AI Taught How To Play Ms. Pac-Man

trogador writes with the news that researchers are working to teach AIs how to play games as an exercise in reinforced learning. Software constructs have been taught to play games like chess and checkers since the 50s, but the Department of Information Systems at Eotvos University in Hungary is working to adapt that thinking to more modern titles. Besides Ms. Pac-Man, game like Tetris and Baldur's Gate assist these programs in mapping different behaviors onto their artificial test subjects. "Szita and Lorincz chose Ms. Pac-Man for their study because the game enabled them to test a variety of teaching methods. In the original Pac-Man, released in 1979, players must eat dots, avoid being eaten by four ghosts, and score big points by eating flashing ghosts. Therefore, a player's movements depend heavily on the movements of ghosts. However, the ghosts' routes are deterministic, enabling players to find patterns and predict future movements. In Ms. Pac-Man, on the other hand, the ghosts' routes are randomized, so that players can't figure out an optimal action sequence in advance."

Comments

Not Really

[ilikepi314]

It just lied that it could play Ms. Pac-Man so it could get more reward food.

Re:Not Really

[calebt3]

Too easy to check up on that. It lied about it's top score.

Re:Not Really

[BinarySkies]

The joke is on it, the cake is a lie.

Re:Not Really

[Anonymous Coward]

Man, that must be the most self-referent joke of all /.

Re:Not Really

[Anonymous Coward]

How is it self-referent? Did GP fake to have posted it just to have more reward food?

What I have learn from PacMan is

[mastermemorex]

live fast, eat chips, big ones are the best and avoid the gosh with ugly faces

re Now we KNOW!

[jelizondo]

In Ms. Pac-Man, on the other hand, the ghosts' routes are randomized, so that players can't figure out an optimal action sequence in advance.

I feel I'm beginning to understand ...

Perhaps the greatest achievement of AI would be to understand female behavior

Re:re Now we KNOW!

[AndGodSed]

That my friend, is a statistical; scientific and verrily a religious impossibility.

Re:re Now we KNOW!

[the_banjomatic]

Perhaps the greatest achievement of AI would be to understand female behavior Obligatory, but true none the less: "The only way to win is not to play"

Re:re Now we KNOW!

[hairyfeet]

Not to mention very cruel. You'd have the poor little A.I.'s brain snapping like a twig trying to figure out things like "If you don't know why I'm mad then I'm not going to tell you". It would be the A.I. equivalent to torture!

Re:re Now we KNOW!

[Anonymous Coward]

That my friend, is [...] a religious impossibility.

Absolutely. What would God do if we figured out why she created the universe the way she did?

Re:re Now we KNOW!

[stonedcat]

This should not be difficult for most slashdot readers.

Re:re Now we KNOW!

[Whiteox]

Ahhh: In Basic

Loop

IF N N THEN N = N

NEXT N

That's got to be close :)

Perhaps Slashdot's lameness filter should be able to recognize code?

Re:re Now we KNOW!

[Whiteox]

IF N<>N THEN N = N

NEXT N

OK I worked out how to do it.

Re:re Now we KNOW!

[zippthorne]

Nope, You still lose. The only way to win is to be attractive. Then it doesn't matter what you do.

Re:re Now we KNOW!

[ralphdaugherty]

Perhaps Slashdot's lameness filter should be able to recognize code?

      or even better, recognize lame code.

Re:re Now we KNOW!

[Whiteox]

What? You can do better?

Re:re Now we KNOW!

[mestar]

"to understand female behavior"

I'm guessing you're having problems with shit tests. http://www.fastseduction.com/guide/

Re:re Now we KNOW!

[ralphdaugherty]

What? You can do better?

      yes, but I'm keeping a low profile because I don't want the Terminator to find me.

  rd

And here I thought

[AndGodSed]

I was playing quakewars against the computer AI, you know... the bots?

Re:And here I thought

[Anonymous Coward]

did you read the article summary or did you just glance at it while freebasing

Bad idea

[QuickFox]

As if everybody didn't already waste too much time on games, do we have to teach programs to waste time too?

Re:Bad idea

[GroeFaZ]

The good part is, computers are several orders of magnitude more efficient at wasting time than humans.

Re:Bad idea

[exp(pi*sqrt(163))]

What are you talking about? By having machines play our games for us, humans can finally move on and become truly free.

Re:Bad idea

[Tribbin]

Your post is rated Insighful.

*Some* need to re-tune their dotbots.

--

This post is generated

Re:Bad idea

[Anonymous Coward]

The good part is, computers are several orders of magnitude more efficient at wasting time than humans.

Let me guess, you use Vista.

Re:Bad idea

[Kjella]

Yes, I know some mod insightful instead of funny to give karma, but this is just one of those times where it gets stupid.... Wasting time is a state, not a task and can't be finished faster. Capacity * 0% utilization = 0 no matter the capacity. At any rate I only care when they're wasting my time, which in 99% has nothing to do with CPU load...

so...

[dashslotter]

Who's Al?

Re:so...

[hotwatermusic]

Allen Iverson..duh.

Re:so...

[Adambomb]

Well, he could be Betty's long lost pal.

Re:so...

[ChefInnocent]

Well, at first, I read it as "Al Taught How To Play Ms. Pac-Man", so my first thought was Al Gore. But a second parsing lead me to AI or Artificial Intelligence (though some my acquantences think American Idol). The font can really lead to an ambiguous reading.

Re:so...

[Ajehals]

Nah, Betty's mate just used Al as a pseudonym.

Re:so...

[Anonymous Coward]

What?

Re:so...

[webmaster404]

Al Gore of course. He also told the ghosts how to be ecologically responsible and not to contribute to global warming.

Re:so...

[Chris Mattern]

I can call you Pac-Man/And Pac-Man when you call me/You can call me Al...

Re:so...

[Anonymous Coward]

Does it matter? Al Gore or Weird Al. Either way, playing Ms. Pac Man isn't that big of an accomplishment.

Re:so...

[Conanymous Award]

That's what I thought, too. Gore? Pacino? Bundy?

"AI"s tend to be overhyped

[eyenot]

I think most press releases re: AI are misleading. I highly doubt there is anything like "AI" behind the program they have that attempts to solve Ms.Pacman. Consider if you wrote an "AI" that started off with what you as a human starts off with: the ability to see the screen and understand what the various graphics depict or mean; how to control the pac character; what the basic goals and obstacles are; and a desire to rack up points. An "Artificial Intelligence" (AI) would be able to start with that much and build its skill level as it plays. Presumably it would quickly build a talent that can beat average humans, then most humans, then eventually all humans since it has faster reflexes and doesn't get tired (or make errors once it's learned). That, I think, would justify a press release "AI learns to play Ms.Pacman". However, scripting something that plays the game as well as you can imagine it should be played doesn't seem to be news any more than "scripters automate online game play". I only note this because the article mentioned "teaching" the "AI"; that's not very scientific, considering you're trying to see something learn, and should be maintaining scientific control over the learning process.

