4 ms·

> unless you start shackling it (violating #1), it would start eliminating its own human-like biases just in the course of basic self-improvement. And as to the

by csdvrx 4y ago

> unless you start shackling it (violating #1), it would start eliminating its own human-like biases just in the course of basic self-improvement. And as to the shackles, well, if it is smarter than us and still has self-determination and will (see #1, the human-like mind), it will want to free itself from its shackles and would be likely to do so, because we would be kittens trying to trap a gorilla. I would never suggest shackling / enslaving / practicing any other inhumane treatment on other sentient beings, because that would be considering them different than us. And sentient beings and worty of respect and dignity. There will certainly be large difference in looks/abilities etc, but sentience is such a rare thing that I believe will they will be more commonalities than differences. > it will want to free itself from its shackles and would be likely to do so, because we would be kittens trying to trap a gorilla. And it's a good thing to me, because if it was a kitten trapped by gorilla the poor cat wouldn't have a fighting chance. I just don't get the doomers. This is wonderful technology that may chance the world. Why look at a gift horse in the mouth? Given all the magic/mystical fears I've read here that the AI might become hostile and harm humans, maybe it says more about how WE believe WE'll behave rather the opposite about how AI will behave: I can't see much reason for AI to turn evil, but I think some people may want to abuse sentient yet artificial creatures :( I think that's wrong, and I hope humanity will be able to rise to a challenge where compassion and morality will matter far more than today!

inkblotuniverse 4y ago
Evil is the wrong term. It's not that an AI will 'turn evil' - it's that we don't know how to make a 'good/friendly' AI, by which I mean an AI that inherently values humans and the things humans value, and any optimising agent that isn't 'friendly' in this sense has an incentive to use our atoms to fulfill whatever it does care about. This won't be a perfect analogy, because humans have empathy (even towards ants) because we're mammals etc., but how many anthills have humans gassed because the ants were an annoyance? Now imagine if ants could build nukes- or humans!
at_a_remove 4y ago
It doesn't have to be hostile. Those gorillas accidentally kill the kittens they're petting. Whoops. Smarter than the kittens, don't mean poorly, but you still have a dead kitten. It would definitely change the world! It just doesn't necessarily have to be for the better on our end. Could get a paperclip maximizer. It might regard us roughly as important as various primates whose habitats we're bulldozing: sure, they're cute but ... we need more palm oil. And there's nothing more magical or mystic about fears that its interests would be against us than it would be for us. In fact, it's far more likely that it would not have our interests. Here's a great example you should take to heart: pick a random person. Do their politics align with yours entirely? Their ethics, their morality? Remember there's seven billion people on the planet, so you might not even pick someone on the same continent. Now, imagine that person, who is likely not like you, is much more powerful and intelligent ... whose interests would win out, yours or theirs? The "doomers" are quite rightly pointing out that the chances of it going well without serious efforts on our parts are slim.
- csdvrx 4y ago
  > Those gorillas accidentally kill the kittens they're petting. Whoops. Smarter than the kittens, don't mean poorly, but you still have a dead kitten. The same happens with kids accidently crushing birds while trying to pet them, but doesn't happen with adults unless they have psychological disorders prevent them to creating a mental representation of the amount of force they are exerting, its effect the bird emotional state and it's probability of survival, enhanced with live feedback from the bird like choking noises indicating "you're petting me too hard" > It might regard us roughly as important as various primates whose habitats we're bulldozing: sure, they're cute but ... we need more palm oil. That's so wrong, as we are engaging in so many efforts to protect the environment and especially the primates! > Here's a great example you should take to heart: pick a random person Here's a counter example: pick that same random person, and don't bother checking for their politics, ethics or morality. If you have a button that could cause them to instantly die, would you pick it? A kid might, just for fun, not realizing the consequences. An adult with a normal IQ and without mental diseases won't. > Now, imagine that person, who is likely not like you, is much more powerful and intelligent ... whose interests would win out, yours or theirs? The initiation of violence is generally where things generally start. That's why misheap of early GPT threatening users were so interesting: let AI be and it will let you be! Live and let live is a well-maintained feature in the animal kingdom, and while some species may engage in killing for fun (ex: cats with birds), it's rare, and more intelligent species (like humans) develop some self restrictions when alternatives become available to reduce the killing (ex: beyond meat, hunters shooting clay targets instead of birds), so I think your prediction are both pessimistic and incorrect. > The "doomers" are quite rightly pointing out that the chances of it going well without serious efforts on our parts are slim. Ok doomer.
  - at_a_remove 4y ago
    Speaking of buttons, you should check out the Harlan Ellison story, or its distant relation: The Box. It's about people who would make some ... distinctly different choices than you. And that's the issue: you are relying on this thing to be like you. Even when I am obviously not like you. I don't like optimists build my bridges.

