AI System Invents New Card Games (For Humans) 112
jtogel writes "This New Scientist article describes our AI system that automatically generates card games. The article contains a description of a playable card game generated by our system. But card games are just the beginning... The card game generator is a part of a larger project to automatise all of game development using artificial intelligence methods — we're also working on level generation for a variety of different games, and on rule generation for simple arcade-like games."
Good luck with that (Score:5, Insightful)
Creating games or levels is pretty simple (well, relatively speaking) in comparison to making them fun. Bu the myriad of bad games out there, I would say that making good games or levels is something not even natural intelligence masters routinely. It is bound to fail trying to do it with AI. Nonetheless a nice research benchmark. But please stop trying to imply real-world usability where there is none. It is unethical and unprofessional.
Re: (Score:3, Insightful)
TBH, I don't think we even play the best games out there routinely - at least for board games. Monopoly, for instance, is a really shitty game with everything in favor of the guy who lands on good properties and then drags on forever. But it's one of those board games nearly every family has in their closet.
Re: (Score:3)
Re:Good luck with that (Score:4, Insightful)
The game mechanic itself isn't. The human interaction that arises out of the game mechanics is.
Most of the fun in board games is about having a shared experience; a context to talk about.
Buying and selling streets and houses is boring. Continuously calculating and counting money is tedious. Making jokes about it is fun.
Re: (Score:2, Interesting)
Interestingly, Monopoly is a lot better when you play with the auction rule that everyone ignores. The official rules also include a couple of altered games with fixed time limits, to prevent the dragging-on that occurs when you omit the auction rule.
Re:Good luck with that (Score:4, Informative)
Re: (Score:2)
Re: (Score:1)
You are the only person I have ever met that plays that way, then.
uh, not to kill the mood here but ... you haven't actually met this person.
just sayin'
sorry, still no friends.
Re: (Score:2)
Re:Good luck with that (Score:4, Insightful)
But please stop trying to imply real-world usability where there is none. It is unethical and unprofessional.
But it keeps the grant money coming.
What they've created is a method for representing card games symbolically (probably the hardest part of the project). Then they searched through many permutations of games, and keeping the ones that pass an acceptance criteria. It's AI the same way Prolog is AI.
Or depth first search. Is depth first search AI? Does an A* search make a machine intelligence? We need a new tag, SearchIsNotAI or something.
Re:Good luck with that (Score:4, Interesting)
Re: (Score:2)
Re:Good luck with that (Score:4, Interesting)
It's AI the same way Prolog is AI. ...
SearchIsNotAI or something.
HUH? You definitely lost me there. First, Prolog is a computer language more than any kind of algorithm, just one more declarative and suited for logic. Definitely a lot of AI has been coded in Prolog.
Second, how is search not AI??? Almost any AI algorithm I can think of is a search problem. Chess (or other games) AI is nothing else than a search for a close to optimal set of moves (based on a scoring function). SLAM and Path-finding in general is also a search. Watson performs a search for potential documents matching the query. Classifiers search for an optimal decision boundary to divide the data. Clustering searches for a stable configuration of centroids (for example). Object recognition searches for matches that maximize the likelihood between object... etcetera, etcetera, etcetera. I mean, almost any algorithm that I have been teach in Machine Learning and Robotics has been introduced as a search problem!
Re: (Score:2)
Do you consider a sorting algorithm to be AI? Creating ontology and searching through it does not require intelligence, for reasons similar to those described in this paper [cogprints.org].
Re: (Score:2)
Re: (Score:2)
What I'm saying it that a lot of what is considered AI by the people that do AI is based on search.
This is true, you are right. It is also true, unfortunately, that a lot of what is considered AI by the people that do AI has nothing to with intelligence.
Re: (Score:2)
a lot of what is considered AI by the people that do AI has nothing to with intelligence.
No it has to do with automating reasoning. Intelligence is so vaguely defined that two people could have an opposite opinion on the importance of rational tough in the definition of intelligence and they would both be right be right depending on which school of thoughts you belong. I suggest you read a little bit in the following encyclopedia : starting at that page [stanford.edu]
Re: (Score:2)
I suggest you stop messing around with encyclopedias and start reading journals. Here is a good place to start [cogprints.org], as mentioned earlier.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
But please stop trying to imply real-world usability where there is none. It is unethical and unprofessional.
But it keeps the grant money coming.
What they've created is a method for representing card games symbolically (probably the hardest part of the project). Then they searched through many permutations of games, and keeping the ones that pass an acceptance criteria. It's AI the same way Prolog is AI.
Or depth first search. Is depth first search AI? Does an A* search make a machine intelligence? We need a new tag, SearchIsNotAI or something.