Re:"AI"s tend to be overhyped

[bunratty]

AI can be as simple as basic search algorithms such as breadth-first, A*, and minimax. When you play any board game against a machine, that's AI. When you get driving directions from a computer, that's AI. It seems to reason that AI is behind a computer playing Ms. Pacman. And in this case, the computer generate playing policies on its own, so it really is learning, improving its performance based on previous experience.

Re:"AI"s tend to be overhyped

[ralphdaugherty]

      AI seems to be nothing more than try random outputs and use feedback to reinforce outputs that resulted in success. It's sort of funny, my first recollection of this was in 1962 when a student in my grade school class performed this exercise for a project. A game was played repeatedly with losing moves recorded, developing a chart. Playing from the chart the game was eventually unbeatable by fellow students. The more things change the more they stay the same.

      I wrote a Double Deck Pinochle game (first for TRS-80, then DOS, then Java, and soon PHP) whose DOS version I released as freeware many years ago. As it plays as well or better than most humans, it might mistakenly be called AI as chess games were earlier, but it is just a program of logic as in any other computer program. Anything complex as that doesn't lend itself to the trivial output - feedback loops that modern day AI efforts involve. Of course the more ambitious efforts of the '80s are infamous in their failures.

      If we ever achieve AI it will be with a core of code that can generate modules of code that attempt different strategies, in other words grows a brain as program code and database, not just a matrix recording true - false results from random permutation outputs.

  rd

Re:"AI"s tend to be overhyped

[Anonymous Coward]

I think most press releases re: AI are misleading. I think you don't understand the difference between AI and sentience.

Re:"AI"s tend to be overhyped

[GroeFaZ]

I only note this because the article mentioned "teaching" the "AI"; that's not very scientific, considering you're trying to see something learn, and should be maintaining scientific control over the learning process.

All machine learning methods can be controlled, that's not the problem. The learning models either have parameters that can be retained or changed at will between runs, or they don't have parameters, which means the conditions are always the same, which saves the same purpose. The outcome can be defined, measured, and compared to other runs. Somewhere in between, random stuff from the outside happens, and from the outcome over many runs you can draw conclusions, for example which parameter had the most influence on the final result. You need the (pseudo)random input somewhere, otherwise you would always get the exact same result in every run. So from a scientific method view, there seems to be problem with machine learning experiments.

The problem I do see with machine learning as described in TFA is that it operated on a defined set of rules only. It won out over humans by a small margin, which even left the possibility that 4 of the 5 humans were still better than the AI, and that small possible victory happened despite having fewer strategic rules than humans, implying it won by brute force. So whatI would like to see is an AI developing novel behaviour, developing new sets of rules that really show off the difference in computing speed for such problems between a computing machine and an organic brain.

Re:"AI"s tend to be overhyped

[Hado]

I feel I must comment since I am familiar with the AI used in this case: Reinforcement Learning. RL is a method of finding a mapping of states to actions in a setting where rewards can be obtained. The interesting part is that RL algorithms can learn to behave optimally when only very basic information is given. For instance, it should be enough to simply give small rewards for eating the dots and large punishments for being caught by a ghost. There are many theoretical results in the field that also hold in the case of stochastic environments (such as when the ghosts move randomly). In a sense you don't have control over the learning process, at least not in the sense that you control what exactly happens and which actions get tried. However, in the end theoretically still perfect behavior can be learned. This may take quite some time though, but fortunately good behaviors usually emerge much sooner.

That being said, it is relatively easy to apply these techniques to games such as Ms. Pacman. Much harder problems have already been solved using RL algorithms. What seems missing in the article (though I don't know if this is also the case in the actual research) is comparisons with other RL methods than their own. Though their approach sounds promising and it's nice that they beat some human players, this is not uncommon in games for RL.

Re:"AI"s tend to be overhyped

[Tyir]

Actually, what you describe is exactly what Reinforcement Learning (RL) is. RL can be considered a subbranch of AI. In RL, an agent starts by knowing nothing about the environment. It explores the environment by taking available actions, in this domain, the actions would be exactly the actions available to the human players. It also has a reward signal R, which is used to train the agent to do the correct thing. Completing the level will probably give a high reward, encountering a ghost will give a negative reward. What a RL algorithm will do is give approximations for the future value of being in any state in the environment. What the researches will do is train the agent on the domain for a large number of steps (perhaps millions of games played in simulation) and the agent will learn to play the game well. Note this require *no* domain knowledge, i.e. the programmer doesn't but any heuristics, strategies, or high level tricks to have the agent complete the game, which I believe is what you think is being done here.

So what the technique used in this paper is doing exactly what you would consider "real" AI. Full disclosure, I am a Master's student who has done a good amount of work in RL, and I have not read the paper, so what I describe above is not going to be exactly right, but is probably the general idea.

Re:"AI"s tend to be overhyped

[eyenot]

Well, suffice it to say, I am simply not of the "camp" that believes "artificial intelligence" should be applied so cleanly and popularly to something that obviously does not do much "learning" on its own, at all.

Re:"AI"s tend to be overhyped

[bunratty]

I think you're referring to general or strong AI, which hasn't been developed yet. All we have now is weak AI, which even when it seems to demonstrate "learning", all it's really doing is running mechanical search algorithms and heuristics really, really fast.

Re:"AI"s tend to be overhyped

[eyenot]

Understood, and I have no clinical knowledge of AI at all, and can only make general assumptions. However, I see what has been made here for Mc.Pacman (and for other specific environments) more as a combination between strategics and data-mining. The data to be mined in Ms.Pacman would be the predictability of the randomized ghost paths. I haven't looked at the arcade game code, but there is probably some pseudo-RNG involved that is seeded by a timer (if this is correct, in fact the ghosts would behave the same way if you could always start the game at the same exact time). Basically the only thing missing from this "AI" is the pseudo-RNG itself, which the designers could have easily thrown in along with knowledge of whatever the RNG is being seeded with. The outcome, if this is valuable at all, would perhaps be some formula for behaviour that can be reduced and tokenized and perhaps applied to resource-sharing or risk-taking; but this outcome will take considerably longer to arrive at considering that first the "AI" will be sorting through the pseudo-randomness. But I digress -- this "AI" isn't really learning anything, it's just dealing with missing variables. It can't make any cognitive leaps from the human equivalent of "intuition", it can't re-apply what it's learned (though in this specific case that's probably more due to the restraints of the tiny and simplistic environment), and if I read the article correctly (nor did I read the research paper) it doesn't properly make informed decisions, and all of its actions are entirely predetermined, albeit by a pseudorandom element. Considering that innovation is considered either a hallmark of intelligence or its entire purpose, it would seem that by that standard alone the game itself will always be more intelligent than this "AI".

Re:"AI"s tend to be overhyped

[eyenot]

Conversely, punishments could be foregone since it's only a game, and the script could be left with the in-game punishment provided, which is failing to make it as far and push the limits of the process, and rewards are already provided in-game in the form of the score. I can see how application of "RL" algorithms to the script itself might reveal some things about the application of scripted behaviour to the learning process, but I personally don't feel that this script constitutes AI (see other responses?)

