Slashdot Mirror

← Back to Stories (view on slashdot.org)

Google DeepMind's AI Beats Humans At Even More Computer Games

Posted by samzenpus on from the better-than-ever dept.

An anonymous reader writes: Google DeepMind's learning algorithm has trumped human performance in an even greater range of games from the Atari 2600. The system's performance in classic games for the 80's games console has improved steadily since it was revealed in April last year (video) and a paper released yesterday shows it besting people in 31 titles.

50 of 96 comments (clear)

  1. 80's console? by LordStormes · · Score: 3, Informative

    The Atari 2600 was released in 1977.

    1. Re:80's console? by __aaclcg7560 · · Score: 2

      But the home video game revolution didn't implode until 1983. Hence, 80's console.

    2. Re:80's console? by Minwee · · Score: 1

      The Atari 2600 was released in 1977.

      If you're going to be pedantic, at least be accurate. The Atari VCS was released in September of 1977. It was renamed the Atari 2600 in 1982 to distinguish it from the newly released Atari 5200.

      So, to answer your question, yes, that date is within the decade known as "The 80s".

  2. Color me shocked by marcle · · Score: 3, Interesting

    That a computer can beat humans at a computer game.

    The real question is, can a computer beat a human at a human game? Chess, yeah. Go, not so much.

    Hasn't reverse engineering been around for a while now? If a computer wasn't better and faster at that than a human, that would be the true surprise.

    This just in -- maybe it doesn 't require "intelligence" to win most computer games, just good memory and fast reflexes.

    1. Re:Color me shocked by penguinoid · · Score: 5, Funny

      Computers have been steadily beating humans at more and more games, "real life" ones or not. Yes, this includes Go. Ironically, humans still beat machines at things that any idiot could do, such as walking or talking or seeing. But even those things are they are getting better and better at (and we aren't), enough to beat us at various surveillance things like recognizing people or license plates.

      Humans still beat computers at Calvinball, so there's that.

      --
      Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
    2. Re:Color me shocked by dejitaru · · Score: 1

      Meh, i'm just waiting for computers vs computers. I want to see if they computationally breakdown... And hopefully more entertaining then the "game preview play" before you start playing arcade games.

    3. Re:Color me shocked by pr0fessor · · Score: 5, Insightful

      I don't want my computer to play classic atari games better than me I want it to make my work easier so I can have more time to play classic atari games, just saying.

    4. Re:Color me shocked by Anonymous Coward · · Score: 1

      Well, eventually the computer will be so good at your job that your company won't need you anymore, and at that point you will have tons of time to play classic atari games. Just saying.

    5. Re:Color me shocked by ShanghaiBill · · Score: 3, Informative

      The real question is, can a computer beat a human at a human game? Chess, yeah. Go, not so much.

      Neural nets are rapidly gaining on old fashioned hand coded algorithms. Here is a Go playing NN, that can beat Gnu Go after only a few days of learning. Progress is rapid, and computers will overtake the best humans at Go within a few years.

    6. Re:Color me shocked by mspring · · Score: 2

      ...and no money.

    7. Re:Color me shocked by pr0fessor · · Score: 1

      I don't need a super learning neural network artificial intelligence to automate 30-40% of all the jobs at my company.

      Yes the TPS reports will be in on time with the new cover sheet... I just added it to the script that makes them for me.

    8. Re:Color me shocked by vivian · · Score: 1

      You assume that in a world where robots do all the work humans who do not own robots get nothing.

      That assumption is correct, under a purely capitalistic society.
      Unless there is also a move to a more socially oriented society, there will be a lot richer few and much poorer many, when the majority of all work is done my machines.

    9. Re:Color me shocked by Dutch+Gun · · Score: 4, Funny

      Pedantically speaking, computers have been beating humans at videogames since they first appeared in arcades.

      --
      Irony: Agile development has too much intertia to be abandoned now.
    10. Re:Color me shocked by penguinoid · · Score: 1

      You assume that in a world where robots do all the work humans who do not own robots get nothing.

      This depends on whether robots replace soldiers before replacing too many of the other jobs...

      --
      Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
    11. Re:Color me shocked by Jeremi · · Score: 1

      Yes the TPS reports will be in on time with the new cover sheet... I just added it to the script that makes them for me.

      Right, but if they can get a robot to know when and how to modify the script, they can downsize you also, and improve company profits that much more.

