When I read an article called "Man VS. Machine," in the Atlantic, which took the position of this thread title, I thought, "Another one bites the dust"--one more prediction of what technology could never achieve, proven wrong sometime in the near future. After reading the article I give more credit to the writer's argument. When we think about our minds, it's not really the power and sophistication that are so distinguishing, but the aspects that depend on our interactions with each other, our sensitivities about what we're experiencing in any moment. We haven't given full credit to his side of us as we've been engaged in ramping up AI. Ironically, now that we've been able to create computer programs that have some level of success in fooling judges into thinking they're humans, we've also been taken back to what is best and unique about our humanness, and that isn't the strictly analytical part. It certainly isn't the information storage part, either; we've known that for a long time. So the writer, Brian Christian, says that advancing AI projects tends to make us more appreciative of our own humanity.

Christian was the winner of "The most Human Human" award at the 2009 Turing Test (actually the Loebner Prize competition). There's an award given to the confederate, as the human conversers are called, who does the best job of showing that he's human. The more famous award is the one for the "Most Human Computer." In 2008, the computers came close to passing the Turing Test, which would be when at least 30% of judges picked computers as humans. In 2009, though, none of the computers were able to fool a judge. This leads Christian to disagree with a statement by the co-founder of the competition: "One thing is certain: whereas the confederates in in the competition will never get any smarter, the computers will." Christian disagrees with the first part. The results of 2009 show that. He himself had studied previous transcripts of the 5-minute exchanges. He figured out how he could highlight the weaknesses of computer-generated responses through consciously emphasizing what humans really do when they talk to anyone. It worked. Even if computers pass the Turing Test on any given year, it's not as though a milestone has been permanently passed. Humans could deny them the prize the following year. Christian believes this is what could have happened with Big Blue, the chess program that defeated Garry Kasporov at their second confrontation in '97. Kasporov proposed a rematch, but IBM got edgy and dismantled the program.

http://www.theatlantic.com/magazine/arc ... hine/8386/

...much less the Chimpanzee mind!

Right, that's the video that johnson posted, and it's amazing. But I think a computer could be made that was better at instant recognition.

oops... didn't realize it had already been posted.

When we think about our minds, it's not really the power and sophistication that are so distinguishing, but the aspects that depend on our interactions with each other,[...]back to what is best and unique about our humanness, and that isn't the strictly analytical part.

Yes, what is distinct about human interaction?! #1 is it that we can talk? How did this come about? I can see how a flying bird might evolve from a non-flying bird in that those who could jump the highest would be the ones most likely to leap to safety and reach more food than those of their less gifted relatives - so I can see how a result (in this case, flying) could result from a feature not directly associated with it (arms/wings to allow for greater jumping ability) but what is the rationale for the development of the human capability of speech? Why hasn't the chimp developed it? Why not, say, cats? I'm asking because it's my opinion both these characters could whip humans in IQ if only they could talk amongst themselves. They already do communicate of course, but I mean talk. I'm thinking now of Rex Harrison's Dr Doolittle speaking ability. Birds in a given area will know which humans to be particularly be wary of prior to their ever having seen them... and speaking of birds, what's the deal with the vocal chords of parrots and their close kin?

There is (IMO anyway) an interesting book by Justin Leiber (the son of his more famous father, Fritz) called Can Animals and Machines Be Persons? My answer is but of courserous! Well, what makes a person? Are all humans persons? Are no animals persons? Are no machines persons?

I am rambling... EDIT: Dr Doolittle has a fine soundtrack! My favorite 60s musical probably... EDIT 2: Just yesterday I beat the Windows chess program at the level 5 seting for the first time!

There's a common misconception which extends to probability and other scientific subjects, and which is summed up in the "infinite monkeys" problem, often phrased thus: if we give an infinite number of monkeys a typewriter each, eventually one will type the complete works of Shakespeare.

While this seems superficially true, it's based on a huge assumption: that the monkeys are physically capable of typing the complete works of Shakespeare. If they're not, it doesn't matter how long we wait, it will never happen.