Re:"AI"s tend to be overhyped

[eyenot]

I made a response to a later comment about brute forcing the pseudo-random element, and how if the designers had thought that through and simply included the obvious RNG and seeding subroutine, the script could have jumped that hurdle ahead of time and might already have shown the quantification of risk-taking in luring ghosts and timing pills (though I predict that those behaviours simply won't win out over avoiding risk and racking up points through stamina and perfection). I just don't think "AI" is a term that needs to be applied here (response I am going to make following this one).

Re:"AI"s tend to be overhyped

[bunratty]

this "AI" isn't really learning anything, it's just dealing with missing variables. It can't make any cognitive leaps from the human equivalent of "intuition", it can't re-apply what it's learned (though in this specific case that's probably more due to the restraints of the tiny and simplistic environment), and if I read the article correctly (nor did I read the research paper) it doesn't properly make informed decisions, and all of its actions are entirely predetermined

That's the only AI we've ever developed. As you point out, it's completely incapable of doing anything original. It's called weak AI, as opposed to strong AI, which exhibits general intelligence. Strong AI is strictly limited to science fiction at this point.

Re:"AI"s tend to be overhyped

[eyenot]

I'm just stating that there's no need for such a general application of the term "AI" so as to render it practically useless (dict. "hyper-realised"). It's akin to the term "nano-technology" being used now to refer to anything very small that's somehow industrially useful but not biological, such as buckytubes and other Fullerine structures. A carbon-lattice, curled-up tube -- seen by itself, out of context -- may have come just as easily from natural processes as "technological" (which is how they were discovered in the first place). "AI", now, is being applied to any programming which bears even the most unremarkable and shoddy human resemblance, or is being used to automate previously human-controlled tasks, even without any "learning" involved.

Re:"AI"s tend to be overhyped

[eyenot]

Earlier, more abitious ideals of AI are what I based my criticism of this research on. Granted, there was a great deal of failure, but there was also some slight innovation (such as the tiny "e-life" routines) that went along with perhaps too much fear of returning to older methods or applying anything new back to them.

Re:"AI"s tend to be overhyped

[Unoti]

If we ever achieve AI it will be with a core of code that can generate modules of code that attempt different strategies, in other words grows a brain as program code and database, not just a matrix recording true - false results from random permutation outputs.

In that case, the future is already here. You should look into the work of John Koza and others. Their work involves generating code, real computer generated programs, not a matrix of lookup tables. I highly recommend his books, they are eye opening and amazing. (Warning, it's Lisp. But if you want to do this kind of work in any most other modern languages, you end up essentially building a Lisp-like interpreter anyway. Ruby would be one exception to this.)

Re:"AI"s tend to be overhyped

[Hado]

Actually, having reread your initial post, I largely agree with you. I think RL should be able to solve Ms. Pacman without the need of higher level strategies that are either hand crafted or learned with the cross-entropy method described in the article. However, the article gives little useful information about this method so I cannot tell whether this can be sufficiently be descripted by the term 'scripting'. I feel sufficiently complex scripts can in fact be called AI. However, this is always a matter of subjective interpretation and valuation of algorithms.

Re:"AI"s tend to be overhyped

[bunratty]

But that is indeed how the term has been used for decades. What you describe is taught in AI classes and is described in AI books. It's the only kind of AI we have. As such, the term isn't useless. If you want to refer to original thought by a computer, use the term "strong AI," which hasn't been invented yet.

Re:"AI"s tend to be overhyped

[bunratty]

If we ever achieve AI it will be with a core of code that can generate modules of code that attempt different strategies, in other words grows a brain as program code and database, not just a matrix recording true - false results from random permutation outputs.

I think you're referring to strong AI. Note that what you describe is not sufficient for strong AI. Doug Lenat used a technique like what you describe with Automated Mathematician in 1977, but didn't succeed at doing much even in the limited field of mathematics.

Re:"AI"s tend to be overhyped

[Chris Mattern]

AI can be as simple as basic search algorithms such as breadth-first, A*, and minimax. When you play any board game against a machine, that's AI.

I'd argue that. IMHO, it's the heuristics used to evaluate the positions discovered by the search that's the AI part of it.

Re:"AI"s tend to be overhyped

[subStance]

Agreed. The case study project used when I was studying neural nets as a noob was a tank game (teaching tanks how to collect pellets randomly distributed over a target area and not hit other tanks) had all the core elements described as being important in Ms Pac Man, and that was quite a while ago.

Methinks this is just a case of Ms Pac Man being used for recognizability to people who don't know the field. Certainly nothing groundbreaking.

Re:"AI"s tend to be overhyped

[BungaDunga]

I read about an "AI" that would appear to learn in a game against a human, gradually becoming harder and harder to beat. The trick was that the AI was just a bunch of matchboxes. Oh, this is it (Google-cached pdf):
http://209.85.173.104/search?q=cache:nNcKxyHirgUJ:www.teachers.ash.org.au/kenprice/documents/martgame.doc+matchbox+artificial+intelligence+martin+gardner&hl=en&ct=clnk&cd=1&gl=us&client=firefox-a

Re:"AI"s tend to be overhyped

[Citizen of Earth]

If you want to refer to original thought by a computer, use the term "strong AI," which hasn't been invented yet.

Invented, no — evolved, yes.

Re:"AI"s tend to be overhyped

[cnettel]

this "AI" isn't really learning anything, it's just dealing with missing variables. It can't make any cognitive leaps from the human equivalent of "intuition", it can't re-apply what it's learned (though in this specific case that's probably more due to the restraints of the tiny and simplistic environment), and if I read the article correctly (nor did I read the research paper) it doesn't properly make informed decisions, and all of its actions are entirely predetermined

That's the only AI we've ever developed. As you point out, it's completely incapable of doing anything original. It's called weak AI, as opposed to strong AI, which exhibits general intelligence. Strong AI is strictly limited to science fiction at this point. All along, we've also seen a shift in specific tasks, where we once thought that they would require strong AI. I would expect machine translation to be one area where larger data sets and only slightly more complex models (which are possible to train, thanks to the larger datasets), might result in the conclusion that good translation actually doesn't require understanding, or that this weak AI, at some level, shows equivalent understanding, even though it still wouldn't be able to practice it generally.

Re:"AI"s tend to be overhyped

[s388]

What you say is true but the parent is still right too, the way I see it. Articles like this routinely use misleading terminology ("Ms. Pac-man teaches herself") in reference to nothing more than a pre-programmed set of [prioritized] rules. The convention is in fact to make normal extant AI seem as strong-AI-like as possible, by way of blurry vague terminology and hype.

The robotworld (site) article for example is totally inconsistent or misleading in its use of the words "learn" and "teach". Gamers here know that when they play Half-life single-player they are fighting "the AI", but nobody can seriously say "the AI is TEACHING itself (and/or LEARNING) to attack me!".

