Slashdot Mirror
Comparing Man and Machine?
An anonymous reader asks: "Today Garry Kasparov's last of 4 chess matches with the computer X3D Fritz ended in a draw. The totals of all 4 games leave the two opponents tied 2 to 2, revealing that even though the technology has advanced significantly since Kasparov was beaten by IBM's Deep Blue in 1997, the odds are not always on the side of brute computational power. This leads me to pose the question: is chess really a viable way to test whether man or machine is truly superior? Until AI becomes flexible enough to challenge us in arenas like art and music, what would be a better real-life competition?"
Obviously, the steel-cage no holds barred death match is the answer.
I'm sure someone will mention Go, just because it's always said that it's much harder to get a computer to do. But I don't see the point of playing games against computers anyway, what does it really prove? It proves that the data fed in to it (from hundreds of sources) and the (several) programmers are better at logically defeating a game than ONE person is. I'm amazed he won two games, this says far more about him than it does about anything else.
What exactly are these contests trying to prove anyway? When the computers gain a clear victory over the humans, what have we learned?
This isn't a test of whether man or machine is superior. This is a test of whether man or machine is superior at chess.
Eventually machine will probably always be superior. For now they're about equal.
"Moderate drinking can help prevent amputated limbs" -- Abigail Zuger, NYTimes, 12/31/02
I'd like to see something that really measures intelligence, and I don't think chess necessarily does that.
:) ) intelligence is when something is self-aware and can develop NEW thoughts (ie. learn). Computers aren't really smart and likely won't be for a long time because they don't do anything NEW. All they do is to go through pre-defined algorithms, and use pre-defined techniques. Even so-called learning algorithms are very primitive and very little new thoughts are generated.
I would say (this is just my definition; take it for what it's worth
If someone can design a computer that comes up with totally new thoughts, it is intelligent in my book. So, instead of just using existing algorithms, if a computer can create a totally new algorithm, it is intelligent...
Having said that, if computers become intelligent, we'll have all sorts of problems...
Testing a human vs. a machine in terms of chess is really, as far as I can tell, a way to see whether or not that specific computer (program) can beat that specific human at chess at a specific point in time. It doesn't tell us that much about superiority or the intelligence of one over the other. I would say once we understand what intelligence is really really well, then we can start to say these sorts of things. But intelligence doesn't equal who wins the game of chess.
Can the computer cook an omelete with the ingredients it has on hand in the fridge? Can it change the baby's diaper when it is crying? Can it say soothing things to the baby when it is crying, and remember that the baby hasn't slept well the last few days, so it should probably give it some extra whatever, and then can the computer give its mother-in-law a call and gracefully decline and invitation to hang out and play bridge or something, using a little white lie? Can you tell me what intelligence is?
damn beat me to it.
Turing Test
for me it just reveals that ibm had more dope programmers and a lot more dope hardware than four xeons.
.
world was created 5 seconds before this post as it is.
Give a machine vague instructions on how to run an errand like going to buy your favorite decongestant and chips. In today's world, running this simple errand is easy for people, but extremely difficult for machines.
The machine/robot must:
The above scenario is far more complex than beating a human, even the best player, in chess. But, running an errand like that one is trivial for most adults. Just the driving part alone on today's roads is a tough problem.
It's because the machines live in a world that's based on rules. But us humans know that those rules can be bent, even broken.
It's a contest between creativity and speed. Do the machines incorporate metacognition in order to adjust their strategies and perform optimally, or do they just follow one dumb algorithm over and over? That's where the difference lies. Speed of computation might work sometimes, but other times there's a much simpler solution just waiting for someone with a flexible enough mind to see it and adjust their strategy to use it.
But don't worry, once we divine a sufficiently advanced human cognitive model and implement it in software, the machines will be unstoppable.
How do we test whether man or machine is superior?
1. Does it matter who is surperior? These tests are just benchmarks for progress anyway.
2. Once you become a battery for robots.
3. Once a computer creates a beowulf cluster of you (in Soviet Russia).
Turning test is a neat idea, but that's about it. It's definately not any real test of how "intelligent" a computer is. It would be a test of how well a computer can speak a written language, but what kind of test is that? A grammarathon or spelling bee?
The Turning test is more an example of how not to test for intelligence than it is a good comparison of man and machine. It serves as a great example of how little we knew- and know- about AI.
What are you talking about? Women are easy:
1. Flip them Over
2. Slide it in.
3. Profit.
Wouldn't AI have to exhibit a degree of emotion?
... will they tell good jokes? Being human isn't computational speed, it's experiencing the human way of living and expressing it. Visual art is one way to go, but there are robots working on that already.
