Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Eisensturm]

i very much doubt it will go that fast.

But, apparently, the experts do.

Correction. According to that article:

"SOME experts do. Then theres plenty more camps of thinkers."

[Lord Shonus]

i very much doubt it will go that fast.

But, apparently, the experts do.

Correction. According to that article:

"SOME experts do. Then theres plenty more camps of thinkers."

Most notably, that article's poll consists almost exclusively of people that have made developing such an AI their life's work. Mainstream computer science answers the question of "When will a godlike super-AI exist" with a range from "no less than a century or two given the fundamental breakthroughs that would be required" to "it is a completely impossible quack theory, just like Alchemical Transmutation or FTL by acceleration."

[MetalSlimeHunt]

It's a completely implausible quack theory. Flat out. Going in any direction with it you rapidly run into a hard problem that suggests we're less "kept away from the goal" and more "don't even know where/what the goal is or if we're even on a racetrack to begin with".

Chinese Room
Intelligence Explosion
Medium Limitation
Problem of Consciousness
Nature of "Thought Directives"

This is but a short list of everything, everything humanity would have to conclusively answer for a singularity AI to even be possible, much less actually created.

The singularity is bullshit and it will never happen, Roko's basilisk is blind, and everyone who makes their dramatic sci-fi predictions for the future always sets them in their remaining lifetime.

[Graknorke]

The Chinese Room "problem" is a terrible objection to AI. It's clear from the description of it that the system of instructions contained within the book is what understands Chinese and can hold a conversation.

[Spehss _]

Searle's Chinese Room argument isn't exactly the best choice for debunking the possibility of AI. Searle tries to argue that making an AI as competent as a human is impossible because the biological human brain has some "stuff" that makes it work the way it does which an AI cannot have. Searle makes no attempt to scientifically quantify this "stuff" and it's for all intents and purposes just a person saying "but muh brain is more special than a computer".

Also, in the Chinese Room argument, if the guy in the room meant to translate chinese is supposed to be representative of an ai program made to translate chinese, but nothing else, then the ai would be a "narrow" ai or "weak" ai which is not at all meant to be  as intelligent as a human. It's a program written to do its job, which in this case is translate chinese to english.

edit: ninja'd by grak.

[Lord Shonus]

Also, in the Chinese Room argument, if the guy in the room meant to translate chinese is supposed to be representative of an ai program made to translate chinese, but nothing else, then the ai would be a "narrow" ai or "weak" ai which is not at all meant to be  as intelligent as a human. It's a program written to do its job, which in this case is translate chinese to english.

That's the entire point of the argument. The Chinese Room problem posits that -just as there is no way to tell the difference between a fluent Chinese speaker in a box and someone with zero Chinese skills and a good Chinese-to-English dictionary in a box- it is impossible to tell the difference between a "strong" AI and a "weak" one with any basic tests because all you can do is record what it puts out, not determine whether or not it is capable of really understanding the data being processed and use that to actually think.

[miauw62]

also we're already hitting quite a few barriers with regards to processor speed. i doubt that anything short of a big parallel supercomputer would even be able to run a powerful AI in the first place.

[MetalSlimeHunt]

Moore's Law has escaped its fate more than once before, and we have reason enough to believe it'll do so at least a few more times, but computronium it ain't.

[Spehss _]

That's the entire point of the argument. The Chinese Room problem posits that -just as there is no way to tell the difference between a fluent Chinese speaker in a box and someone with zero Chinese skills and a good Chinese-to-English dictionary in a box- it is impossible to tell the difference between a "strong" AI and a "weak" one with any basic tests because all you can do is record what it puts out, not determine whether or not it is capable of really understanding the data being processed and use that to actually think.

If the chinese room AI can do anything besides its primary reason for being programmed - to translate chinese to english - then it's a strong AI. Otherwise it's a weak AI that's good at what it was made for but bad at anything else it wasn't made for.

also we're already hitting quite a few barriers with regards to processor speed. i doubt that anything short of a big parallel supercomputer would even be able to run a powerful AI in the first place.

There's been engineering and science and stuff working on computing with things besides silicon chips. Silicon chips are the current step in computing before finding the next big thing to work with, like vacuum tubes were before silicon chips.

[Knit tie]

What about the other problems MetalSlime listed? Assuming that computing speed is not a problem - i.e. we've got a hypothetical computronium processor that is infinitely powerful - is creating something that will be, as the article states, "thinking on a level that we can't possibly comprehend" even theoretically feasible, or is it just a case of people trying to promote their own line of work too hard?

[scrdest]

What about the other problems MetalSlime listed? Assuming that computing speed is not a problem - i.e. we've got a hypothetical computronium processor that is infinitely powerful - is creating something that will be, as the article states, "thinking on a level that we can't possibly comprehend" even theoretically feasible, or is it just a case of people trying to promote their own line of work too hard?

Well, do bear in mind, humans are thinking on a level we cannot comprehend, at least currently.

[Facekillz058]

I've been doing a lot of heavy roleplaying lately.
Had to have a real life conversation today and the word "Aye" just came out.
Everyone looked at me funny and now I have to consciously control my speech, T.T

[TD1]

Come to Northern Ireland. I regularly say aye in my speech, though typically leave it out of my writing.

[MagmaMcFry]

What about the other problems MetalSlime listed? Assuming that computing speed is not a problem - i.e. we've got a hypothetical computronium processor that is infinitely powerful - is creating something that will be, as the article states, "thinking on a level that we can't possibly comprehend" even theoretically feasible, or is it just a case of people trying to promote their own line of work too hard?

Consider the problem of how to get the AI up to the intelligence, knowledge and wisdom level of an average human in the first place. How do we get humans up there? By letting them muck about in society for thirty years. How do we get AIs up there?

[Tack]

Wanna do bunches of forum game stuff.
Phone unsuitable for text dump.
Must wait hours for next computer availability.

i wouldn't mind being ridden tbh

You're a bear? I thought you were skinny Canadian.