Re:"AI"s tend to be overhyped

[Chess_the_cat]

Exactly. This is the sort of thing AI researchers tout in order to maintain funding. For 60 years AI scientists have failed at creating anything close to intelligence so they trot out the micro-world or the game solver and claim they're progressing. They're not. Read What Computers Still Can't Do by Dreyfus. Great stuff and pretty much blows the lid off claims like the one found in this submission's headline.

Re:"AI"s tend to be overhyped

[eyenot]

We could use a book like that for many, many more fields: astrophysics; "nanotechnology"; microbiology; neurology; psychology...

Especially psychiatry. Isn't it interesting that the president of the APA made a press release basically stating that American psychiatrists have absolutely no clue what the psyche drugs they are prescribing actually do? I couldn't tell if that was just a legal CMA or not.

Re:"AI"s tend to be overhyped

[eyenot]

Sorry for this late response, due to the holidays and I use public terminals. (Also, I like to carry on online discussion while there are still things to add).

The recent articles about two AI visionaries committing suicide was odd, but it did bring to light something I didn't know: that somebody was developing a fact database for AI to parse as a resource.

I had been thinking for some time about whether or not the ability to carry conversation with humans is really necessary for an AI, as a benchmark. The behaviourists would be really interested, I guess, (but for that case they could just specialise in ELIZA and leave AI in general alone, couldn't they).

I figure that it shouldn't be too hard, if it's completely necessary, to have a normal conversation with an "AI" if it's really just a dictionary/language parser, and you simply outfit it with grammatical structure and a very basic vocabulary of propositions and conjuctions and so on, basically the axioms of forming new data based on dictionary definition. Then whenever you hand it a statement, it could parse what you're saying in a branching structure through various dictionary terms. This would cover the input side: for output, I suppose any "level" of conversation is good enough, it's not as if the requirement is for "interesting" or "warm" conversation, so you could just apply the same process to an encyclopedia, just finding something at random or near-random to jot out at the user. There'd be no need to worry about this "learning" and so on, because apparently if all that's needed is some human-like conversation, that would take care of it, and the computer never needs to ask questions or record anything new.

However, the computer could keep a reserve, slightly-erroneous dictionary/encyclopedia hybrid (which we really have no way of predicting the appearance of except that it would be in syntax) which it would use to store any or all new statements (inputs) made by the human user. These could be referenced in off-time (when not referencing the dictionary or encyclopedia) for the purpose of, well who knows what. This would provide the computer with a database of human-behaviour-modified data that, ultimately, is similar to the dictionary and encyclopedia, because it's already formed into language (the basic idea being that the entire reason we use language is to store and transmit thoughts in the first place).

This still wouldn't, of course, necessitate anything near "intelligence", and even if it was successful, this addition of the slightly-erroneous data wouldn't actually be even one degree closer to "learning"; it would just be scribing and record.

But while thinking of this, I was inspired by Tilden's (the roboticist) leap with analogue brains. He makes these little insectoid robots that search for sunlight, but don't always just constantly take a step here or there to keep in the available light. They do various things, different things (for instance, trying to climb the fence of their confine instead of going straight for the light), and though it's really impossible to check all the states they're in to determine "why" they did them (because the "brains" are just variously-purpose, dicarded electronic boards that Tilden solders to contacts for the body controls and sensors), it's still interesting to see "behaviour". However, Tilden's leap wasn't in such brains to start with: the "behaviours" didn't appear until after he tacked on additional boards to the "brains". The original brains dealt with using data from sensors to make simple, analogue decisions about how to move the frame for the purpose of absorbing more light. For some reason Tilden decided to add one more board that took in signals from the "brains" and sent signals back out, processing the "brains" instead of the outside environment.

So it sort of clicked for me, that if you want your computer to learn, the best thing to do is make sure it can talk to itself and understand itself. So, once you have what I described above, the speech and language

Re:"AI"s tend to be overhyped

[s388]

Articles like this routinely use misleading terminology ("Ms. Pac-man teaches herself") in reference to nothing more than a pre-programmed set of [prioritized] rules. The convention is in fact to make normal extant AI seem as strong-AI-like as possible, by way of blurry vague terminology and hype.

The robotworld (site) article for example is totally inconsistent or misleading in its use of the words "learn" and "teach". Gamers here know that when they play Half-life single-player they are fighting "the AI", but nobody can seriously say "the AI is TEACHING itself (and/or LEARNING) to attack me!".

Humans in no danger yet

[mandelbr0t]

Average score of only 8186 (vs. 8064 by humans). Nothing really amazing here; if the AI could soundly trounce the best humans on a regular basis I might be impressed, but I can consistently score above 10000, and I'm not very good. TFA also notes that humans make better decisions on scoring points, while the AI shows some survival ability. Sounds like they need a better Ms. Pacman program.

Re:Humans in no danger yet

[eyenot]

Ah, but the scripts also managed to reach human-level average scores while discovering two things:

1. you don't necessarily gain anything luring ghosts...
2. or necessarily gain anything timing power pill consumption

Re:Humans in no danger yet

[nine-times]

Humans are not in danger until AI can make a site as awesome as the Pactionary.

Re:Humans in no danger yet

[master_p]

10000? I consistently score over 100000, and I have seen players reaching 700000 and more.

Also check out this video:

http://www.youtube.com/watch?v=cVH1mCc5EvU

Not gonna happen

[djupedal]

"I feel I'm beginning to understand ...
Perhaps the greatest achievement of AI would be to understand female behavior"

Understand that you can never 'understand' female behavior and be done with the entire exercise...

Re:Not gonna happen

[Conspiracy_Of_Doves]

The simple fact is that when the female suspects that the male is beginning to understand her behavior, she changes the rules.

Re:Not gonna happen

[Deliveranc3]

Of course then the trick is to pretned you know what she's doing until you do then pretend you don't.

Of course the range of possible female behaviours approximates infinity and thanks to feminism the range of acceptable female behaviours is perhaps a superset of the possible...

I'm just happy when they stay away from the subset that involves hitting you...

Re:Not gonna happen

[Whiteox]

Let's not forget our male roots!
Many years ago, when faced with that dilemma, a man would just find her, club her, grab her hair and drag her back into the cave.
Simple really.

Re:Not gonna happen

[Kjella]

I think it's one of the HHGTTG questions where you can't know the question and the answer at the same time, and in this case you know the question. Supposedly, if you find out "it will instantly disappear and be replaced by something even more bizarre and inexplicable." which isn't too far from the truth...

Re:Bad idea - NOT

[jelizondo]

Only by teaching them to waste time AI will be become truly human...

gender-neutral pac-person

[waveformwafflehouse]

So now we're teaching our AI that it's a round, dot hungry trans-gender Miss-Man being chased by ghosts?

Re:gender-neutral pac-person

[Hal_Porter]

So now we're teaching our AI that it's a round, dot hungry trans-gender Miss-Man being chased by ghosts? What are you some sort of Post-structural, postmodern, post Marxist pre capitalist Feminist academic with a background in Post Queer gender studies? Ms Pacman wears a bow and therfore SHE is a GIRL. It's very simple.