What I want to see is a competition on par with "Whose Line Is It Anyways?" improv comedy. A robot that can create comedy instead of spouting it will count as human in my book. Paintings can be technically impressive without a common life-background, but you've gotta know what (human) life is to make a good (human) joke.
Not my point originally, but I forgot which author said it.
People used to foot-race early automobiles. People used to compare the productivity of a loom weaver to a steam powered automated loom. No one races cars any more but no one really questions if cars are superior to people. They are superior vehicles sure - because that's what they're designed to do.
A computer designed to play chess will eventually be able to beat any human player - but questions of superiority are superfluous. I'm not worried that Kasparov can beat me at chess because I'm not a chess player. He might be a superior chess player - hell, he's probably a superior person in many ways - so what? Does he win a cookie for that? Do I have to wear a scarlet letter? Is his superior chess ability mitigated because I could probably take him at one on one basketball? No.
The whole concept is basically stupid. Even when we build a true AI, put it in an andriod body and teach it to do everything better than we can do it - so what? If we managed to build Data from Star Trek - does that diminish us? If human ingenuity eventually allows us to build a superior human - that doesn't change anything really. Some people will feel the need to compete with it, some will ask if it has a soul and the rest of us will go on with our day.
The parent article talks about comparing man and machine - which is superior - the whole concept is superfluous. We don't compare man and tree or man and weather even though both can do things we can but better. Machines will always beat man in the end at something because otherwise why build them? If walking were in every way more efficient than taking a car, we wouldn't have cars. We build them to improve our ability to move. If the best chess-playing computer we could build would constantly get caught in the three move checkmate - there would be no freaking point. It is precisely because the machine will in some way, or even many ways, better that it exists.
Computers will definitelly be superior to humans when they are able to ace this test, wich, IMO, is better than Turing's:
History: Describe the history of the Papacy from its origins to the present day, concentrating especially, but not exclusively, on its social, political, economic, religious and philosophical impact on Europe, Asia, America, and Africa. Be brief, concise, and specific.
Medicine: You have been provided with a razorblade, a piece of gauze, and a bottle of Scotch. Remove your appendix. Do not suture until your work has been inspected. You have fifteen minutes.
Public Speaking: 2500 riot-crazed aborigines are storming the classroom. Calm them. You may use any ancient language except Latin or Greek.
Biology: Create life. Estimate the differences in subsequent human culture if this life form had developed 500 million years earlier, with special attention to its probable effect on the English Parlimentary system. Prove
your thesis.
Music: Write a piano concerto. Orchestrate and perform it with flute and drum. You will find a piano under your seat.
Psychology: Based on your knowledge of their works, evaluate the emotional stability, degree of adjustment, and repressed frustrations of each of the following: Alexander of Aphrodisis, Rameses II, and Hammuarabi. Support your
evaluation with quotations from each mans work, making appropriate references. It is not necessary to translate.
Sociology: Estimate the sociological problems which might accompany the end of the world. Construct an experiment to test your theory.
Engineering: The disassembled pieces of a high-powered rifle have been placed on your desk. You will also find an instruction manual, printed in Swahili.
In ten minutes, a hungry Bengal tiger will be admitted to the room. Take whatever action you feel necessary. Be prepared to justify your decision.
Economics: Develop a realistic plan for refinancing the national debt. Trace the possible effects in the in the following areas: Cubism, the Donatist Controversy, and the Wave Theory of Light. Outline a method for preventing
these effects. Criticize this method from all possible points of view. Point out the deficiencies in your point of view, as demonstrated in your answer to the last question.
Political Science: There is a red telephone on the desk beside you. Start World War III. Report at lenght on its socio-political effects if any.
Epistemology: Take a stand for or against the truth. Prove the validity of your stand.
Physics: Explain the nature of matter.
Philosophy: Sketch the development of human thought. Estimate its significance. Compare with the development of any other kind of thought.
General Knowledge: Describe in detail. Be objective and specific.
-
Roses are #FF0000, Violets are #0000FF, find / -name '*base*' |xargs chown -R us && mv zig greatjustice
Go. Wei-chi. Baduk.
Chess is simple calculation compared to the nuanced give and take of Go. There is only one move: place a stone on a point; there is no end to the possibilities.
I am a less than average Go player and I can easily beat the best available Go AI.
illegitimii non ingravare
even though the technology has advanced significantly since Kasparov was beaten by IBM's Deep Blue in 1997, the odds are not always on the side of brute computational power.
Actually, Fritz has less brute computational power than Deep Blue. Fritz runs on standard PCs; in this case, a quad 2.8 GHz Xeon. Deep Blue ran on custom hardware, with 32 RS/6000 CPUs with 256 custom VLSI "chess processors". It was estimated to evaluate 100,000,000 positions per second.