      --


      I don't care if it's 90,000 hectares. That lake was not my doing.
    12. Re:Color me shocked by Coisiche · · Score: 1

      On Caprica *all* the humans got nuclear missiles, the Cylons didn't differentiate.

    13. Re:Color me shocked by pr0fessor · · Score: 1

      Here is the thing if every company has learning AI doing all the work and nobody has a job anymore who the heck is going to buy the companies products and services?

    14. Re:Color me shocked by Jeremi · · Score: 1

      Presumably all those nouveau-riche AIs will have cash to spend... :)

      --


      I don't care if it's 90,000 hectares. That lake was not my doing.
  3. It's all in the reflexes by BlackHawk-666 · · Score: 4, Interesting

    Computer with sub-millisecond reaction time and ability to perfectly calculate matrices, vectors and quaternions as well as predict positioning in x amount of seconds beats person. No-one should be surprised.

    --
    All those moments will be lost in time, like tears in rain.
    1. Re:It's all in the reflexes by Anonymous Coward · · Score: 2, Informative

      Deep mind is a neural network based computer. This isn't a competition to aim a laser at a brightly colored balloon. This is a competition to teach a deep neural network strategy and game mechanics. Highly abstract concepts which are not easily encoded in expert systems.

    2. Re:It's all in the reflexes by ShanghaiBill · · Score: 4, Informative

      Computer with sub-millisecond reaction time and ability to perfectly calculate matrices ...

      This is NOT about computers being able to play well. It is about computers LEARNING to play. The point of TFA, is that DeepMind was simply given the goal of "winning", and then learned on its own how to play the game and maximize the score.

    3. Re:It's all in the reflexes by dsmatthews9379 · · Score: 1

      Yeah computers can interact with themselves faster and more accurately than humans can, what a surprise, not. Oh well look on the bright side, the domestic robots of the future will be agile enough to keep up with the manoeuvres pulled by hyperactive two year olds while painlessly negotiating a floor covered with Lego blocks.

    4. Re:It's all in the reflexes by dinfinity · · Score: 3, Insightful

      Well, sortof. From TFA:
      "However, the system's continued poor performance in Ms Pacman exposes a weakness that DeepMind discussed earlier this year. The limitation stems from the DeepMind system only looking at the last four frames of gameplay, about one fifteenth of a second of the game, to learn what actions secure the best results." (my emphasis)

      GP misunderstands the ML aspect of this, but it does come down to reflexes and precision in this specific project. It is nevertheless interesting to investigate which games the net performs badly on and which ones it doesn't.

      In a way, this is also a manner of 'ranking' games: the harder it is for such a system to perform well at it, the more cerebral and less primitive/physical it probably is (although I don't want to imply that one type is better than the other)

  4. Let's play global thermonuclear war with it. by Joe_Dragon · · Score: 4, Funny

    What side do you want?

    1. USA
    2. USSR
    3. China
    4. United Kingdom
    5. France
    6. India
    7. Pakistan
    8. North Korea
    9. Israel
    10. NATO
    11. Iran

    1. Re:Let's play global thermonuclear war with it. by Anonymous Coward · · Score: 1

      12. Cylon

    2. Re:Let's play global thermonuclear war with it. by waTeim · · Score: 1

      obligatory: "Strange game, the only winning move is not to play... How about a nice game of chess?" -- at which the computer will totally own your ass! (updated for 2015).

  5. But can it win against a human players.... by mark-t · · Score: 1

    ... playing "Cards Against Humanity" (perhaps appropriately titled if it was humans vs computer)?

    --
    File under 'M' for 'Manic ranting'
    1. Re:But can it win against a human players.... by guruevi · · Score: 1

      Actually, a computer with access to Google's databases may be able to. CaH is partially luck-based, partially meme based but also based upon understanding the round interpreter's background, emotional reactions, feelings, humor etc.

      It might know better than anyone which card selection would resonate best with someone's senses as well as the groups' senses (because there is also a group pressure subject based around the group's laughter and what the group as a whole thinks is 'funny').

      Ever played with super-religious older people? I have (thanksgiving family thing) and it doesn't work very well, in that case (until we ended the game after a few rounds) the 'cleaner' answers worked best.

      --
      Custom electronics and digital signage for your business: www.evcircuits.com
    2. Re:But can it win against a human players.... by mark-t · · Score: 1

      CaH is partially luck-based, partially meme based but also based upon understanding the round interpreter's background, emotional reactions, feelings, humor etc.