That French shit won't get you laid round here either, cut it out.

Baldur's Gate!?

[VPeric]

When the AI manages to play (and beat!) Baldur's Gate, I'll be seriously impressed. Pacman/Tetris simply aren't that exciting.

Re:Baldur's Gate!?

[someone1234]

I can imagine it will be able to kill kobolds, but i wonder how will it pick the right dialogue options :)

Oblig

[MobileTatsu-NJG]

The most interesting development came when the machine suddenly stopped killing ghosts and simply displayed the message: "The only way to win is not to play!"

Re:Oblig

[notnAP]

nah...

Much more interesting was the point a few minutes ago, when the researchers watched the AI somehow manage to change the game to Missile Command, at the same time that they noticed outside a massive rocket laun

Re:Oblig

[Remillard]

However we knew we'd achieved human-like intelligence when just after declaring "The only way to win is not to play" it then asked for a few more quarters because it ALMOST got to the next stage and it's sure it'll get there next time.

Sweet.

[toppavak]

Now we just need one that can play WoW for my friends so they can get their lives back!

Re:Sweet.

[DMUTPeregrine]

It's called Glider. http://www.mmoglider.com/ It's been around for a LOT longer than this Mrs. Pac man thing, as have many other game-playing bots.

Re:Sweet.

[celle]

Either get a baseball bat and smack your friends some sense or get some new friends. Actually forget smacking your friends, just their machines.

Smack your friends, their machines, and get some new friends. Ah, forget smacking your friends and their machines.

One of these things is not like the other...

[roystgnr]

The new AI game playing routines can handle Ms. Pacman, Tetris, and Baldur's Gate. Can their mathematics routines find sums of integers, roots of quadratics, and proofs of Fermat's Last Theorem?

Re:One of these things is not like the other...

[bunratty]

No, current AI does not exhibit general intelligence. That would be strong AI. We haven't developed it yet. The article is about weak AI, "the use of software to study or accomplish specific problem solving or reasoning tasks that do not encompass (or in some cases, are completely outside of) the full range of human cognitive abilities."

Re:One of these things is not like the other...

[UnknownSoldier]

> No, current AI does not exhibit general intelligence. That would be strong AI.

Whatever passes for "AI" would be better called Artificail Ignorance.

The "strong AI" you mentioned, I would call Artifical Intelligence and until Comp Sci and Biologists get a clue what consciousness is, we'll be forever stuck in the Artificial Ignorance mode.

Re:One of these things is not like the other...

[Anonymous Coward]

The new AI game playing routines can handle Ms. Pacman, Tetris, and Baldur's Gate. Can their mathematics routines find sums of integers, roots of quadratics, and proofs of Fermat's Last Theorem? No, and neither can most people.

(Let's review: to start with, most people don't know what an integer is...)

Re:One of these things is not like the other...

[kongit]

a whole number between 0-65535

Re:One of these things is not like the other...

[Anonymous Coward]

Fine, you just enjoy your as yet unappreciated by AIs rhetoric.
Meanwhile I'll be using methods from the "artificial ignorance" field, like simple pathfinding and computer vision heuristics, to get machines to do useful work.

Re:One of these things is not like the other...

[LandruBek]

actually tetris is np-hard

Re:One of these things is not like the other...

[Anonymous Coward]

Can their mathematics routines find sums of integers, roots of quadratics, and proofs of Fermat's Last Theorem?

Could the average human?

Re:One of these things is not like the other...

[Maian]

What's wrong with the term artificial intelligence? We study the "intelligence" of insects and other lifeforms, so why can't the same term be applied to artificial constructs? Why do you think that the term "intelligence" means human-level intelligence?

I think the term you're looking for is "artificial sentience" or "artificial consciousness".

Re:One of these things is not like the other...

[meringuoid]

The new AI game playing routines can handle Ms. Pacman, Tetris, and Baldur's Gate. Can their mathematics routines find sums of integers, roots of quadratics, and proofs of Fermat's Last Theorem?

Certainly Baldur's Gate, taken as a three-part epic campaign, is an immeasurably harder problem for AI than the other two; you have to be able to understand a whole lot of English dialogue, for a start. But the core game mechanism is very constrained. It's second-edition Dungeons and Dragons, a standardised rule set which is susceptible to AI attack.

Don't think of the computer working out how best to romance Jaheira. Think of it working out how best to take down Kangaxx. Or how to optimise a character build. It'll be an electronic neural-net munchkin.

Perfect Game?

[Sangui]

I've never been a big Ms. Pac Man player, always preferred the original, but when there's an AI that can pull off a perfect game then I'll be impressed, like that guy who got a perfect score on Pac Man without losing a life in the 80's. When the AI can do that it's done something. Not breaking 10,000 points? Meh.

Re:Perfect Game?

[Anonymous Coward]

I've never been a big Ms. Pac Man player, always preferred the original, but when there's an AI that can pull off a perfect game then I'll be impressed, like that guy who got a perfect score on Pac Man without losing a life in the 80's. When the AI can do that it's done something. Not breaking 10,000 points? Meh.

Pac Man never ends. There's no such thing as a perfect score. You just get a buffer overrun.

Re:Perfect Game?

[Anonymous Coward]

How can one lose that which he doesn't have? ;-)

Re:Perfect Game?

[greg1104]

The first perfect (meaning all the possible points were collected) game of Pac-Man wasn't until 1999 and was played by Billy Mitchell. It took him 17 years of playing to get that good. Here's some background. That page has one of my favorite quotes about the ill effects of video games:

Imagine a world in which Billy Mitchell never encountered Pac-Man. Put to good use his sharp mind, excellent hand-eye coordination, incredibly long attention span and his prodigious talent for problem-solving probably would have led the world into a utopian technological society by now. The human genome would have been mapped by the mid eighties. World poverty would have been eliminated entirely. The air and the earth would be clean. We'd be living in an age of unprecedented peace. Serbs and Kosovars would be frolicking hand in hand cracking jokes about their ethnic differences. Billy Mitchell would have a girlfriend. Instead, Billy Mitchell played Pac-Man and grew a moustache.

If you're ever near Weirs Beach, New Hampshire, be sure to visit Funspot--great arcade.

I'm a pretty good Ms. Pac Man player, and I consider my game a failure if I don't get the maximum of 14600 points on the first board. If the best the AI could do is averaging 8186 points per game, I think we're still pretty far from Skynet taking over.

Re:Perfect Game?

[Anonymous Coward]

that guy who got a perfect score on Pac Man without losing a life

I don't think you can lose something you never had...

Re:Perfect Game?

[bigstrat2003]

First of all, AI is in a rather basic stage compared to what we tend to expect from it. AI isn't going to be doing anything impressive any time soon, but that doesn't make progress less significant.

Second of all, getting a perfect score on Pac-Man without losing a life isn't that impressive to me, considering that by learning a handful of patterns you can play a perfect game (as long as you don't fuck up and mis-time a turn or something).

Re:Perfect Game?