Search is an integral part of AI. Always has been. One famous AI scientist once said that all AI is just search and knowledge representation. It is also the case that once we find an AI algorithm that works sucessfully it seems to no longer be considered AI.
The algorithm here is a small incremental improvement, thats how breakthroughs work - bit by bit. The scientists may not have claimed it as a hugh breakthrough, I would guess the journalists did that if anyone did.
Re: (Score:2)
Re: (Score:1)
But it keeps the grant money coming.
What they've created is a method for representing card games symbolically (probably the hardest part of the project). Then they searched through many permutations of games, and keeping the ones that pass an acceptance criteria. It's AI the same way Prolog is AI.
Or depth first search. Is depth first search AI? Does an A* search make a machine intelligence? We need a new tag, SearchIsNotAI or something.
Yes, it's AI. It came out of AI research, as part of the path to full AI, and is a natural part of what the only intelligent species we know of, does.
Re: (Score:2)
as part of the path to full AI,
Probably not, but that's speculation anyway.
Re: (Score:1)
It is bound to fail trying to do it with AI.
And still you know it will work very well in the long run, like almost every other task that was set to fail with AI.
Re: (Score:2)
People keep attributing magical properties to their own intelligence that no mechanical machine could possible possess.
The simple truth is that our brains are just as mechanical as any machine, albeit a very impressive machine.
Re: (Score:2)
I don't know about the impressibe part - Have you seen *some* people?
Re: (Score:2, Funny)
As far as I can tell, "AI" has succeeded only in keeping the same name after endless redefinitions resulting from it's numerous failures.
Your blind faith in AI seems to indicate that you're either hopelessly misguided or one of those singularity nuts. I can only hope that you've merely been mislead...
(To Ray's deluded followers: Just get over the pretense and just worship Kurzweil the prophet outright. Failing that, at least get robes. Not only will they keep you warm, they make you and your fellow cult
Re: (Score:3)
As far as I can tell, "AI" has succeeded only in keeping the same name after endless redefinitions resulting from it's numerous failures.
Your blind faith in AI seems to indicate that you're either hopelessly misguided or one of those singularity nuts.
Absolutely not. The improvement in fields that were said to be impossible for AI are just astonishing, and I am among the first surprised by such successes. Let's state it clear: computing power is increasing, theoretical models are improving, practical implementations are getting more efficient. So yeah, basically Turing was right, we are just impressively capable computers and nothing more.
Around 5 or 6 years ago, there were some image classification benchmarks that were incredibly tough and said to be al
Re: (Score:2)
Everything that allows a computer to make a statement that you thought was only possible by a human is AI.
And the redefinition continues. This one is great, as it's both overly broad and completely subjective. As a bonus, it defines "AI" exclusively in terms of successes.
Hey, now it can't fail!
Re: (Score:1)
Re: (Score:2)
There are a LOT of games that have random level generation. Rogue being one of the first. It's random level generation being ones of its most important features. I could see this sort of thing being used in that sort of way.
Automate all game development? (Score:5, Funny)
Shall we play Global Thermonuclear War?
Re: (Score:3, Insightful)
How about a nice game of chess?
Re: (Score:2)
Re: (Score:3)
Shall we play Global Thermonuclear War?
Sure, I love that game [wikipedia.org].
OK, so it's not what you were referencing, but it is a card game, and the name is close, and it is quite fun. And, it's probably not the sort of game that an AI could come up with.
Re: (Score:1)
The Past, also: (Score:5, Interesting)
I see MMO expansions someday taking this route to expedite content generation. Players complain there's not enough content? Drag and drop your quest generator with a bit of human tweaking and you're good to go. I'm sure some of the systems in Eve were generated partly through random generation.
It turns out that procedural generation is conceptually pretty easy; but making it good is much harder. Early videogames(from the era where memory and storage constraints were Serious Business) and demoscene stuff(where the constraints are wholly artificial; but that's the point of the exercise) are pretty much forced to rely on it heavily because they simply didn't have the option of storing canned content.
Today, though, you see games with substantially greater amounts of (not inexpensive) artists and designers thrown at them, and gigabytes of art assets, with hand-tweaking especially evident in places where the player is likely to look closely(eg. generic NPCs will be thrown together from parts, giving the world a varied population without requiring the art people to hand-model 10,000 different 'bandit' characters; but the risk of output that just looks a little off, or hit a few branches of the ugly tree on the way down, means that those critical NPCs that follow you around for half the game had their appearance nailed down precisely). The fact that artists are slow and expensive has created a demand for procedural generation tools, and quite a few exist(I'll just mention SpeedTree, purely because the phrase "SpeedTree for Games has been the gaming industry's premier vegetation solution since 2002" amuses me); but the problem of creating really good environments continues to be vexing enough that titles that can afford it throw a lot of humans at the problem.