      And it's that understanding that makes me think a computer wouldn't do it very well. It might be able to gauge a popularity score with the general public for a response, but it ultimately depends on the personality of each round's caller.

      Ever played with super-religious older people?

      I wouldn't want to.... it would cross a line, IMO. I can respect their values when I am around them and not deliberately do things that I *know* will offend them, even if I don't subscribe to those values myself.

      --
      File under 'M' for 'Manic ranting'
  6. Re:The dream of the 80s is alive at Google! by chipschap · · Score: 1

    What I've found interesting is that neural nets are getting better at deterministic games. It's no news that neural nets play backgammon at and beyond world championship level, but if I understand the literature correctly, neural nets are now playing Go better than programs based on just calculating move trees. I also understand that there have been inroads into chess (even though chess already is played beyond world championship level by computers).

    So no, Google has not found something at all new here, but they seem to be generalizing and extending the approach.

  7. I hate these stories by WOOFYGOOFY · · Score: 4, Interesting

    I hate these stories. Games were designed (albeit evolutionarily, through generations of culture) to exploit specific human cognitive limitations in exhaustive search and look ahead, and thereby force us to fall back on things like heuristics and strategies. This makes games unpredictable and interesting.

    But computers don't have those limitations. Of course they can out play us at games. They also add faster than we do.

    This is all IBM's DeepBlue was, a massive, massive lookahead machine which used a little human-discovered / human programmed rules of thumb to reduce the search space and then human-discovered, human programmed rules of thumb for judging the relative goodness of each move.

    The fact that computers are good at beating humans at something specifically designed to make humans perform badly is not an advancement in A.I.

    Well, OK it is, but that's not saying much.

    1. Re:I hate these stories by Moridineas · · Score: 4, Insightful

      But that's not at all the point of this article. The point of this article is that a computer program learned--in a manner SOMEWHAT analogous to human learning--through practicing how to play certain video games without having any game-specific special programming. AI opponents have existed as long as there have been video games (or close to it) and you're right, if that's what this article was about, it would be be boring. Neural net learning by examining visual output--now that's pretty cool.

    2. Re:I hate these stories by Anonymous Coward · · Score: 1

      It didn't learn how to "play" anything. It just mapped some inputs to some outputs. That's it. It can't plan, and it has no memory. It totally fails at anything needing a goal or series of steps, such as needing to get a key before opening a door in Montezuma's Revenge.

    3. Re:I hate these stories by WOOFYGOOFY · · Score: 1

      You're point is valid. WRT to neural networks, I am not impressed really. In a nutshell, I think it's a disguised way of doing statistics. An iterative, on-analytical way. With neural nets, after it's trained, no one can tell why the neural net functions as it does and no one can tell you when the neural net will do something completely insane.

      There is no analytical framework which decomposes a neural network which has arrived at THESE weights on THESE node with THIS many layers using THIS algorithm to update itself. It's just a standalone thing that works OK ... until it doesn't.

      My uni was so huge on neural nets especially my dept. If you wanted to irritate the profs,who had been made famous by their research in NN, ask them a few of these questions. If you wanted to really destroy your career whilst still an undergrad, press the point.

      Not impressed then, not impressed now. OK not impressed is not quite right; more like "underwhelmed".

      What WOULD be interesting is if there were analytical rapprochement between neural nets and statistics or at least an analytical framework whereby number of layers number of nodes connections weights, update algorithms and reward algorithms were understood well enough to support a predictive framework such that given a goal, you could write a NN without having to "train" it or given an arbitrary NN, you could analytically decompose it and understand what it was capable of.

      OK now THAT would be interesting.

    4. Re:I hate these stories by swillden · · Score: 2

      In a nutshell, I think it's a disguised way of doing statistics. An iterative, on-analytical way. With neural nets, after it's trained, no one can tell why the neural net functions as it does and no one can tell you when the neural net will do something completely insane.

      Just like training a biological brain. And yet, those seem to be somewhat useful.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    5. Re:I hate these stories by WOOFYGOOFY · · Score: 1

      OK that's the conceit of NN in a nutshell- just like a biological brain, so you said it. To me that's like saying a camera is like an eye. The brain is more than just neurons firing over synapses and reinforcing the ability to communicate across synpases. For example, nitrous oxide diffuses through the brain and is used in signalling. There are other things like that going on.