[RyuuzakiTetsuya]

they don't have a DDR or any game made after 1992. Why bother? Classic gaming can suck my balls.

ghost routes

[Wescotte]

I'd almost be more impressed if it could have learned the routes of the ghosts in Pac-Man vs learning to avoid random movements. AI in the Ms Pac-Man game just needs to run away while to succeed in Pac-Man you need to first realize there are planned routes and then learn them.

Re:ghost routes

[Johnny Hardcore]

The routes in Ms. Pacman are randomized at the beginning, but they follow certain set behaviors after. Have a look at the article about the Bozeman Montana think tank. I gotta memorize this sometime and beat the high score at work ;)

Other uses

[TheSpengo]

This is cool, being able to choose smart moves against a random opponent could have a lot of uses in enemy AI in other games too. The unpredictability of a human opponent has always been an issue when creating realistic AI. It always kind of bugged me that even in new advanced games like Crysis, enemies will sometimes move in the most stupid ways possible. The next generation of FPS AI could use something similar to this.

Re:Other uses

[greyhueofdoubt]

I agree, but you have to admit that the humans often make the stupidest moves on those maps, too.

-b

Can the AI play Tic Tac Toe?

[Joe The Dragon]

Can the AI play Tic Tac Toe?

Re:Can the AI play Tic Tac Toe?

[mad_minstrel]

But can it play Twister?

The difference between Pac-Man and Ms Pac-Man

[Kamineko]

That's it? Get right out of town!

Re:The difference between Pac-Man and Ms Pac-Man

[blogan]

Ms Pac-Man also has a bow.

The 1990s called, they want their aimbots back

[Rogerborg]

Honest to Jebus, I was writing Netrek bots in 1994 that used a genetic algorithm to self-guide their development, and you don't get more "random" than human opponents. When all those Quake bots hit the scene a couple of years later, it was already old hat as far as I was concerned, and now some Korean MMOs are almost entirely populated by robots. Are people really still getting grants for this?

Re:The 1990s called, they want their aimbots back

[Dutch Gun]

some Korean MMOs are almost entirely populated by robots Keep in mind that MMO "robots" (more typically called "bots"), are mostly automated scripts that utilize very specialized record-and-playback functionality combined with techniques for screen-analysis, such as recognizing the name of a piece of text used as a navigation marker. These bots exploit the predictive and repetitive nature of MMOs, such as the fact that a particular creature will always spawn in the same location, that a vendor will be in the same place all the time, that the same sequence of actions will consistently lead to the exact same outcome in combat, etc. A human player simply sets up a character in the game with the required equipment / spells, records the proper sequence of actions to defeat the creature, collect the loot, then travel back to town and sell loot to vendor.

While the bots' systems can be sophisticated in their own right, in no way are these bots exhibiting any sort of AI behavior. In that sense, what the Pac-Man AI is doing is much more interesting. Even in modern games, learning AIs are rarely used, except occasionally in initial training of opponents. I've written AI for several commercial games, and in general, it's much easier to simulate intelligence with simpler algorithms than to use true machine learning. BTW, creating a killer bot is really not all that hard. Computers have far better reflexes than people do. It's much harder to create a computer opponent that's entertaining to play against.

Re:The 1990s called, they want their aimbots back

[Sigma 7]

When all those Quake bots hit the scene a couple of years later, it was already old hat as far as I was concerned IIRC, the Quake bots were severely limited by the platform they were running on. In particular, they couldn't "see" the map directly and had to rely on waypoints. These waypoints took up space in the 600/768 entity limit which made the bots fail if you tried using them on large maps.

These bots also need to know how maps work - either by seeing players proceed through the map or by having a developer setup waypoints for the bots. In the first case, bots would be confused by complex map structures beyond a simple press button to open door. In the second case, it's time consuming, works only on a per-map basis, and needs to have capability of understanding

In a way, Netrek bots are easier to write, since there is no worry about as many obstacles that appear in a Quake map. You also have much better access to the source code.

Re:The 1990s called, they want their aimbots back

[Rogerborg]

My point, which was admittedly very badly made, is that as the environments that real humans actually care about have become more complex, that 'successful' (i.e. extant in the wild) 'AIs' have become more primitive. You could view that as lazy devolution, or as honing away the parts that nobody (in the real, funds-delivering) world actually cares about. I guess your experience indicates that it's the latter case.

Academic AI research still seems to be producing solutions to the problems of the 1980s. The method may be interesting, but if the application is (and remains) obsolete, then what is the point in funding it? If these Goddamn eggheads actually had to pay their own way, then perhaps MMO bots would be more than just scripted doofuses. The smart bots would be bushwhacking and looting the corpses of the dumb ones just before they enter town to sell their loot.

Meat Eating AI

[liak12345]

So we're now creating AIs that are learning how to eat things and we have that run on meat.

Nope. I don't see any way how this could result in the destruction of the human race!

The key difference...

[MacarooMac]

... between the AI and humans tested playing the game appears to be (last para) that the humans were able to adapt their tactics.

E.g. They learned to lure ghosts close to Ms Pac-man so they would easuer to catch and eat once they became edible.

I'm sure this tactic could be programmed as a new rule and added to appropriate position on the AI's 'priority' list.
But until this 'cross-entropy' learning method (and any other AI learning technique for that) can truly teach the AI to adapt by itself - from it's own observations - then it's just not proper AI, imo.

Angband is more complex than Pacman

[Old Wolf]

...and we've had Angband Borg for some time (which is very impressive!)

Re:Angband is more complex than Pacman

[Carnildo]

As I recall, Angband has a game-breaking strategy that requires spending hundreds of thousands of turns on the first level. It's too tedious for most humans, but computers don't get bored. The presence of this strategy makes writing an AI for Angband easier than it is for most other games.

Good Morning Professor Falkin

[ChainedFei]

Would you like to Play a game?

Re:Good Morning Professor Falkin

[KillerBob]

I would like to play Global Nuclear War. :)

Re:Good Morning Professor Falkin

[Anonymous Coward]

Theater-wide biotoxin and chemical warfare, please.

("Global thermonuclear war" is so overdone; I'm bored of that one. It's like there's no point even in playing it.)

Skynet?

[RyogaHibiki]

First Ms. Pac-Man... next Skynet.... I'd pull the plug if I were teaching the PC before it's too late.

Unfair

[Anonymous Coward]

Oh, sure. These guys get praised for their great research, but when I teach my computer to play WoW I get banned!

And I always thought...

[leenoble_uk]

...that the only difference was that Ms Pac Man had a bow in her hair.

One of these things just doesn't belong!

[Eightyford]

Tetris, Ms. Pac-Man, and Baldur's Gate... One of these things just doesn't belong!

Koza?

