by "Spanfeller is a twat" (theaspiringengineer)
Published 11/09/2017 at 16:50
by "Spanfeller is a twat" (theaspiringengineer)
Published 11/09/2017 at 16:50
No Tags
STARS: 1
I was having this discussion with a friend: “Can AI simulate humans?”
Key asumptions:
To simulate something you must understand it.
IA is based on logic and math, IA uses patterns.
Argument timeline:
So far, we’ve used computational power beyond our own to help us understand things faster, think calculators and computers.
Thus only computational power limits our capacity to understand ourselves.
But computational power is only logical, thus, humans must be logical if to be understood by ever advancing, logical AI.
To which she said “We are so illogical we end up being logical”
Expand “We are so illogical we end up being logical”:
Maybe human logic is so complex that at first glance it seems illogical. In order to understand it, machines would need to find complex algorithms and patterns in order to simulate what we understand as illogic.
This would only be true if human thought was finite, thus we arrive at a common question in literature: “Is there truly a unique thought?” If the answer were yes, it means we have infinite thoughts, thus illogical. if the answer were no, then we have finite thoughts and thus we are logical.
To which we concluded a thought experiment (I think)
If a computer AI is given freedom to use other computers to help it compute, and it could understand itself, then the only barrier to understand ourselves is computing power. thus, we are logical.
If it couldn’t, then we are illogical.
"Aaron M - MasoFiST" (amarks563)
11/09/2017 at 16:53, STARS: 0
If you’re interested in brain emulation as a computing problem, read Permutation City by Greg Egan. Definitely sci-fi but gives you a lot to think about.
"Highlander-Datsuns are Forever" (jamesbowland)
11/09/2017 at 16:55, STARS: 2
See the movie Ex Machina, pretty cool.
http://www.imdb.com/title/tt0470752/
"Spanfeller is a twat" (theaspiringengineer)
11/09/2017 at 16:55, STARS: 0
Will! But the issue (at least for now) remains, Is human thought finite? me and my friend think that would be a great indicator (mind you, we’re in first year engineer/psychology. Take what we say with a huge grain of salt.
"Chariotoflove" (chariotoflove)
11/09/2017 at 16:57, STARS: 0
"gin-san - shitpost specialist" (gin-san-)
11/09/2017 at 17:05, STARS: 0
That’s an interesting argument.
On the topic of unique thoughts, how would you argue for human creativity then? There must be unique thoughts since works of art are unique pieces themselves.
There’s also the aspect of human “instinct” for which I don’t know if it’s classified as or grouped with “intelligence.” I think we’ve all had times where our thinking may lead us to a “logical” conclusion but for better or worse we go against it with a gut feeling (which is also a bit of a misnomer since it’s still thinking but with emphasis on emotions/feelings rather than purely information/evidence).
Maybe I’m rambling, but I’m not a professional philosoraptor.
"and 100 more" (nth256)
11/09/2017 at 17:06, STARS: 0
This doubles back into Simulation Hypothesis somewhat, in that all reality is simply data, and matter is a medium of computation, so then we are all just algorithms, just tiny little subroutines in a grand calculation. Reality as we know it is a finite construct within a much larger intelligence.
"66671 - 200 [METRIC] my dash" (66671)
11/09/2017 at 17:08, STARS: 0
I think it’s all about the physical construction, maybe it can come close to a human if you can build the computer similar to a human brain (and maybe the rest of the body??), otherwise, you’re going to set awkward parameters and limitations for something that is capable of not only more, but different as well.
Part of AI is learning, and once it learns by itself the only way to make it learn like a human is to give it a human brain, I would assume, otherwise it would just learn (probably) way faster and not be human at all.
"jasmits" (jasmits)
11/09/2017 at 17:13, STARS: 2
You are using a flawed definition of simulate, to simulate means to “imitate the appearance or character of.” To imitate something you don’t need to understand it, you just need to be able to replicate it. For any input, give the appropriate output. Thus, with enough information a computer can “simulate” a human. However, a computer cannot and can never actually think because it is only doing as it’s told.
A common analogy in computer ethics is the Chinese room. Imagine you are in a room with a bunch of Chinese-speaking people but you do not speak, read, or write Chinese. Nobody is allowed to speak verbally, you can only communicate through written notes. You have a huge book of every possible phrase a Chinese person write to you with the(or a) appropriate response. Every time a Chinese person gives you a phrase, you look it up in the book character for character and copy the response, character for character and give it to them(for the sake of the analogy, assume you are in a booth and they cannot see the book). They will think you can communicate with them and are a fluent Chinese speaker but in reality you have zero comprehension of what you are hearing or saying and have no way of learning to comprehend it. You can only respond how you have been ‘programmed’ to respond through your book. Similarly a properly programmed and powerful enough computer can perfectly simulate a human in every way but lacks intentionality(meaningful, deliberate thought) and can never actually do something it hasn’t been programmed to.