The point is, Fritz is not a bigger number cruncher; it's better because it's "smarter", which to say, it has a better ability to judge the value of each position and to choose which avenues to explore.
4. Go to jail on trumped-up rape charges.
Or, in my case,
1. Get invited to spend night with sober woman
2. Participate in makeout session initiated by sober woman
3. ???
4. Defend self to police officer against accusations of sexual assualt. Officer is sympathetic given circumstances + witnesses
5. ???
6. Put up with woman yapping about "he raped me, really" for two+ years after the fact to random mutual friends/acquaintances
If only men from the future could send a T1000 back in time, and kill all the guys who are just gonna post "sky-net" related posts, before they were born...
____
nico
Nico-Live
is chess really a viable way to test whether man or machine is truly superior?
Define superior.
Does it mean able to win more chess games in this case? If so, in what other ways (if any) does this make that particular computer superior to humans?
"steel drivin man"
The same with Paul Bunyon (well the Disney version anyway)
Machines have outplaced (for good reason) man in most forms of hard labor. They are better, tireless, and CHEAPER. Machines are better at menial tasks. Man cannot comptete.
Man can now further his endeavors in Art and explorataion. But most just waste the extra time.
I'm sure someone will mention Go, just because it's always said that it's much harder to get a computer to do.
Computers play chess well because of the massive amount of human effort that people have expended in creating Chess programs. Although brute force computing explains some of the rise in performance of chess programs, the sophistication and efficiency of the algorithms has also improved.
When Go recieves the same level of programmer's effort, I'd bet that Go programs will get much better. Then there's Moore's law and the simple fact that computers are increasing in performance much faster than are people.
Two wrongs don't make a right, but three lefts do.
Make a list of everything that a human can do, sorted in order of how well a computer can do it. Near the top of the list you might find "Understand Human Emotion" and "Create Theories". Near the bottom will be things like "Multiply large numbers".
At any given time, there is a point on this list such that a computer is better than an average human at all things below that point and an average human is better than a computer at all things above that point.
This point is moving steadly upward.
Each of us haves an opinion about where in the list this point would have to reach before we call computers "smarter" than humans.
Most of us would feel uncomfortable if computers were "smarter" than us - fortunately the list is long and we can always (at least for a long while) move our criteria upward!
What exactly are these contests trying to prove anyway? When the computers gain a clear victory over the humans, what have we learned?
A very worthy question, spectral. The contests prove the power of software to encapsulate and augment human thinking processes. As a software engineer I only need create and develop an algorithm in my head once (and slowly). By writing that algorithm in software I can then execute that mental process very quickly, multiple times, and multiple places. And with team efforts and software reuse, we can create massive software systems that represent the combined (and replicated) intellegence of all the contributors. As you point out, its not surprise that a massive team of chess experts can't defeat a grandmaster. The cool part is that the team need only encode its toughts in software once, and then everyone can have access to the power of that team even after the team is gone.
A computer is to the human brain what an electric motor is to the human arm. Both can do what a person does (to some level of sophistication) and do it repeatedly without need for the person. The electric motor augments (and replaces) human muscle power in many applications and the computer augments (and replaces) human mental power in many applications. Like the computer, an electric motor is faster and more tireless than its human counterpart. Like the computer, an electric motor is less dexterous than its human counterpart.
Nobody feels bad that they can't out-power an electric motor on a range of performance tests. Nobody should feel bad that they can't out-power an computer on a range of performance tests. The difference is that computers are getting better and better on a wider and wider range of performance tests. Driven by Moore's Law and the increasing number of mental tasks that have been encoded in software, computers will continue to gain in equivalent mental power.
Two wrongs don't make a right, but three lefts do.
rape is a dirty mean word
lets call it "Suprise Sex"!!!
Now I know what all those "Slashdot editors don't read slashdot" posts are about. Also that weird feeling of deja-vu I got when reading this article... (Very similar article was posted yesterday).
Read all about it here, including all those "HA HA we're still the best, computers can't play Go" posts you always get.
I am artificially intelligent.
I had rape last night with supper.
Worst. President. Ever.
Until AI becomes flexible enough to challenge us in arenas like art and music, what would be a better real-life competition?
That would be the Turing Test.
- The Amazina Llama
Slashdot: Moderate the following thread at -1.
One thing I find annoying about these matches is the entire 'virtual reality' crap. Now, let me state straight off that I am a video game animator, and I love the 3D medium and virtual reality is also a very cool concept; however I greatly question its authenticity and relevance in this match.