      I am not saying that I think NN are worthless. I probably came across like that; a delay reaction from years of overexposure to NN cheerleaders. It may be the start of a good way to model the actual working of the brain. I am all about Rodney Brooks' bottom up approach. The thing is, so far, it's NOT a good way to gain insight into the brain. It's all going the other way- NN models built on Hebbian learning and all that.

      I get sick of people pushing AI as "on the verge of an incredible breakthrough" or worse , doing what IBM did with Deep Jeopardy or whatever they called it, Deep Thought . That whole thing was a total dog and pony show that potentially misled the general public, potential investors and government funding authorities about what they had achieved (virtually nothing significant) and where it might lead (nowhere).

      AI has a very very long and ignoble history of overhawking its wares, dating back to the 60s then the 80s then the 90s.. oh fuck it, every 10 or 15 years..

      In '68 HAL9000 was completely plausible by 2001 in the eyes of the AI establishment.

      Great things have come out of basic AI research. Expert systems. Defense systems. But the idea that we're anywhere close to "the Singularity" which a lot of naive people believe, anywhere close at all is just not true.

      So Google is big on AI. I give you 100 to 1 that their self driving cars FAIL as a general mode of transportation by the year 2050.

    6. Re:I hate these stories by WOOFYGOOFY · · Score: 1

      We all benefit from 80s-style expert systems every day. Even your GPS is a form of expert system.

      The idea we'll soon be of replacing humans, which is causing a sort of mini moral panic amongst some engaged and intelligent but non-expert part of the population is totally off course.

      Just focusing in on driving applications, lane awareness and accident avoidance are two great uses of AI but the thing is, they serve only to take minor, stereo-typical (get back in your lane! Break right now as hard as you can!) actions in response to common re-accident scenarios in which there is only one-best-thing-to-do (so do it now!) It is assumed that the driver will take over to address the possibly more complex real-time situation.

      AI is not going to parse out every possible set of real world events which a driver may encounter. Google is finding that out now. The situations are too varied, too unpredictable (the technical term for this is "fucked up") and engage too many independent actors whose reactions are unknowable but critical.

      Sure , it can be used on highways with a human driver, and warehouses (maybe) and golf courses and it can , you know replace the guy who drivers the smoother over your local ice skating rink, OK now HE'S got a job to defend against AI, but when you mix in the liability issues there is about zero chance Google cars are going to be how people get around.

      When Palm finally quit trying to recognize all possible handwriting from everyone and started instead to make people write using their stylized alphabet, then they became successful. So also here. We'll all drvie Google cars when and if Google convinces the legislature to re-engineer the roads, the rules of the road, and the liability laws to support their technology or, to be more precise, the limitations of their technology.

      Society moves forward and changes so don't say THAT won't happen. It might. It might not. It's a lot of money, but then it would save a lot of lives, save a lot of money and reduce crime a lot also. Maybe THIS is their REAL end game.

      At any rate, something like the above changes is going to happen sooner than a Google car is going to replace your car, with everything otherwise staying the same.

    7. Re:I hate these stories by swillden · · Score: 1

      OK that's the conceit of NN in a nutshell- just like a biological brain, so you said it. To me that's like saying a camera is like an eye. The brain is more than just neurons firing over synapses and reinforcing the ability to communicate across synpases. For example, nitrous oxide diffuses through the brain and is used in signalling. There are other things like that going on.

      Meh. So there are some additional interconnections. Are those actually essential? It seems unlikely to me, but they could certainly be added if they are.

      It may be the start of a good way to model the actual working of the brain.

      Irrelevant. Oh, I suppose it may someday be relevant to neuroscientists whose goal is to understand the brain rather than to create useful systems. But for the people interested in being able to create automated systems that can be taught to make complex decisions effectively, what really matters is that it seems to work very well. Sure, the fact that we don't understand how they work means they may occasionally do insane things, but that's also true of complex decision-making algorithms we do understand (or think we do). And it's also true of living brains.

      AI has a very very long and ignoble history of overhawking its wares, dating back to the 60s then the 80s then the 90s.. oh fuck it, every 10 or 15 years.

      You say that as though everyone here isn't fully aware of it. But it's obvious, and it's common. It's in no way particular to AI. People are tremendously bad in general at predicting future technology for the -- rather obvious, actually -- simple reason that future technology will be built based on knowledge that we don't yet possess. You can't accurately predict the results of applying knowledge you don't yet have.

      the idea that we're anywhere close to "the Singularity" which a lot of naive people believe, anywhere close at all is just not true.