"Spanfeller is a twat" (theaspiringengineer)
11/09/2017 at 17:32, STARS: 0
If the world is simulated then we could hypothetically find the last digit of Pi, because it would have to be saved into a finite memory bank
"Spanfeller is a twat" (theaspiringengineer)
11/09/2017 at 17:38, STARS: 0
Hmm, I understand what you mean. I think me and my friend meant “Emulation”
"and 100 more" (nth256)
11/09/2017 at 17:59, STARS: 0
Possibly, yes. Unless there is some function of quantum computation that allows for truly infinite computational space. Or whatever is beyond quantum computation, that allows quantum computation to exist within our simulation. Some kind of meta-math that describes how math works in our limited reality, but allows for truly infinite yet fully-defined long numbers.
"If only EssExTee could be so grossly incandescent" (essextee)
11/09/2017 at 18:14, STARS: 0
A true AI is by definition a simulation of a human, so yes.
"jasmits" (jasmits)
11/09/2017 at 18:26, STARS: 0
Same thing functionally, emulation means to match by imitation. A computer can emulate/imitate anything, with a big enough sample set and enough computing power. Even if we were illogical, it could mimic that close enough that we wouldn’t notice. It still isn’t actually capable of thought because it’s still just doing precisely as it’s been told, however complex that instruction set is. The idea that computers can, completely on their own accord become ‘more intelligent’ than humans is absurd, they are just better at accessing data. They couldn’t ever ‘want to take over’ or whatever people seem to get afraid of, they don’t want anything it’s just a rock getting zapped with electricity.
"Spanfeller is a twat" (theaspiringengineer)
11/09/2017 at 19:47, STARS: 0
Clearly I need to visit Dictionary.com more often!
But I guess that regardless of my argumentation, you’re right.The truth is that computing hasn’t been able to have will other than the “will” we program into it.
"jasmits" (jasmits)
11/09/2017 at 21:06, STARS: 1
And won’t. It can only follow logic, even if that logic is complex enough to seem like something else. Even programs that can create new programs(for lack of a better explination) are programmed to build new programs how they build them. Anything horrible that a computer does is just a reflection on human beings. And we already know human beings are capable of doing horrible things.
Also the Turing Test is BS, that’s more of a test of a programmer being able to fool the tester than anything else.
"BiTurbo228 - Dr Frankenstein of Spitfires" (biturbo228)
11/10/2017 at 04:43, STARS: 0
I agree with you on the point about ‘imitation’ not being the same as being something, and the analogy of the chinese-lookup device is a perfect one :)
However, I don’t agree with this point: ‘However, a computer cannot and can never actually think because it is only doing as it’s told.’
I don’t believe that in the slightest. Our brains are nothing more and nothing less than massively complex biological computers. There is absolutely no reason to assume that the media something is made of (neurons vs circuitry) has anything to do with whether something is able to think for itself or not, simply its complexity.
Even then, I’m personally of the opinion that people don’t actually have the self-determination that you’re alluding to there. Why would we? We can certainly make decisions by ourselves, but the notion that we actually have ‘choice’ above and beyond sticking stimuli into our biological processors and spitting out an action is in all likelihood a fallacy to make ourselves feel special.
It’s more likely that the mechanisms by which our circuitry produces an action from any given stimulus is simply complicated enough that we don’t understand it, so we have this illusion that there’s some higher ‘self’ that makes decisions all of its own accord. The other option is that there’s something magically special about us compared to other similar mechanisms which is, in all likelihood, nothing more than human exceptionalism.
"Aaron M - MasoFiST" (amarks563)
11/10/2017 at 08:06, STARS: 1
The book takes a different tack, which is a little less philosophy and a little more engineering. We know that the brain is made up of neurons, and vaguely that different neurons are specialized for different tasks. We also know, from studies of people with traumatic brain injuries, that neurons are at least somewhat elastic and can do things other than what they were originally intended for. Furthermore, we have some understanding of the signals that neurons send, to the point where we can use those signals to, say, produce a computer output of what an eye is seeing. Not well, but well enough to gain an understanding of how vision works.
In theory, a good enough emulation of a neuron could eventually lead to a good enough emulation of a brain. But would this emulation lead to consciousness? We have no idea. Another book which explores the brain in the jar simulation a bit is Walkaway , by Cory Doctorow...in that case though, the brain uploading discussion is not primary to the plot (in Permutation City it definitely is) at least for the first half of the book. Still a great book though.
"jasmits" (jasmits)
11/10/2017 at 13:43, STARS: 1
I do agree that a better argument seems to be if we actually have intentionality or if we only have an illusion of intentionality.