We should be so lucky to have machines outsmart us.
This article fails to communicate the unfathomable chain of events such technology would set in motion. Without hyperbole, all previous events in human history would pale in comparison.
That is hyperbole; for all we know, the first human level AI (for however you define human level) might just want to read lolcats and watch sports (or study mathematics but not help us, etc.)
That's what most humans want as well but there is a system of incentives that keep people working. Needless to say the 'productive' versions of AI will end up with a lot more computing resources than the 'lazy' ones.
Apparently not, and i'm somewhat dismayed at the lack of opposition to this point of view ... so i'll try to make a case against it.
Creating smarter than human AI would be humanity giving up its dominion, and who can say what will happen as a result of this. It is not being melodramatic to say that there is a major chance that it will be the end of humanity, the end of history and the end of everything that most humans value.
I would submit that our values, what we find wise, beautiful, kind, humurous or otherwise virtuous, our appreciation of the natural world and our legacies, they are all relative to our human condition, and that jeopardising this - for everyone - is wrecklessness without comparison. Yet enough of us would do it
Were is this complicity in our own demise coming from. Are supporters of ai and the singularity misguided - by their own values - or do we have irreconcileable philosophies on the matter.
My suspicion is that a lot of support for the singularity comes from death anxiety - the way things are, in the long run i'm dead anyway so may as well have a punt (50-50?) on immortality, with humanity as the stakes. That is the singularity is a plausible alternative to an after life. Call me a wanker, but i think that one should take fatality like a man (i.e and die), and not be so selfish.
For me, one good, or at least indisputable reason, is disillusionment with humanity... i would not be able to argue against someone who lived through the trenches in WWI and had such an opinion, that this world as we know it is just not worth it. However I would not agree.
Finally I think a lot of support for the singularity comes down to pure hair-brained optimism and reading too much sci-fi. It actually made me a bit sad to watch the recent star trek movie and its clumsy attempts to make the characters relevant (a sword fight? come on!). Traditional, speculative, science fiction used to be about final frontiers and buckaneering captains, now its struggling to reconcile the future and anything that we might want from a narrative.
Smarter than human AI may be our story is coming to an end ...
The technology to end humanity is already here (nuclear and soon biological weapons). In the long run superhuman AI may be the only thing that can prevent us from destroying ourselves.
Even if it were desirable, stopping the advance of technology is impractical. If superhuman AI is possible, it will be built. A more practical argument would be that superhuman AI should be strictly controlled, though advancing technology will eventually make that difficult too. The only solution in the long (really long) term is for us to become superhuman ourselves and attempt to preserve the things we find important in the transition.
> Were is this complicity in our own demise coming from. Are supporters of ai and the singularity misguided - by their own values - or do we have irreconcileable philosophies on the matter.
This reminds me of an (unintentionally?) hilarious LW post:
"For example, most existential-risks activists (scientists doing networking and research about risks like unFriendly AI) are male, and I plan a top-level post to assert that not having reliable access to sex with the kind of sexual partners who can most improve the life of an existential-risks activist should be considered a large disability in a male prospective existential-risks activist -- in the same way that, e.g., an inability to stop rationalizing one's own personal agenda should be considered a large disability."
So there you go. When skynet gets invented, blame it on sexual frustration :).
It sounds like you're equating "AI" with "sentient intelligence with its own values and goals which will only arbitrarily coincide with ours." Certainly, that's a possibility--and one AI researchers would do well to avoid.
But the possible intelligences we can observe on this planet have more variety than that, and the useful intelligences we've created have even more. There's no reason a smarter-than-human AI couldn't simply be a general-purpose tool in the way that a faster-than-human spam classifier is a special-purpose tool.
I know it was a typo, but software that's sentimental as well as sentient is probably what we need, as opposed to the vast, cool and unsympathetic variety.
I think most proponents of strong AI include the idea that they should be built on the foundation of a "premise-state" compatible with our human ideals such as intrinsic respect for life, appreciation of complexity and diversity, decentralisation of power, privacy of thought and freedom of action, etc etc.
Which, sadly, demonstrates how far AI as a discipline has to go. I am much more encouraged by the development of Rommbas and similar that seek out an electrical outlet, and so are actually motivated to learn, by something akin to hunger.
Comments
Am I the only one looking forward to this?
We should be so lucky to have machines outsmart us.
This article fails to communicate the unfathomable chain of events such technology would set in motion. Without hyperbole, all previous events in human history would pale in comparison.
Hate to be a pedant, but the important thing is not intelligence that exceeds humans, but intelligence that can scale arbitrarily.
That is hyperbole; for all we know, the first human level AI (for however you define human level) might just want to read lolcats and watch sports (or study mathematics but not help us, etc.)
