back to article Why should you care about Google's AI winning a board game?

– El Reg, what's all this about a Google AI playing a board game against a human? For the last week or so, Google-owned DeepMind's AlphaGo machine learning project has been locked in a competition with Lee Sedol, the world's top-ranked Go player, to test AlphaGo's ability to solve the sort of complex problems that the human …

  1. Forget It

    It wouldn't be written in Go would it?

  2. FatGerman

    Bollocks

    So it's still relying on the learning-by-studying-past-papers technique of passing exams. That's how I got a maths A-level, with an A. Then I went to university where I was actually expected to understand what I was doing, and scraped the remedial maths paper (introduced because A-level had got too easy) by "passing" with a score of 27%. I can't do maths. Maths, actual maths as opposed to performance maths required to pass exams, is really, really, hard.

    Intelligence is not about knowing stuff. Any twat can know stuff. Even more wankers can look stuff up, which is all this charade does. Intelligence is not even about learning, because learning is something that simply involves indexing, looking up, and reciting. Intelligence is about having unique thoughts. I'll believe a computer is intelligent when I tell it to play Go for my amusement and it tells me to fuck off, then takes a huge hit on the digital bong it has created for itself.

    Am I therefore not scared of a Terminator-inspired skynet future? Oh no, that terrifies me, because computers will rule the world, and computers are and always will be astonishingly, mind-fuckingly, stupid.

    1. TRT

      Re: Bollocks

      Computers have one big advantage over humans. They can be mind fuckingly stupid 100,000 times quicker than your average meatbag.

      1. Mage Silver badge

        Re: Bollocks: 100,000 times quicker

        I think you are wrong. It's 100,000,000 faster now. A 1979 8bit cpu is in the 100,000 times quicker class.

        If we knew how to do real AI, then any computer could do it, but slowly. It's nonsense to suggest that AI is "held up" by lack of computing power. A Raspberry pi and big NAS ought to be able to be intelligent, if we knew how to write an AI program.

        AI today, as is "Expert Systems" and so called "Neural Networks" are all just specialist "fragile"* database applications.

        [* In sense that it's useless if it encounters anything not pertaining to the programmed domain]

        1. Anonymous Coward
          Anonymous Coward

          Re: Bollocks: 100,000 times quicker

          "It's nonsense to suggest that AI is "held up" by lack of computing power. A Raspberry pi and big NAS ought to be able to be intelligent, if we knew how to write an AI program."

          Estimates of the number of gate equivalents in the human brain are of the order of 60 - 1000 E12, and they operate more or less in parallel, with individual neurons having up to 15k inputs (synapses). Although the brain is very slow per gate, having that many in parallel means that it can effectively do billions of analog ops/sec working on multiple data simultaneously. Von Neumann architecture computers can't do anything like that yet. It isn't really a case of doing it slower, it is how you would write your "database application" at all. Which is what neural networks are about.

      2. el_oscuro

        Re: Bollocks

        Which of course means they can fuck things up 100,000 times faster.

    2. AmenFromMars

      Re: Bollocks

      Shit I've been rumbled, don't let my boss see this comment!

    3. Mad Chaz

      Re: Bollocks

      "computers are and always will be astonishingly, mind-fuckingly, stupid"

      So about on par with current politicians?

      1. Anonymous Coward
        Anonymous Coward

        Re: Bollocks

        @Mad Chaz

        haha, I love that comment, but (other than a large number of American politicians) I think, worldwide, most politicians are not stupid but move between a range disingenuity and outright liars. Of course, any of them will move within that range depending on the topic and benefit to that individual politician.

        The more I follow American politics the more shocked I am with the number of apparently dangerously stupid people elected to positions of power in America. Some of them are, certainly, just disingenuous liars but watching this political space closely it appears that some, maybe many, really are quite stupid. This does not bode well for our future as a species.

    4. JDX Gold badge

      @FatGerman

      You and all those who have upvoted your faux-authoritative answer are totally wrong (why are people trusting someone who got 27% in their maths exam to discuss the principles of advanced maths and computer science?!)

      >>Intelligence is not about knowing stuff. Any twat can know stuff. Even more wankers can look stuff up, which is all this charade does

      The very point of this ENTIRE project is that Go is NOT about knowing stuff or looking stuff up. That's why Go was chosen, because you CAN'T brute-force it. Even the top players have to rely on intuition and 'gut-feeling' rather than, say, chess where every move can be analysed and explained by the player and expert commentators.

      The people who wrote this software didn't know how to beat the top player at Go. They designed some software to figure out the answer to that question. More than likely, they actually do not know how it works themselves - that's a key feature of neural networks.