The same is true of the AI question: although computers model how our brains work rather well in some respects, there are others in which they are very dissimilar indeed. There is no guarantee that any computer with our current architecture model will be capable of human-like intelligence; it might be impossible due to the very nature of the machine.

The Deep Blue experiment was a great example: that machine didn't really "play" chess. All it did was calculate all the possible games from the current position then choose a move which was part of a game in which it would win. That's impressive, but it's not how a human being plays chess, which is an altogether different and more complex activity.

Squelch wrote:The same is true of the AI question: although computers model how our brains work rather well in some respects, there are others in which they are very dissimilar indeed.

Hello. And the flip side is that our brains model rather well how computers work. I think another assumption often made, (and one of a terribly sour note!), is that an AI or an animal is the inferior of a human when it doesn't model our process of thinking. It doesn't "play" like us therefore it is considered to be something of a parlour trick. And if it nonplays to a higher level than we play we'll accuse it of cheating! Some humans will never be capable of human-like intelligence. They still think in their own way. Well, I think, there is a vicious reductionism at play here! So-and-so doesn't play, but he sure does calculate. Mr Jones is a machine. Mr Jones is an animal. Good? Bad? Anthropomorphic poobah? That old mystic Descartes had some funny ideas about man, machines and animals. He worked himself into a box however, in-between his bouts of butchery and bloodlust, and arrived at a solution - when you have no other alternative call on God. Is humanity a sore loser? Here is what humanity is the undisputed champion of: not chess, rationalization.

There is no guarantee that any computer with our current architecture model will be capable of human-like intelligence; it might be impossible due to the very nature of the machine.

The Deep Blue experiment was a great example: that machine didn't really "play" chess. All it did was calculate all the possible games from the current position then choose a move which was part of a game in which it would win. That's impressive, but it's not how a human being plays chess, which is an altogether different and more complex activity.

I think another assumption often made, (and one of a terribly sour note!), is that an AI or an animal is the inferior of a human when it doesn't model our process of thinking.

Yeah, I think a lot of people make that mistake. Even rather modest desktop computers have extraordinary, prodigious abilities of calculation and logic which are far beyond any human being in sheer speed and scope. In fact I'd take this a step further: just as it's possible that computers might be incapable by their very nature of reproducing human thought, it might be possible that human minds mind be incapable of reproducing computers' abilities.

That sounds facile but consider this: supposing human beings were discovered to be incapable by nature of understanding the fabric and structures of the universe. Let's imagine that it requires a type of super-logic or super-mathematics which human brains are physycially incapable of processing. What then if computers were capable fo processing it? Where would that leave us?

It doesn't "play" like us therefore it is considered to be something of a parlour trick.

When we say "play chess", we refer to the combination of skill, intuition and experience that human players use. Deep Blue wasn't capable of any of that. That's not pejorative, it's a statement of fact.

The Deep Blue experiment was a great example: that machine didn't really "play" chess. All it did was calculate all the possible games from the current position then choose a move which was part of a game in which it would win.

Uhh... This is not really correct. As it happens, even Deep Blue was unable to calculate all the lines to the end. Heuristics and positional analyses were always necessary until the very end of the games. You underestimate the difficulty of this problem immensely, I fear.

Just a further thought on this topic, the question isn't really one of machines 'beating" human minds, but being made to be like human minds. If you take the properties of human minds--properties that can be troublesome for us, but that we still value--you would never want to create a machine with the same properties. Emotional, intuitive, unpredictable machines would be pointless ways to accomplish work. The ways in which computers have bested human minds are already evident, and though it's hazardous to say this, the advantage of computers may never change in kind, only in degree.

"Never" is a long time...

Beat which human mind? At what task?

As soon as you specify the problem at issue, I think you only have to wait a few years before you either have to re-specify the problem more narrowly, or accept defeat. Thus, what counts as "intelligence" is an ever-shrinking set of abilities. Even if computers/programs only manage to achieve above average "intelligence" in some key fields, those talents, once aggregated under one "suite" of programs, interlinked, will certainly seem intelligent to the average person. It will be the most intelligent "person" most people will have much of a chance of conversing with, I'd imagine.