Re:The Past, also: (Score:5, Interesting)
Re: (Score:2)
Don't get me wrong, I hardly think that procedural generation is an intractable problem, or that the AAAs are on the leading edge of good game design(though it is worth noting that some have actually retreated from doing procedural generation, rather than merely not embraced it: TES II: Daggerfall was enormous, ~2x the size of Britain and the thick end of a million NPCs, TES III: Morrowind, moved to a teeny hand-built environment because TES II was judged as having a rather... lifeless... and barren feel de
Re: (Score:2)
This is what really messed up Daggerfall (ie, Elder Scrolls 2). Just about everything was randomly generated. Randomly generated mad-lib style quests, random 3d dungeons (some of which prevented you from reaching the quest objective), hundreds of towns, thousands of NPCs. After awhile it was just the same old thing.
Even humans have troubles creating some sort of content that is interesting, a computer churning it out won't do any better. It's a "safe" bet to just get more artists and require higher end
The rules from the TFA ... (Score:5, Funny)
Re: (Score:2)
Re: (Score:2)
Ah Programmers... (Score:5, Funny)
1) Obsolescence of all current vocational knowledge in their field on 5-15 year scales.
2) The ultimate goal of their work is the removal of their job position from the market (the singularity which can hack in C).
Re:Ah Programmers... (Score:4, Insightful)
Arguably, at least some branches of science and medicine have some major overlaps(though the timescale tends to be longer because reality is stubborn and complex, and just gets worse the deeper you go).
Major credit accrues to those who develop new models that render the old ones obsolete or deeply incomplete, and discover new phenomena that require a course of study distinct from the old ones.
And, while there isn't any major risk of them succeeding themselves out of business, Team Epidemiology is always trying to wipe out one pathogen or another. It doesn't have quite the same finality as 'the singularity'; but that's mostly because they are chasing a bunch of moving targets.
Re: (Score:2)
2) ...at which point the individuals who actually succeeded would be extremely wealthy (and really so would the human race)
Except for the ones who get fed into the matter decompilers to produce more computronium...
Re: (Score:3)
Believe me, when you look at what programmers routinely generate, there is no risk of them becoming obsolete anytime soon (unless those that pay them recognize the scam and either go non-computer again or hire those few that are actually good at it). Most code is so bad that you should actually throw it away before it makes it into production. Most other code will have to be replaced in a few years. When you find something that it really good, it very often is old and whoever write it did just happen to und
Re: (Score:2)
I agree, but I think these are covered by architecture, design, implementation, as these are just more specific aspects of each that depend on the actual business case for creating the software. One thing I missed: You have to be a reasonable technical writer! Otherwise no useful documentation, project plan, presentation to management, etc.
And you are spot-on about the salary: Highly capable, experienced experts should at lest be at 5-10x of those that can only program somewhat. No surprise true experts are
Re: (Score:3)
Programmers make me laugh hysterically sometimes. Seriously, when in the history of man has an entire portion of an industry been dedicated to the following two goals: 1) Obsolescence of all current vocational knowledge in their field on 5-15 year scales. 2) The ultimate goal of their work is the removal of their job position from the market (the singularity which can hack in C).
OK. I'll bite:
0) Machine Intelligence systems are great for helping humans -- Luxury car break controls for when your attention is lacking, Segues, for when your balance is lacking, Self driving car for when you need to take a nap on that long drive, Automatic terrain creation so you don't have to piddle with setting each stone and tree, you can just generate a bunch of settings until you find a cool one, then sculpt the land a bit more way you like to add more visual interests afterwards, Which is how
Re: (Score:2)
The entire reason behind technology is the elimination of human labor.
Not necessarily. There are all sorts of incentives for technology-- increasing human happiness, increasing safety, preserving the environment. A few (randomly chosen) examples of the above: the electric guitar, quick release ski boots, catalytic converters. None of those had anything to do with "the elimination of human labor".
Re: (Score:2)
"Those who fail to learn history are doomed to repeat it."
The same was said about steam powered machines at the dawn of industrialization.
Re: (Score:2)
Programmers make me laugh hysterically sometimes. Seriously, when in the history of man has an entire portion of an industry been dedicated to the following two goals:
1) Obsolescence of all current vocational knowledge in their field on 5-15 year scales.
2) The ultimate goal of their work is the removal of their job position from the market (the singularity which can hack in C).
practically every industry is geared towards removal of job positions. all industrialism is part of that. you can just go back to burning spinning jennies though.