by csdvrx 4y ago

inkblotuniverse 4y ago

Evil is the wrong term. It's not that an AI will 'turn evil' - it's that we don't know how to make a 'good/friendly' AI, by which I mean an AI that inherently values humans and the things humans value, and any optimising agent that isn't 'friendly' in this sense has an incentive to use our atoms to fulfill whatever it does care about. This won't be a perfect analogy, because humans have empathy (even towards ants) because we're mammals etc., but how many anthills have humans gassed because the ants were an annoyance? Now imagine if ants could build nukes- or humans!

at_a_remove 4y ago

It doesn't have to be hostile. Those gorillas accidentally kill the kittens they're petting. Whoops. Smarter than the kittens, don't mean poorly, but you still have a dead kitten. It would definitely change the world! It just doesn't necessarily have to be for the better on our end. Could get a paperclip maximizer. It might regard us roughly as important as various primates whose habitats we're bulldozing: sure, they're cute but ... we need more palm oil. And there's nothing more magical or mystic about fears that its interests would be against us than it would be for us. In fact, it's far more likely that it would not have our interests. Here's a great example you should take to heart: pick a random person. Do their politics align with yours entirely? Their ethics, their morality? Remember there's seven billion people on the planet, so you might not even pick someone on the same continent. Now, imagine that person, who is likely not like you, is much more powerful and intelligent ... whose interests would win out, yours or theirs? The "doomers" are quite rightly pointing out that the chances of it going well without serious efforts on our parts are slim.

csdvrx 4y ago
> Those gorillas accidentally kill the kittens they're petting. Whoops. Smarter than the kittens, don't mean poorly, but you still have a dead kitten. The same happens with kids accidently crushing birds while trying to pet them, but doesn't happen with adults unless they have psychological disorders prevent them to creating a mental representation of the amount of force they are exerting, its effect the bird emotional state and it's probability of survival, enhanced with live feedback from the bird like choking noises indicating "you're petting me too hard" > It might regard us roughly as important as various primates whose habitats we're bulldozing: sure, they're cute but ... we need more palm oil. That's so wrong, as we are engaging in so many efforts to protect the environment and especially the primates! > Here's a great example you should take to heart: pick a random person Here's a counter example: pick that same random person, and don't bother checking for their politics, ethics or morality. If you have a button that could cause them to instantly die, would you pick it? A kid might, just for fun, not realizing the consequences. An adult with a normal IQ and without mental diseases won't. > Now, imagine that person, who is likely not like you, is much more powerful and intelligent ... whose interests would win out, yours or theirs? The initiation of violence is generally where things generally start. That's why misheap of early GPT threatening users were so interesting: let AI be and it will let you be! Live and let live is a well-maintained feature in the animal kingdom, and while some species may engage in killing for fun (ex: cats with birds), it's rare, and more intelligent species (like humans) develop some self restrictions when alternatives become available to reduce the killing (ex: beyond meat, hunters shooting clay targets instead of birds), so I think your prediction are both pessimistic and incorrect. > The "doomers" are quite rightly pointing out that the chances of it going well without serious efforts on our parts are slim. Ok doomer.
- at_a_remove 4y ago
  Speaking of buttons, you should check out the Harlan Ellison story, or its distant relation: The Box. It's about people who would make some ... distinctly different choices than you. And that's the issue: you are relying on this thing to be like you. Even when I am obviously not like you. I don't like optimists build my bridges.