That's what most humans want as well but there is a system of incentives that keep people working. Needless to say the 'productive' versions of AI will end up with a lot more computing resources than the 'lazy' ones.
Though the first human level AI won't stay alone for long.
Apparently not, and i'm somewhat dismayed at the lack of opposition to this point of view ... so i'll try to make a case against it.
Creating smarter than human AI would be humanity giving up its dominion, and who can say what will happen as a result of this. It is not being melodramatic to say that there is a major chance that it will be the end of humanity, the end of history and the end of everything that most humans value.
I would submit that our values, what we find wise, beautiful, kind, humurous or otherwise virtuous, our appreciation of the natural world and our legacies, they are all relative to our human condition, and that jeopardising this - for everyone - is wrecklessness without comparison. Yet enough of us would do it
Were is this complicity in our own demise coming from. Are supporters of ai and the singularity misguided - by their own values - or do we have irreconcileable philosophies on the matter.
My suspicion is that a lot of support for the singularity comes from death anxiety - the way things are, in the long run i'm dead anyway so may as well have a punt (50-50?) on immortality, with humanity as the stakes. That is the singularity is a plausible alternative to an after life. Call me a wanker, but i think that one should take fatality like a man (i.e and die), and not be so selfish.
For me, one good, or at least indisputable reason, is disillusionment with humanity... i would not be able to argue against someone who lived through the trenches in WWI and had such an opinion, that this world as we know it is just not worth it. However I would not agree.
Finally I think a lot of support for the singularity comes down to pure hair-brained optimism and reading too much sci-fi. It actually made me a bit sad to watch the recent star trek movie and its clumsy attempts to make the characters relevant (a sword fight? come on!). Traditional, speculative, science fiction used to be about final frontiers and buckaneering captains, now its struggling to reconcile the future and anything that we might want from a narrative.
Smarter than human AI may be our story is coming to an end ...
The technology to end humanity is already here (nuclear and soon biological weapons). In the long run superhuman AI may be the only thing that can prevent us from destroying ourselves.
Even if it were desirable, stopping the advance of technology is impractical. If superhuman AI is possible, it will be built. A more practical argument would be that superhuman AI should be strictly controlled, though advancing technology will eventually make that difficult too. The only solution in the long (really long) term is for us to become superhuman ourselves and attempt to preserve the things we find important in the transition.
> Were is this complicity in our own demise coming from. Are supporters of ai and the singularity misguided - by their own values - or do we have irreconcileable philosophies on the matter.
This reminds me of an (unintentionally?) hilarious LW post:
http://lesswrong.com/lw/134/sayeth_the_girl/y4i
"For example, most existential-risks activists (scientists doing networking and research about risks like unFriendly AI) are male, and I plan a top-level post to assert that not having reliable access to sex with the kind of sexual partners who can most improve the life of an existential-risks activist should be considered a large disability in a male prospective existential-risks activist -- in the same way that, e.g., an inability to stop rationalizing one's own personal agenda should be considered a large disability."
So there you go. When skynet gets invented, blame it on sexual frustration :).
It sounds like you're equating "AI" with "sentient intelligence with its own values and goals which will only arbitrarily coincide with ours." Certainly, that's a possibility--and one AI researchers would do well to avoid.
But the possible intelligences we can observe on this planet have more variety than that, and the useful intelligences we've created have even more. There's no reason a smarter-than-human AI couldn't simply be a general-purpose tool in the way that a faster-than-human spam classifier is a special-purpose tool.
I look forward to cheap machines that are smarter than man. I fear expensive smart machines which only the richest of the rich will be able to afford.
The first machines will probably be used to make money, and game the market.
...Well, those already exist.
Eventually a good part of our economy may become a capital war for silicon that makes money.
I figured the first would be used to kill, like almost all modern digital technology (digital computers, internet, GPS, etc.)
You're not the only one looking forward to this.
I welcome the day when software are not only living, but also sentinment.
Alas, we're only limited to crafting software system into never needing rebooting while software system are being modified on the fly.
I know it was a typo, but software that's sentimental as well as sentient is probably what we need, as opposed to the vast, cool and unsympathetic variety.
I think most proponents of strong AI include the idea that they should be built on the foundation of a "premise-state" compatible with our human ideals such as intrinsic respect for life, appreciation of complexity and diversity, decentralisation of power, privacy of thought and freedom of action, etc etc.
Which, sadly, demonstrates how far AI as a discipline has to go. I am much more encouraged by the development of Rommbas and similar that seek out an electrical outlet, and so are actually motivated to learn, by something akin to hunger.
Ah yes, a world with networked sentient software controlling the next-next generation of automated labs for making biotech bacteria.
How could that go wrong?
I am a big tech optimist, but sentient software could arguably be classed as WMDs.