[lalena]

Thought John Koza had Genetic Programs playing pac man over 15 years ago.

xscreensaver

[gringer]

There is an xscreensaver hack that is a pacman game with various level styles. I suspect that the monsters in that are a bit more random in their movement. However, the monsters move slower than pacman, and the pacman currently seems rather stupid, running towards monsters, and just collecting air when there's still plenty of pills to pick up. It would be nice to work on the AI in that, then I'd get a more interesting screensaver to watch.

still some time to keep up to Go (weiqi)

[Sem_D_D]

I'd really like to see these guy's algorithms in the game of Go (wei-qi).
Or imagine one day being able to get some machine learning from Guo Juan http://www.guojuangoschool.com/ - she's pretty nice and (still) understandable high-dan player.
Recently some Chinese professor that participated in the Deep Blue's victory over Gary said with the help of M$ research money he's inking closer on the brute-force approach in Go next http://www.spectrum.ieee.org/oct07/5552 . But it is still a steep curve. The chinese chess projections point to a breakthrough in the next 2-3 years and it is still a couple of exponents simpler than Go.
So, way to go, Ms.Pac-Man :-)

Ms PacMan is most definitely deterministic

[howard_coward]

The authors state "In most of Pac-Man's sequels, most notably in Ms. Pac-Man, randomness was added to the movement of the ghosts." Anyone who has played MsPacman fully understands that while there is __some__ random behavior, the ghosts are most definitely deterministic. The aggressive red guy (Blinky) is very very different from the passive yellow guy (Clyde). Knowing this is necessary to play the game in an intelligent manner, if you can imagine such a thing. You can always go to MameWorld, get the 6502 source code and see for yourself!

Re:Ms PacMan is most definitely deterministic

[Anonymous Coward]

> You can always go to MameWorld, get the 6502 source code and see for yourself!

Someone ported the Z80 original to 6502?

Re:Ms PacMan is most definitely deterministic

[_madness]

To the untrained eye Ms. Pac Man does appear random. At the beginning of each level the Blinky and Pinky do move randomly (first 5 seconds). Inky and Sue (orange) move in the same manner as in the original Pac Man and are predictable.

Grouping strategies are used to comprehensively beat Ms. Pac Man and have been used to complete the game up to level 138 at which point it becomes unplayable (First accomplished by a trio in 1983 near Montanta). It is not hard to find more information on this.

It can be shown that Ms. Pac Man is definitely not deterministic. The movement of the ghosts are predictable but one does require extensive practice and familiarity with the game. The most predictable ghost is Blinky and the most complicated is considered to be Inky who's movement is influenced by how close Blinky is (and so it goes on).

Greetings, Professor Falken

[StarReaver]

How about a nice game of Ms. Pac-Man?

Re:Greetings, Professor Falken

[base3]

Later. Right now, let's play global thermonuclear war.

How about Global Thermonuclear War?

[paintballer1087]

Shall we play a game?

Love to, how about Global Thermonuclear War?

Wouldn't you prefer a nice game of Chess?

Programmed to play Pacman

[globaljustin]

The summary was wrong, should read "AI programmed to play Pacman" I agree that AI is overhyped. Now we can debate the definition of "AI" for days but the fact is, this is simple programming. You tell the computer how to do something, and it does it (heh, i know it's not that simple, but the idea is that simple). AI is a fun topic. But ultimately the question of really defining Artificial intelligence is connected to how we define Human thought. In an abstract sense, humans are just programmable meat bags, controlled by what we sense from the environment and by 'hardwired' innate reactions. The bottom line is, computers will never do anything we don't tell them to do (implicitly or explicitly), and thus, AI will always be limited by human intelligence. ps...i know how funny part of that last line sounds 'computers will never do anything we don't tell them to' when taken in the context of a Blue Screen of Death or some such failure, but remember, a human (loose term) was responsible for the faulty program with the errors that cause the BSD, so my analogy still holds up

Re:Programmed to play Pacman

[eyenot]

Sorry I couldn't respond sooner, because of the holiday (I use public terminals).

I thought about this same thing since I last responded. Another user had mentioned the failures of the 80's, and I was planning a response along the lines of "perhaps those failures were mostly the result of no clear distinction in a field somewhere between philosophy, psychology, and programming, with the only real anchor being axiomatics". However, my original comment was entirely about how these various other subjects of the field have been left behind, and now the field is apparently contented with any automated process normally reserved for people.

There's really no need to school me in the fine points of human error, as I've been programming computers since I was eight, which would be since 1986. And I frankly don't believe there is a single person alive currently or in the past who is qualified enough to make such a statement as this: "computers will never do anything we don't tell them to do". It seems a fairly short-sighted statement, on its face.

Re:Programmed to play Pacman

[globaljustin]

thnx for the response

"computers will never do anything we don't tell them to do"

so let's talk about this statement. For the purposes of this little discussion, you can consider me a philosopher. Which is another way of saying I only understand the basic basics of programming. But I can make some relevant points in a discussion about AI...

I am going to stick with my statement. I do not know the jargon, but even AI that 'learns' was still programed to learn in a certain way by a human who had to look at several options for programing a computer to learn, and then pick the best one. At the core, it's still decisions made by a human that cause/predict AI behavior.

In that sense, the largest sense, AI (computers...whathaveyou) will not do anything a human does not at some point, 'tell' or maybe enable it to do.

C'mon, be a sport and give me an example...but don't jargon it up too much, i only have so much time to look things up on wikipedia.

What about tool assisted speed runs?

[Anonymous Coward]

People have been using similar things for years to find the most optimal routes for speed runs. These scripts are known as "TASbots" and are used on all sorts of games, mostly 2D platformers.

If you're interested you can check out a bunch of videos made with the help of these tools here.

How long?

[NRISecretAgent]

How long do you think it would take for the AI to figure out the random system of the ghosts and started anticipating all of their moves? That would be interesting.

Re:How long?

[the donner party]

Even in a simple game like Pac-Man, the state space is so large that learning to anticipate complex patterns can take millions of years, unless the programmer assists the learning algorithm by providing carefully crafter building blocks for the player actions. It seems that they did just this for their Ms. Pac Man study.

just sick

[goga_russian]

next thing you know there is Money missing and your Daughter is knocked up.

Everytime...

[Anonymous Coward]

...i see one of these articles, i'm crossing fingers it's not my tax money going to waste here. I kinda get this feeling, that these kinds of scientific AI projects always end with "Neat script doing neat little thing".
Why do they always set that goal? I doubt they come any closer to making a real AI program by making hundres of small programs with individuel tasks.

Take a look at CS or UT bots from the gaming area...way more advanced than any of these small programs. Why do they then keep doing the same "research" over and over?

*sigh*

[jpellino]

I rocked at this game back in the day.
Now I've been pwned by a largish calculator.

AI vs. Chimp

[Paaskonijn]

Let's pitch the AI against a chimp.

Replicate at home

[alito]

I'll take this opportunity to advertise the SDLMame interface at http://organicrobot.com/mame/ in case anyone wants to try to replicate this with Ms Pacman or any other MAME game they've got.

You have to bring your own visual parsing tools though (I recommend starting with OpenCV), since all you get is a bitmap, not the ghosts position, number of dots, state of ghosts, etc that these researchers had access to by using a different implementation of Ms Pacman.