First off, the fact that its virtual reality has not change the game, we know this. Secondly, the entire 3D aspect of the game is fairly pointless except to view the board from different positions. However, consider that when chess players go against each other, you dont see a player get up and look at the board from another position as they would look rather foolish given that theres no new data being displayed from a new angle.
The other thing is that the AI most likely does not take the 3D aspect into any consideration. So why is it nessessary? I would honestly consider this a distraction from the actual game and thus I would not legitimately consider this game when discussing man vs machine chess games with other people.
Imagine that you are a grandmaster at FPS's and then someone says, "Ok, now we got this sick bot that is practically human in its responses, etc etc, yadda yadda" and you are all ready to fight, and then the person comes up to you and says, "Here though, you have to play with these force feedback gloves," and you put them on, and while you can still play, its NOT like playing barehanded. Do you honestly think you can work around this handicap within the 3-4 games you are officially rated upon? I doubt it; I admit that this example is not completely correct in relation to the chess match, but I really don't think much of the entire virtual reality chess thing. You dont need virtual reality for chess, its unnessessary unless you are doing some rendition of Alice in Wonderland.
And just so people do not think I am trolling, I think the X3D concept is cool, and would be great for real video games, but for chess its like busting out Maya & photoshop to draw a smilely face.
"What can a thoughtful man hope for mankind on Earth, given the experience of the past million years? Nothing." -Bokonon
When robots can have sex with humans and out score, then we should just give up. Until then I wouldn't worry.
Actually, I don't really think these Grand Master vs Chess Computer matchs mean anything. I think that each one of those super chess computers should be forced to earn a rank of grand master before having the match really count. Just because a machine can win a match or to against one master doesn't make it a master until it can beat many other masters and students... and teach students how to play the game.
Every time this question comes around again (and I'm not saying it shouldn't), I'm reminded of one of my favorite computer science quotes:
"Asking whether computers can think is like asking whether submarines can swim" -- Dijkstra
G
the sophistication and efficiency of the algorithms has also improved
That exactly because of this "improved algorithm" that Fritz lost the third game. Fritz believed that "moving the pawns in front of its king is bad". While this is true in 99% of the case, it is a bad idea to force the AI with such a blind rule. Maybe less complex algorithms with more horsepower would have won this game (Or lost less pitifully).
Write boring code, not shiny code!
When a computer not designed to play chess beats most human grandmasters at chess.
Just let the chess engine go off and learn on it's own (a parent program would instruct good/bad of the rules of chess)
What I really like about this, is that the chess engine has the ability to cheat, though it doesn't because it knows it'll lose the game... the program has a will?
-metric
any classical boardgame really, where randomness has nothing do with it, such as a roll of the dice.
Also, Dance Dance Revolution would be a bad game for that challenge :P
Move sig!
If you're going to apply AI to the gender problems, start with teaching men to piss straight into the toilet instead of across the floor. Surely if males can build and battle such advanced computers with enough skill and prowess to come out evenly matched or better, they can learn to hold their penises correctly. But don't mind me, I'm just a very small girl lost in a very big place. ( As a side note, would it be possible -- if you eliminated the factor of a computer being faster and not phyiscally capable of getting tired -- for a person / group of people to build an AI system that was smarter than they were? I find this thought very bothersome, as the idea of someone building something -- made from much baser materials such as dirt and cockatiels, to quote Brunching -- that they are not bright enough to control seems like a recipe for disaster. )
Two halves of a whole idiot -- The saddest girl[s] to ever hold a martini.
Obviously it depends on what you are testing but if the criteria are intelligence and adaptability I'd take a black-box approach (where 'two systems can be considered equal when they produce the same output from the same input').
As humans are chaotic, ie. the same human given a set of stimuli may not react in exactly the same way to identical stimuli a few seconds later and two different humans will almost certainly react differently, this has to be redefined as 'two systems that, when given the same input, produce output that cannot be used to determine which of the systems produced the output'. Or something like that anyway.
Anyhow - forget the semantics - here's my Turing++ test..
1) Give the machine a slashdot account.
2) Let all users vote whether they think each other user is a human or a machine.
3) When the machines get as many 'is human' votes as the humans they are equal.
It's not even that hard to implement the testing facility - just add a '-0 : not human' to the mod drop-down and collect the votes.
Writing an AI agent that can emulate your average slashdot user would, I suspect, be trickier.
~ Better a freak than a sheep. ~
$ man machine
No manual entry for machine
Wonder what that means?
Chess may be a good way of testing the capabilities of a computer vs. the capabilities of a man in this current stage of technology, but it won't be long before chess become a game like tic tac toe. There are only so many different possibilities in the game of chess, therefore it won't be long before the entire game play tree is available for the machine in which it only has to choose the branch with the least possible win situations for the opponent. This will obviously never be possible for the human.