Why do people believe superintelligent AI could solve all our

File: AI.png (262KB, 697x534px) Image search: [Google]

262KB, 697x534px

>>7846244

>>7846260
>people keep posting an image of a book they probably didn't even read themselves

Well excluding the magical stuff, a lot of our problems as a society can be solved be a smarter problem solving, state budgets for example, there are many factors, some of them are interwined some are not, over all this is a kind of task AI's are good at, AI doesn't have to be perfect, and by AI I don't mean the conscious bullshit, it's simply a program that preforms a task that wan't explicitly programmed and improves somewhat with experience, AI can be exremely helpful in the future, but it's not the second coming of jesus.

>>7846244
Too much Terminator/space odyssey bullshit

>>7846314

It's a good overview for non-experts who don't know that much math

>>7846326
>state budgets for example
nah, I dont think its about beeing smart. Its more about having connections, bringing people onto your side and finding solutions that everybody can live with

>>7846326

The special thing about machine learning and AI is that it's theoretically posibble for it to make valid decisions to improve that before being tackled by the AI weren't really a possibilty, we're not talking about brute forcing here, some machine learning algorithms (networks) expose a set of moves that were previously thought to be useless, for example when trying to tackle a game like pacman for example, AI algorithms would expose bugs and special moves like going through the ghosts, which is benefitial to the score.

Well, Super AI will understand all of the rules of any countries and aply this without any mistake

>>7846342

Well smart may be the wrong word, but broad "insight" about all the different many factors and how they interact together and affect the final result, is closer to what it is.

Not to mention that eliminating the human factor is an important one, no (human) errors, no corruption, feedback from previous failures is what makes a machine more attractive (to me) in that kind of position

>>7846244
Because people believe problems can be solved by intelligence, that there aren't problems that are simply impossible.
So if there was something super intelligent, it could super solve problems, right?

>>7846364
then maybe we shouldn't call such A.I. superintelligent, but supergood with people or something.
The implication here is that intelligence is the most important skill to find good solutions, which is simply not true

>>7846357
what if the rules are inheritly flawed, because they were designed by humans?

>>7846384
Then, let the AI be our "slaves" and never use their capacities, the both possibilities are possible, they have their advantages and their disadvantages, it's an endless debate

>>7846409
I dont think refering to AI as slave makes any sense.
It is not human, it doesn't have the desire of freedom or power or love or whatever.
It does what it is programmed to do

>>7846364
I'm really not sure whatever AI we could create would be in mesure to model the complexity of a world filled with billions of humans.

But anyway, seeing how humans are, I just don't think their is a way to "solve our problems". We are the problem. Whatever happens, some people will like it, and some people won't. Some will go angry and break stuff, etc ...

>>7846420
Absolutly right but a judge is not supposed to feel love or whatever that looks like love, it must be objective and fair,so let AIs make what they are programmed to do, humans are intelligent and probably can program a judge robot or anything else which can do what he is made to

>>7846378

Define intelligence? The whole reason "Artifical Intelligence" is a term that exists is becuase it's a sexier title that is more likely to get funding, the alternative, "statistical dynamic decision planning" isn't as likely to get $$$ thrown at the researchers.

>>7846460
>Define intelligence?
The ability to solve IQ tests

>>7846244
"Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an 'intelligence explosion,' and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control."
- I. J. Good

>>7846602
>implying more intelligence doesn't require more computation power
>implying you can keep on creating more and more intelligent A.I. with a limited amount of resources
>implying there is not a point of diminishing returns where amount of resources to create an more intelligent A.I. grows exponentially
I.J. Good sure made a lot of implications

>>7846482

If thats the case then we are closer than you think.

http://www.gizmag.com/artificial-intelligence-algorithm-learning/41448/

https://www.technologyreview.com/s/538431/deep-learning-machine-beats-humans-in-iq-test/

http://www.sciencedaily.com/releases/2012/02/120214100719.htm

http://gizmodo.com/an-ai-program-in-japan-just-passed-a-college-entrance-e-1742758286

http://techxplore.com/news/2015-10-ai-machine-iq-score-young.html

>>7846615
He said AI intelligence will leave the intelligence of man far behind, not that it will shoot upward forever.

>>7846643
Well, my problem is, I just dont see how he can take for granted that beeing more intelligent than man would make the machine much more capable.
If you look at our world today, its not the most intelligent people who have the most power, so it doesn't seem like intelligence is the most important skill

>>7846676
>If you look at our world today, its not the most intelligent people who have the most power, so it doesn't seem like intelligence is the most important skill
I'm pretty sure intelligence correlates strongly with power. But there's also the issue that variance in human intelligence probably isn't very big compared to what an artificial intelligence could achieve.

The most powerful chimpanzee might not be the most intelligent one, but the reason that humans rule this world and not chimpanzees is definitely our intelligence. And the difference between the highest intelligence a self-improving AI can achieve and the intelligence of the smartest human could be much, much greater than the difference between the smartest human and the smartest chimpanzee.

>>7846326
>a lot of our problems as a society can be solved be a smarter problem solving,

The only problem is humanity itself. It is one greedy mother fucker and will fuck up any utopia you dream up in your hippy weed stupor.

>>7846676
>Well, my problem is, I just dont see how he can take for granted that beeing more intelligent than man would make the machine much more capable.
>If you look at our world today, its not the most intelligent people who have the most power, so it doesn't seem like intelligence is the most important skill
What do you think intelligence is? For one thing, it includes such things as charisma, political skills, networking, and "people skills" in general. In that light, the people with a lot of intelligence *absolutely* are the ones holding power in today's world.

>Why do people believe superintelligent AI could solve all our problems and rule the world and do all other sort of magic stuff?
Intelligence is LITERALLY (though not exactly) the ability to accomplish arbitrary goals and solve arbitrary problems; the more arbitrary the problems you can solve and the goals you can accomplish are, the more intelligent you are. A superintelligence being able to solve all our problems is a trivial consequence.

>>7846861
Define "arbitrary"

>>7846861
I like the definition of intelligence as the ability to solve problems efficiently. Everything else being the same, if you can come up with a better solution to a problem than the other guy, or if you can solve the problem faster or with less resources, you're smarter.

>>7846867
>Define "arbitrary"
No need. You can flesh it out in more detail by saying that the more goals you can accomplish, the more intelligent you are; it's a quantitative thing. (You could define an exact measure if you really wanted, but that does not really change the key point so why bother.) "arbitrary" is just an informal description of the above.

>>7846615
Do you think human brains are the most intelligent organization of matter that our universe allows? Y/n

Y -> You must be kidding
N -> Good's implications are fine

>>7846482
IQ is not applicable to computers. It would be just like the chess playing computer, only able to do one thing.