(practically none of game development industry is geared towards true ai though, if you think that then you're a victim of some pretty effective marketing)
Re: (Score:2)
1) To create something that does the hard work for you.
Seen this way, in the history of man, almost EVERY industry has been trying to do this for a long, long time. The issue here is not technical development, is the fact that one we achieve something, we always come up with something else to go for and maybe an issue of how we should distribute the benefits of those developments.
Machine learning game strategies (Score:5, Interesting)
On a tangentally related idea, we're working on a project of machine learning to take games and the rules of play, then derive strategy based on the rules.
Nothing particularly new, except we don't define what winning is, just the rules of the game. No hint is given to what constitutes good play, or even what "playing" is. Although it is a very slow process depending on game complexity (learning can take weeks and sometimes months of processing time), it requires no real programming effort, beause we don't have to know what "good" play is or some series of algorithms; it produces better and better tactics and strategies of play during the learning process, by experimenting with the rules, how to play, and such.
What's cool about this, is that you can watch it teaching itself different strategies and tactics. Some of the "tactics" it creates are many times counter intuitive or plain bizarre, but based on the overall strategies it develops, allows for some really different playing experiences as it doesn't follow human game logic based on experience with "similar" games or "intuition".
Re:Machine learning game strategies (Score:5, Interesting)
Let me clarify that, as that statement was misleading.
We don't program what winning is as any function of the strategy. The system comes up with several strategies, which all play against each other. At the end of a series of competitions, a strategy is told "Hey, you played against a bunch of different people, you won more than the rest. We don't define what winning is, how it won, or even what winning is, we just tell the system that strategy 1532 was the best. The system knows what strategies work better than others, so it can learn what methods are more successful. The system doesn't know why it won, just that when it made certain decisions it won more often. We don't even tell it on each game, we tell it after an aggregation of multiple competitions how it did. By comparing all the strategies it tried, then it develops better and more complex ways to win (which we didn't tell it how to do).
Even more interesting is when it comes up with what is considered doctrinal tactics that humans have arrived at to win as well (or statistically increase the chances of such) although no such logic was included in the programming.
The benefit to this is that although it takes a LONG time to develop "good" strategies, it comes up with completely unique and novel approaches to winning, even though it doesn't know how exactly it won, only that its strategy wins more than everyone else.
The benefit to us is we just tell it the game rules, we don't have to come up with any specific playing algorithm, the learning system figures that out. We just tell it the rules, whether they are concrete like in chess (bishops move diagonally, pawns move one, or start with two, etc) or variable rules based on other complexity factors. Whether its poker or chess or military tactics, the systems job is to come up with the strategy. How good or complex that strategy is allowed to be, is a function of how much processing time we want to give the system to learn the best way to win.
Re: (Score:1)
In other words, it's an evolution of strategies.
Re: Machine learning game strategies (Score:5, Interesting)
We use several forms of evolutionary programming in several sections of the learning systems' areas.
There are hybridized genetic algorithms in the portions involving the strategy blending evolution system, which does a few different forms of strategy selection pressure and evolution controls, which is critical due to training time to not cause premature convergence or genetic instability.
Additionally, we introduce additonal factors such a genetic drift and migration so that out competing strategies can evolve independently as the explore the strategy plane.
There are macro level evolution techniques to handle the complexity growth of the strategy species, so that the complexity can be altered depending on how "advanced" the system needs to be. In a simple sense of a turn based game, it would equate to the number of plies or analysis depth you would go. For more complex multiobjective systems, like military tactics involving minimizing casualties, civilian losses, maximizing kill or capture of enemy units, minimizing structural damage to infrastructure, etc., then it modifies the strategy complexity. For example, you could send eveyone with guns to kill everyone, or you could parallel it on intelligence gathering with drone units to direct fire, long range snipers or diversionary tactics, or factoring logistical support costs.
A lot of the core work is maximizing the efficiency of the evolutionary strategies, as they are the biggest fator in learning time. It's really easy to write inefficient logic that ends up taking much longer to arrive at good solutions without getting lost due to too much noise or oscillation in the system.
Another method that is used is a version of PSO, which is used to optimize subsections of the strategy (depending on what we are trying to find a solution to) that further get to optimal solutions.
So a lot of bachelors level CS is used. Although a lot of customization has been done, the benefit is it uses a lot of basic concepts, and utilized processing power rather than trying to algorithmically come up with solutions. Also, it can be continuously adaptable so it adjusts to situational changes. The strategy isn't locked, it can be reacts based on changes to frontier so to speak. If your opponent changes what they're doing, or doing something new, it can adjust itself to that.
