someone is foolish enough to remove that buffer.
I’m afraid that would be like keeping one’s cake and eating it too. There’s no way a program can become intelligent enough if he doesn’t have massive amounts of raw information. And that kind of information just can’t be provided in a simulated space. Only the internet could provide the needed info flux to allow an AI to outdo a human intellect. Thus, only if we give it access to the “real world” will AI mean anything at all. I don’t think there’s any way around that. But than again, I’m far from being an expert on the field.
Ok, I am not going to make a video response ladyattis … cause I don’t have a webcam and I also agree with alot of it so I am not sure it is necessary.
You mentioned cybernetics and I did as well in my first post here, I tink it will make the robot overloard scenario unlikley.
If you look at the first post here by Justin it says “What do economics and robotic overlord scenarios have in common?”
This far I have mostly been arguing from the scenario that there are robotic overlord.
(It doesn’t necessary have to be robots either, basically just that there is another sentient species that is “biologically and physically superior” to our own … robots wouln’t be biological but you get the idea I hope … they have superior mental capabilities, superior technology and lots of guns.)
In such a scenario I think it would very possible they would choose to wipe us out. Also I don’t think you can make an economic argument to why they shouldn’t.
It is simply a matter of costs versus benefits. The cost of having us around could easily outweigh the cost of obliterating us plus any comparative advantage benefits. I have also argued why we can’t apply a sense of guilt as a cost factor in wiping us out, it isn’t to us when we kill animals and guilt obviously doesn’t have to be an issue when we wipe out other humans either. It is just a matter of dehumanising them sufficiently first and our sense of moral guilt becomes defunct.
Now it may very well be that robots created by us can never create this scenario. That is however a different question.
I think the scenario in itself is pretty interesting to debate because of it’s application to how successfully we can create universal ethics for all sentient life rather then just humans. It also illustrates problems of creating universal ethics across cultures where one is technologically superior.
Economics alone have difficulties to explain why we should include primitive savages in our system of ethics if the technological gap is so vast that the law of comparative advantage won’t really create a lot of value for the superior party.
Now moving on to the possibility of robot overlords coming into existence.
Thanks for the explanation of empathy. But I am still not sure it is applicable.
Despite having empathy humans are capable of extreme cruelty, so a machine would be also.
There doesn’t necessarily have to be several machines involved in the decision making of the “robotic community”. It only takes one of them to obtain access to nukes and unleash armageddon. I don’t think you have to be mentally ill (from not having empathy) to consider wiping out the human race. I have empathy and I kill animals, I would also happily wipe out entire some species of animals if it was possible. Malaria-mosquitoes as well as some nasty predators that doesn’t really serve a function to maintain a healthy biotope anyhow. Further there are plenty of soldiers who have committed mass-murders, I strongly doubt all of them lack empathy. It is just a matter of removing the incentives to relate to the specific victim by fuelling a belief that they are inferior or simply enough hate.
This argument is valid to why human overlords could wipe us all out too. It is not impossible, but not very likely.
However artificial life would not be dependent on the environment for survival and evolve faster then us. That they don’t need the environment makes it possible for them to wipe us at low cost without dooming themselves. That they evolve faster makes it more likely that they would come to view us a non-moral problem. Just a rock to be removed and this would remove any barriers to killing us created by there function of empathy.
However if there is going to be a nuclear armageddon I think humans are a far more likely to be the cause and will continue to be for a very long time. For basically the reasons that you take up in your video, even though I would have expressed them quite differently if you hadn’t educated me in the science. Namely diversity among the robots, that they evolve into existence and don’t spring up like Skynets (which should among other things create plenty of bonds in culture and trust between robots and humans) and the development of cybernetics advancing the human race beyond it’s biological limitations.
I am not sure if I am misinformed or if you didn’t understand what I meant. That applications of AI that I see being developed are stuff like Bigdog and that is also where the technology appears most useful. Basically using the technology for all sorts of tasks we might as well have trained animals to do. It is probably pretty difficult to train a chimp to clean my flat for me … but what I am trying to say is we use them as tools. As we could have used any non-sentient autonomous servants if there had happened to be any in nature capable of learning the tasks.
There appears to be very little value in creating sentience in itself. While these AI tools have lots of value.
That made me conclude that it is research aimed to creating non-sentient autonomous servants that will be funded. So if we ever develop something sentient it should spring from such research. We then use a guided evolutionary process to build such servants like they illustrate over at TED … it seems efficient. And from this the tools could obtain sentience.
We’re so far off real AI it isn’t even funny, so who is to say by that point humans will not have either a) turned themselves into machines in physical terms and/or b) enhanced their bodies significantly by fusion with technologies by the time AI does come to be, if that ever happens (and there are doubts by specialists regarding its possibility; the biggest hindrance is that human reasoning, which it aims to simulate, is not mechanical and is strongly embedded in our embodidness, and is highly integrated rather than resolving into simple abstract rules; it depends on integrating ever wider data, differentiating and taking in relevant data to formulate knowledge and make decisions, so it results in dealing with ever greater complexity - AI theory at least in the past took it more so to be based on abstract rules, though it’s evolved since; humans underestimate their cognitive capacities)? I think it’s a stupid worry to be honest. If anything a more realistic danger is the prospect of computers trying to interpret human commands as they become ever more versatile and widely used, i.e. what happened in the sequel of 2001 as it was revealed that was in fact what Hal was trying to do. If anything, the rise of AI will probably coincide with the apogee of capitalism as only it could sustain that and any intelligent species will ultimately opt for it, especially one capable of discounting emotional arguments.