Comic Discussion > QUESTIONABLE CONTENT
WCDT: 2281-85 (24-28 September 2012) Weekly Comics Discussion Thread
Carl-E:
Ok, first up; I'm a doubled square, so I think that I'm twice madness age, and probably a square and a year past skewbrow... and that should be enough for anyone to figure out all three!
Of course, I could be wrong about the ages of my mathematical compatriots, but you can at least figure out what I think our ages are...
Akima, interesting that you cited John Henry - the point of the song was that he won, though it was a pyrrhic victory, at best. The steam drill failed, as I think we know all technology does at some point. We've gotten used to fewer and fewer catastrophic failures, but even so, were an AI like Station to fail, the results would be a disaster. I imagine there's some pretty sturdy redundance built into a system like him. Quite frankly, it's rather amazing Pintsize has lasted so long!
And those computer game players are using their full computational power to beat us. But when asked, chess masters admit to not thinking more than a few moves ahead, much less than the computers that play the game. There's something else going on, and allegedly that's what's been tapped into in the development of AI - otherwise, something with as little computational power as Pintsize wouldn't be able to pull of even a simple practical joke!
It's an interesting situation, and I think we're just sitting here poking it and finding all the holes, but it does make for some good storytelling!
jmucchiello:
--- Quote from: Akima on 30 Sep 2012, 04:07 ---As for c), I'm sure there would be other applications of Computer Science than AI development, but why would we assume that humans would be as good at any aspect of the field as AIs? Because humans are special? With the advent of truly sentient, self replicating artificial intelligences, maybe not so much. Some areas of CS, like compiler-design, might become completely redundant. Why would an AI need a compiler at all? Once upon a time even humans wrote machine code directly, and it doesn't take much digging to find grey-muzzled old programmers who will assure you that they wrote more efficient code that way too!
--- End quote ---
True, but they also did not write program of nearly the same complexity of programs written today. Imagine implementing a HTML renderer in assembly language that could handle all flavors of HTML and run against any video driver. It could be done but I can imagine it would be very amenable to changes.
But I think this ignores the problems of hard AI and forgets its limitations. Just because the AI is running on computer hardware, doesn't mean the thinking AI is aware of the computer hardware. You, presumably, run your intelligence on a wetware brain that you are mostly unaware of.
To multiply two multi-digit numbers, an AI might perform the calculation "by hand" just like a human would: reducing the problem to n "multi-digit times single digit" problems and then summing the results to get the answer. Performed in this manner the AI will not be as faster than a human as the speed of its microprocessors would imply.
Other AIs might have the ability to spawn off a process to execute an interpreter in their hardware (like Python) and interact with it to get an answer (just like a human might go to computer and launch an interpreter to get the answer, or pick up a calculator to get the answer) and so it might seem like they can perform "faster" than a human. But that has nothing to do with being and AI and has more to do with have a computer "close by" that they can interact with intelligently.
--- Quote ---Or maybe not. Even in our world, where computers are well short of sentience, the arrival of computers that can beat the strongest human chess players some of the time, has not stopped people playing chess. Even if computers were developed that could defeat all human players in every game, people would probably still play against each other, and by studying the games of the computer champions improve their own play. There are, after all, human weiqi players I would have no chance of beating, and yet I still play and enjoy the game, and study the games of the professional champions. If (or rather when) computer players can reliably defeat those human champions, I doubt I will give up the game. Humans compete in marathon foot-races despite being wholly uncompetitive over that distance with motor cars. Perhaps we don't have to be the best at something to find it a worthwhile subject of study.
--- End quote ---
I think the mere existence of the Para-Olympics shows that mankind is competitive even in the face of knowing there are opponents whom one cannot beat.
Weiqi is generally called Go in English. English readers have a better chance of recognizing the name Go than the name weiqi. Though the game is not really popular at all in the States by any name.
Pilchard123:
--- Quote from: Carl-E on 30 Sep 2012, 10:25 ---There's something else going on, and allegedly that's what's been tapped into in the development of AI - otherwise, something with as little computational power as Pintsize wouldn't be able to pull of even a simple practical joke!
--- End quote ---
What if all companion AIs were permanently linked into an AI cloud and did the majority of their processing there, with just enough local hardware to function on a basic level?
Carl-E:
Interesting idea - perhaps that's this "creche" Jeph has mentioned.
Skewbrow:
I don't think there is any compelling reason why AI would understand, for example, algorithmic complexity theory any better than humans. True, they can do arithmetic and number-crunching at blinding speed, but that's not what that part of computer science is about (me thinks). And, yes, even if they were better at it, the humans would still compete with each other.
Akima also raised the comparison between chess playing and go/weiqi playing computers. The former being much better - at least when pitted against human champions. The reasons for the difference originate simply from the fact that a game of go branches much faster than chess. (Sorry about using the Japanese name here. I used to play some with two friends, but unfortunately they have both moved away. We used the Japanese point counting rules, so that's what we called the game. In these parts at least it is much better known by the Japanese name.) Somebody who understands it better than I do also told me that if computers ever became proficient at handling the standard 19x19 board, humans could simply move on to playing with, say a 23x23 board, again leaving a program relying on number crunching largely helpless. I have only seen a pitiful (=inexpensive) go-playing program from early 90s. I'm sure they have improved much since, but cannot comment. Hopefully we get to witness a China vs. Japan go/weiqi-playing computer design competition. It could get nasty (but I would prefer that to them fighting over those tiny islands).
If I interpreted Carl-E:s phrasing of the age(s) problem correctly, it is a simple Pell equation. Finding the algebraic integers of norm minus one in the field Q joint root two is straight forward. As I obviously won't be celebrating my 1st nor 1681st b-day this week, the solution of that equation is, indeed, unique.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version