      You're wrong. And so are they.

      The truth is that we have no idea how close we are to that point, and really won't have any idea until we're there, or until we prove that it's impossible.

      However, none of that has anything to do with the topic at hand. Neural networks (biological or electronic) are almost certainly not the only way to construct the information flows underlying intelligence. It's also perfectly possible that our current NN models are inadequate for producing intelligence. So what? There are huge numbers of tasks for which we don't need general intelligence. All we need is a good automated decisionmaker which makes the right decisions and doesn't cost too much to build.

      That is where NNs are awesome. How many engineer-weeks of effort would it take to produce an algorithm that, fed only the raw pixel data from the screen, can play a video game effectively? With Google's DeepMind NN, it takes one week of computer time, with little or no human involvement at all.

      Neural networks may or may not be useful in reaching toward general AI. But they absolutely are useful tools, enabling us to build useful automated systems now.

      I give you 100 to 1 that their self driving cars FAIL as a general mode of transportation by the year 2050.

      I'll take that action. How much will you put on it? Let's define the terms and work out the logistics. Also, I'd be fine with pulling in that year by a couple of decades. I'll bet that self-driving cars on Google's model (full self-driving, highways and in town, with vanishingly few situations the car can't handle) will SUCCEED by 2030.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    8. Re:I hate these stories by swillden · · Score: 1

      AI is not going to parse out every possible set of real world events which a driver may encounter. Google is finding that out now. The situations are too varied, too unpredictable (the technical term for this is "fucked up") and engage too many independent actors whose reactions are unknowable but critical.

      Actually, I know some guys on the self-driving car team, and Google cars already handle just about everything safely. What they're focused on now isn't so much handling strange situations, but optimizing the car so it behaves like a human driver, to avoid confusing other human drivers -- or being taken advantage of by them.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
  8. Re:Rocket League. by galabar · · Score: 2

    Only if it learned on its own.

  9. The only game in town is the mating game. by dsmatthews9379 · · Score: 1

    I wonder, if they feed thousands of romance novels into it, will it learn what women want better than the average geek can?

  10. Centipede by tomxor · · Score: 1

    At least i can still beat it at my favourite :P

  11. How could they have missed E.T.? by turrican · · Score: 1

    So, they haven't yet exposed it to E.T. The Extraterrestrial, or they already have, and DeepMind refused to continue playing?

  12. Grand Theft Auto? by Irate+Engineer · · Score: 1

    Call me when we catch DeepMind in GTA shagging a hooker in a Bugatti.

    --

    Left MS Windows for Linux Mint and never looked back!

    Vote for Bernie in 2016!

  13. So I'm beat by a bot by Snotnose · · Score: 1

    You're telling me PANDASGOCOMMANDO is a Google bot, that's why I can't beat the guy in CoD?

    / That's a real Modern Warfare 3 name
    // Beats me handily every stinkin time
    /// Doesn't look like he's cheating either :(

  14. Have Patience... by Frosty+Piss · · Score: 1

    I want it to make my work easier so I can have more time to play classic atari games

    One step at a time. This is just the beginning of "real" computer AI iRobot ( or Robot & Frank ) style. Sure, this seems a trivial application, who needs it. But you have to start someplace, and game decision making is a good place for many reasons.

    --
    If you want news from today, you have to come back tomorrow.
  15. huh? by cascadingstylesheet · · Score: 1

    Computers have been beating me at computer games since before you were born, sonny ...

  16. Re:God damnit, make me a sammich by jewens · · Score: 1

    Only if you say the magic word. For any of you who just heard a strange wooshing sound, its "sudo".

    --
    That group of bovine standing over there appears quite portentous. That's right it's an ominous cow herd.
  17. Self-important nigilists by Katatsumuri · · Score: 1

    Looks like 90% of commenters in this thread are too proud of their superior human brains to even try and get the point of the experiment. Researchers made a computer which can learn to achieve goals with no instructions, and you mix it up with custom game AI or bitch about how it is not fair to compare scores with biologically limited humans. This is just depressing.

    Go DeepMind!

    1. Re:Self-important nigilists by babymac · · Score: 1

      Agreed. Slashdot is one of the worst places to go for commentary on AI developments. Is it because of an aging audience?

      --
      "War makes me sad." - Me