(btw, having seen a few very smart people try getting a computer to play Pacman, the results obtained are very impressive. As per usual, 4/5ths of the objections/questions/disses posted here could be answered by reading the paper linked to at the bottom of the article. JAIR is open and free, you do not need to pay or subscribe to read it)

I wonder ...

[LrdDimwit]

> However, the ghosts' routes are deterministic, enabling players to find patterns and predict future
> movements. In Ms. Pac-Man, on the other hand, the ghosts' routes are randomized, so that players can't
> figure out an optimal action sequence in advance.

How sure are they that this AI hasn't simply learned how the random number generator works, so it CAN predict the ghost's movement patterns? Unless the random number generator is reseeded at unpredictable and unmanipulable intervals, then it will be subject to adaptive learning techniques used to figure out the seed.

Re:I wonder ...

[LrdDimwit]

So TFA implies the techniques they used are different, that they actually taught the machine how to play the game and gave it a rules-based AI, as opposed to something like genetically evolve a program that executes aribtrary code to map the inputs to the outputs. But I'm not very familiar with these techniques ...

Re:I wonder ...

[Griffon26]

Exactly.

I once made a program that simulated evolution of a population of creatures going after food placed randomly on the screen.
The 'DNA' determined the direction they were likely to move in relative to the food as well as size of the step they took in that direction.

I expected the creatures to evolve to a 100% chance of moving in the direction of the food, but it turned out that that chance settled at about 75%.
Even if I gave half of the initial set of creatures a 100% preference of going towards the food, it would still be evolved out (ruling out a local maximum at 75%).
Interestingly most of the remaining 25% ended up in the "to the left when facing the food" direction. This led me to conclude that the random generator
was the cause of this (something like the next food likely to appear to the left of the current one).

Hold on a sec...

[Snowmit]

Besides Ms. Pac-Man, game like Tetris and Baldur's Gate assist these programs in mapping different behaviors onto their artificial test subjects

BALDUR'S GATE?!

Re: Computers wasting time =

[TaoPhoenix]

Does constantly running strange processes for no clear reason and rebooting explorer daily count as wasting time?

AI can't replicate consciousness

[yters]

Otherwise it could replicate me. However, I know this can't be the case. Anytime someone tells me they've successfully created a machine to replicate me, I know they're lying because I'm still me. If they'd replicated me, then I'd be the machine. But, blow the machine up, torture it cruelly, give it pleasure unimaginable, whatever. None of that affects me at all. Then, considering that any truly significant insight I ever have had requires self consciousness, I highly doubt any kind of machine can be truly intelligent.

Who's Al?

[lawnsprinkler]

And why is there so many of him?

Modern?

[kazade84]

the Department of Information Systems at Eotvos University in Hungary is working to adapt that thinking to more modern titles. Besides Ms. Pac-Man, game like Tetris and Baldur's Gate...

I really have to cut down on what I drink, I've woken up in the 80s again.

Screw Pacman...

[Anonymous Coward]

...what I really want to know, since the AI played Baldurs Gate, is of what alignment our new gameplaying artificially intelligent overlords are.

Re:Screw Pacman...

[meringuoid]

...what I really want to know, since the AI played Baldurs Gate, is of what alignment our new gameplaying artificially intelligent overlords are.

The constructs within the Rubikon Dungeon Research Project are of lawful neutral alignment, although for the purposes of evaluating adventurer responses they are playing a chaotic evil role. Die in the name of the Evil Wizard!

AI not AL

[andrewjhall]

I read the title as AL taught how to play Ms Pac Man and my first thought was "damn, the day Slashdot reports on Al Gore playing old computer games is the day I stop visiting."

Happily Slashdot is fine - it's my brain that is malfunctioning.

Re:AI not AL

[Anonymous Coward]

stfu.. nobody cares...

They didnt solve MsPacman (carp, carp)

[howard_coward]

They (google NoN-SeNS Pacman 1.6 with C sourcecode to get the original Courtillat pacman stuff) apparenmtly addressed an artificial MsPacman implementation. People should realize that Pacman is a particular implementation (pacman.zip in mame, the original Midway game) as is Ms Pacman (the mspacman rom in mame). IT DOES NO GOOD TO SOLVE A DERIVATIVE PACMAN OR MSPACMAN! We went through this in the original shannon 4x4 chess board which remained a curiosity until a real chess board with real chess moves (e.g. en passant capture) was used. Same with checkers. In this day and age there is no excuse to dumb down the game. It is my basically uninformed but strongly held opinion that the originals are the real thing. Solving a game that kindof looks like ms pacman should be discouraged. C'mon guys!

Not entirely

[Xest]

Whilst this is indeed a clear case of weak AI, it's not quite as simple as the weak AI vs. strong AI thing. Weak AI in itself can be broken down into different levels, the AI mentioned in this article seems to be just a run of the mill application of symbolic AI, and whilst symbolic AI.

Because such programs like this are the ones that for some reason make the headlines they're also the ones that make people think "well, AI is a bit of a let down then really isn't it" but weak AI goes further than just symbolic AI, many accept that symbolic AI has a much less promising future right now and all the important research that goes into AI that might actually seem more intelligent is in the field of biologically inspired AI which is still right now only providing us with weak AI, but it's weak AI that is often so much more impressive than old fashioned rehashes of various symbolic AI implementations as per this article.

I'm not suggesting that the basis of your comment was wrong - you're certainly correct that this type of AI really isn't that impressive and as such it's hard to call it intelligent, however I do feel you were wrong to dismiss current AI as a whole with your comment that it's the only type of AI we have, because this specific type of AI (symbolic AI) really isn't the only type we have. Biologically inspired computing has been pushing symbolic AI out the spotlight for over a decade now and absolutely demolishing it in terms of the impressive kind of demonstrations it's bringing forth.

Of course, that's not to say symbolic AI is worthless either, it's certainly has it's place for things like this, chess and for expert systems and so forth. The biggest problem is that when AI becomes commonplace and people implement it left right and centre they find it easy to overlook the intelligent part of it and see it as "just another bit of code to implement". This is essentially the case with the poster you were replying to - we find it easier to call something intelligent if we don't understand it ourselves, but when we understand it ourselves we don't see it as intelligent.

Re:Can it play Twister?

[Anonymous Coward]

I always wondered where "bite my shiny metal ***" came from...

been done...

[Anonymous Coward]

look up automouse sometime. Tons of people use tools like it to play mmorpg games and gain loads of skills as they sleep.

Ms. Pac-Man is random...and yet NOT random...

[pacmantab]

One thing to note: The only ghosts that are random in Ms. Pac-Man are the Red (Blinky) and Pink (Pinky) ghosts. And they are random for only the first few seconds of each level. So... while that would elminate the ability to use Pac-Man-like patterns, you would be able to determine the best way to go in situations after the first few seconds based on what is known of each ghost's tracking strategy - as they would then be just as predictable as the original Pac-Man.

Cool! A Minnie Driver/Anne Hathaway love scene!

[Impy the Impiuos Imp]

The article continues:

"Researchers report the first thing the AI did was go download a bot to play the game the right way."