How intelligent is a computer - and just what does it mean to be intelligent, anyway? The Association for the Advancement of Artificial Intelligence (AAAI) aims to clarify that question with its Man vs. Machine Poker Challenge, taking place today and tomorrow in Vancouver, British Columbia. Polaris, a poker program developed …
"My money’s on the Unabomber."
Of course. But the cool thing about this kind of challenge is that, even if the computer loses, the scientists can still learn something, and design something that will come back stronger next time.
I'm sure the author knows this, but I thought I'd just highlight this point for any reader who might think that the whole experiment is a failure if the computer loses its virtual shirt.
Scientists *often* learn at least as much from "failures" as successes.
To Boldly Go ....... a Colossal Turing Project ?
Come on on in, the water is fine. Wish you were here, PsID:-)
"Part of the challenge for proponents of artificial intelligence lies in the difficulty of defining exactly what intelligence is: “intelligence” as we typically understand it is a rather haphazard measurement of our abilities, some of which lend themselves to duplication via computer software, and some of which do not."
If Viable Virgin Imagination is Intelligence, a computer is always just going to be a Box of [Magic Pandorian] Tricks which can be Programmed with IT Imaginatively, which renders Memory not so much a redundant component of Computing, for it is valuable for repetitive/copy function, but as an aid to Progress, it has zero input and may actually have a negative effect upon natural progression/evolution should we fail to consider IT so.
Do you think Advanced Artificial Intelligence is a QuITe Alien Intelligent Design for A.N.Other Generation of Human XXXXistence with a Simply Complex ReProcessing of Information/MetaData 42 Show another side to Reality which can be built with Media and Comunications leading the Way?
Some may have even concluded that that is the only logical way Forward out of Chaos and have already launched into ITs campaigning .... http://jamesstgeorge.proboards32.com/index.cgi?board=Worldissues&action=display&thread=1184777300&page=2#1185259404
Skill is not intelligence
Deeper Blue was highly skilled at chess but no more intelligent than a block of wood; moles are skilled at digging tunnels and wasps at turning wood into paper buildings.
After all these years there's still no sign of a better test than the Turing Test and nothing has come close to passing even that rather loose definition of intelligence. AI research is really all about developing a more sophisticated system for extracting grant money from various patrons.
Any ideas how polaris is supposed to work?
So can anyone give an overview of how this polaris program works? I assume it number crunches the card probabilities, and I guess it will gauge its own bet levels based on its own perceived chances of success, but what else does it do, if anything?
What other inputs does it consider? Level of opponents bets? Time taken by opponent to lay a bet? If so what if their opponent is delayed for reasons outside consideration of their bet? What if they went to the bog or had to blow their nose?
Is it equiped with video input and attempts to read the humans body language? I strongly suspect not. Will this body language blindness prove an advantage or disadvantage. And of course the machine will give up no body language to the human player - the ultimate poker face - no tells, unless the humans notice the way its diodes always quiver when it's bluffing :)
What is intelligence?
My humble opinion:
1. Being aware of yourself as a person / entity
2. Having memory
3. Being able to process sensory input, recognize patterns previously learned, and fit it into memory as part of world view / understanding
4. Ability to weigh ideas / objects / images / actions etc and give value to them
5. Ability to learn and use languages, modes of communication.
6. If the "intelligence" is embodied in a machine or vehicle of some kind, ability to learn optimal movement and balance.
Whose a geek?
After decades of persecution and excessively prolonged virginity, it seems as if the IT crowd is no longer at the top of the geek list. Look carefully at the two rubes in the article and consider:
poor fashion sense
excessive amount of pride and posturing
Seems to me that IT geeks could hang out at a poker convention and do fairly well with the opposite sex. The poker guys are going to make the IT folks look like gods.
Well if their "Poker Face" is going to be anything to go by Polaris has it nailed!
An unfortunate coincidence
David Gelernter, the polymath and computer scientist whose book "The Muse in the Machine" dealt with the practicalities of machine intelligence in compelling detail, was a victim of the bombing campaign mounted by Theodore Kaczynski, later known as the Unabomber.
I'm not sure where this comment can go from here, except to say that Phil Laak could conceivably have picked a better nickname. I felt the coincidence should not go undocumented.
Not a test of emotional intelligence
A big part of poker is that it is a game of deception and of inference. Many good poker players win not because they have a perfect knowledge of the odds, but because they can read the other players. Conversely, they can also deceive the other players playing. “Trash talking” and taunting the other players (to put them off balance) can also be a big part of the game. Since the article states that players will be unable to communicate with each other, the only “information” that will be available will be the cards. With that being the only information available, the edge is to the computer as it can impute probabilities vs. expected payouts with perfection. What this experiment ends up measuring will not resemble a real game of poker.
AI and poker
I worked on a team that developed a program to compete in the U of Alberta's Poker competition, and the AI involved is significant. Despite the lack of visual contact (something computers don't do so well anyway), there is a lot of learning you can do based on your opponent's actions. As a simple example, anyone who plays poker knows that you can't treat tight and loose players the same if you want to maximize your winnings.
As for the value of AI in general...
For decades, the state of artificial intelligence has been disappointing everyone who has read Isaac Asimov's robot stories or similar sci-fi. However, the real goal for most researchers (other than Ray Kurzweil) is not to mimic humans, but to enable computers to solve problems that the programmers don't know how to solve themselves. Some of the basic techniques of modern algorithms, such as many search techniques, grew out of AI research into solving games using limited resources. It doesn't look much like intelligence, but the results are better than anything programmers are likely to accomplish using their own understanding of the domain.
This is a MASSIVE step towards true AI.
All they need to do is to implement facial recognition software for the whole picture of poker and the machine will be hard to beat.
This is step 1. If they can get close or beat the pros, then half the battle has been won. Imagine going into a tournament and going head-to-head with a robot - it can see you, you can see the robot. It'd be hard trying to defeat a robot that has no tells.
Intelligence (IMHO) is a multitude of factors - it is NOT JUST how much knowledge you have retained/gained. Your senses also play a part in gathering "intelligence".
You may enter a room and have nothing in it - you know that by looking around. But there may be a certain smell/odour to the room. You have just acquired intelligence from your senses about the room. Or a certain surface may have a distinguishable "feel" to it. More intelligence.
Like anything, over time, the Polaris team will gather more "intelligence" and become better at it.
The reality of playing a robot on the poker table is closer than you think. Bring it! :)
The Turing Test is Rubbish (and Ray Kurzweil is a c*nt)
Much as I worship the late lamented Alan Turing, his test for AI is possibly one of the stupidest things he ever came up with. It measures simply the ability of a machine to convincingly imitate a human being.
While a team of boffins and a gert big supercomputer may fail to convincingly imitate Alan Turing, I can probably program my 90s era palm pilot to convincingly imitate Lisa from down the chippy.
Not that I've got anything against Lisa from down the chippy mind you, she always gives me extra chips.
Funny that Ray Kurzweil's name should come up, I recently found a copy of his "Age Of Spiritual Machines" gathering dust on my bookshelves. Typical AI cheerleader and a complete space cadet. Meh.