"We don't define what winning is...", but you do!! (Score:2)
.
But usually, the "rules of the game" specifically point out exactly "what winning is". An end goal or an end-state is defined as the desireable outcome, whether it's getting to the end of the squares' sequence in the board game Life or whether it's getting a "higher scoring" hand in poker, the concept of a winning move (or equivalently, a game ending move along with a ranking system that defines who the winner is, e.g. monopoly en
Re: (Score:2)
I'm not sure you misunderstood as much as my poor explanation. Although rules of many games specify what winning is, in some cases strategy solutions don't necessarily have a clear definition of winning. Sometimes winning isn't defined as well as treating it as an optimization problem. There are rules of the game, and goals of the game.
As a simple example, take tic-tac-toe. There are rules (you can only put your marker, a X or O in a blank space) which specify what you can do. There are goals that evalu
Re: (Score:2)
Sounds interesting. And a lot more long-term and rational than much of the basically worthless "rockstar" research going on in the area. Have some publications about this?
Re:Machine learning game strategies (Score:5, Interesting)
Portions of it were influenced on a couple of works done.
Chellapilla and Fogel's 2001 work on Anaconda which built a completely evolved checkers program, which did similar techniques at the broad level. The checkers playing strategies in their case were building neural networks which regulated play. Our similarities are in the way that the strategies evolved and that no game specific knowledge was needed, other than movement rules and an aggregation of strategy fitness across competition rather than individual competition values,
Other techniques are in Kewley and Embrechts 2002 work on military strategy which was interesting in that the evolved strategies were good military strategy (with emergent doctrinal tactics) which beat military experts strategies in a simulation, in additional to beating it's own strategy when military experts modified it. This also used evolutionary concepts to evolve its solutions.
Unfortunately I can't divulge our own specific information above and beyond what I've discussed, but we certainly have been influenced by previous work on the subject, and made a few new additions to it in our own work.
Re: (Score:2)
Thanks anyways!
Re: (Score:2)
Re: (Score:2)
Thinking of branching out to simple arcade games, like Bubble Bubble or Pacman? I'd love to see a video of a computer mastering those...
Interesting you mention arcade games. As changes are made to the framework and some of the subsystems, we have a variety of benchmarks that are run that help evaluate breadth and depth of the learning process and quality of the strategies.
One of the benchmarks is a version of Asteroids. Depending on the strategy goals, it measures length of life without firing a shot (movement only while learning about spatial relationships of the asteroids), length of life based on cost of fuel (the ship is a floating pl
Re: (Score:2)
a: highest score after x levels
b: quickest way to defeat x levels
c: complete x levels with minimal thrust
d: complete x levels with maximum thrust
Re: (Score:2)
You are reminding me of Blondie24 [wikipedia.org]. Please publish, or provide a link or something. Would love to read up on your work.
I'll believe they're real games... (Score:2)
...when they require three booster packs and a prime card bought off eBay to be competitive!
Not too impressed (Score:3)
Fizbin! (Score:1)
Get a piece of the action!
I know a couple of kids that do the same. (Score:2)
The latest game they made up (on Friday) was two card draw poker (i.e. hand size is two cards). I worked through the probabilities with them to get the rank of hands correct. (It turns out to be straight flush, pair, straight, flush, high card.)
The Matrix Begins (Score:2)
So this will be the Architect. Who's working on the Oracle?
Some additional info (Score:2)
http://julian.togelius.com/Font2013Towards.pdf [togelius.com]
and
http://julian.togelius.com/Font2013A.pdf [togelius.com]
Similar evolutionary techniques have been used to generate a number of different types of game content, including Starcraft maps, Super Mario levels, rocks, dungeons, weapons... Here's an overview:
http://julian.togelius.com/Togelius2011Searchbased.pdf [togelius.com]
TFS: A Playable Card Game (Score:2)
CC.
Yavalath (Score:3)
There is at least one board game that was computer designed: Yavalath [boardgamegeek.com]. Yavalath was designed algorithmically by Cameron Browne, as described in his PhD thesis "Automatic Generation and Evaluation of Recombination Games". See his publications here:
http://www.cameronius.com/ [cameronius.com]
Re: (Score:2)
http://julian.togelius.com/Togelius2008An.pdf [togelius.com]
In general, this line of research is still in its infancy, as we are trying to figure out new ways of evaluating ga
The future is here at last (Score:2)
http://www.thewb.com/shows/the-jetsons/uniblab/d1a360bc-3f5e-478f-86b8-81e8768c823d [thewb.com]
52-pickup (Score:1)
And for some reason, the computer can always kick your ass at this game, too.