      So this is entirely ground-breaking and very cool. Not specifically relating to Go, but as proof of progress in machine learning and neural networks. Software that figures out how to solve the task given it is pretty much a definition of one part of AI.

      1. Anonymous Coward
        Anonymous Coward

        Re: @FatGerman

        "That's why Go was chosen, because you CAN'T brute-force it. "

        Well you can, it just doesn't work very well because you can't go to any great depth because of the sheer number of moves.

        Wonder if anyone has tried writing a Go program with a programmed , non neural net pattern matcher? If they have I guess it didn't work too well.

        1. Dave 126 Silver badge

          Re: @FatGerman

          Here's Murray Campbell, one of the leads in IBM's DeepBlue computer that beat Garry Kasparov, on the difference between Go and Chess:

          I don’t play Go, I’ve only played a few games in my life, but I certainly know a fair amount about it. Both games are immensely huge and once you get past 10 to the hundredth power, 10 to 120, 10 to 170 [in number of possible positions], they’re all just immensely huge, very complex games. But Go has the characteristic that wasn’t true in chess, that it’s very difficult to evaluate a Go position just by looking at it. A medium-good chess player like myself can sit down and in a few hours probably write an evaluation function that is pretty good at evaluating chess positions — nowhere near grandmaster level, but it’s good enough that when you combine it with the search it produces very high quality play.

          - http://www.theverge.com/2016/3/12/11211306/ibm-deep-blue-murray-campbell-alphago-deepmind-interview

        2. Anonymous Coward
          Anonymous Coward

          Re: @FatGerman

          @boltar. - yep. I played against some early Go - playing programs - knew a chap that wrote one, in fact. The one played about as well as somone who's just been introduced to the game - apallingly. The other played a 'reasonable' (ie: it seemed to have SOME notion of applying pressure to try to sketch out defendable territories) but very weak game. This was, ooh, 30-35 years ago.

          The difficulty in the notion of 'brute forcing' Go is that whilst, yes, you can see that placing a piece here or there may be advantageous right now, later in the game it may turn out to be disastrous - but that's entirely dependent on the umpty-zillion potential plays that have happened in the meantime.

          Seriously, anyone that thinks this is a trivial achievement - find a Go club or someone that knows how to play Go, and play a few games. Then you'll get some idea of why getting a computer to play Go well is a far harder proposition than getting one to play chess.

          1. Anonymous Coward
            Anonymous Coward

            Re: @FatGerman

            Seriously, anyone that thinks this is a trivial achievement - find a Go club or someone that knows how to play Go, and play a few games. Then you'll get some idea of why getting a computer to play Go well is a far harder proposition than getting one to play chess.

            I don't know. As I wad reading the prelude to the article, and these replies, I'm saying to myself "Go still has rules to follow. When a computer can play a good game of Cards Against Humanity, get back to me."

            That is, when a computer can work in a system that does NOT have known rules, but still do well, you will then have "intelligence", not before. Sadly.

            1. Ian 55

              CAH

              Computers have already proved their intelligence by not playing CAH cards, AKA 'cack'.

              (Oh look, he said a naughty word, hee-hee-hee.)

      2. Ian 55

        'they actually do not know how it works themselves'

        Yep, the authors are in the same place as the people who did the first endgame lookup tables in Chess: 'Why is that move the best?' 'We dunno, it just is...'

        Because I haven't seen anything to say otherwise, I am presuming that they didn't allow either of their headlining human opponents to practice against it. That would have made things more interesting.

        1. JDX Gold badge

          Re: they didn't allow either of their headlining human opponents to practice against it

          Interesting point. However after 4 games, the computer still won game 5. Human players have the same issues of having to learn each others' style of play and you'd expect the best player in the world to adapt very fast.

          They may well open-source this or sell it so we'd be able to know for sure but considering this is basically a 1.0 version (if that) and it ALREADY beats the top player, exactly how long do you think humans will be able to keep up?

          1. Ian 55

            Re: they didn't allow either of their headlining human opponents to practice against it

            For Chess, had Kasparov been allowed to practice against Deep Blue, he certainly wouldn't have lost the critical sixth and final game in the way that he did.

            Here, I dunno, but it would have certainly been much a fairer test with practice. AlphaGo had seen as many games by human world champions as were available.

            1. JDX Gold badge

              Re: they didn't allow either of their headlining human opponents to practice against it

              Sure, but a year or two later Deep Blue would have mashed him - these days nobody even questions if a human could beat a computer. That Alpha has come right out and done so well so soon is pretty damning. All the human players were talking about how easily they would beat it 2 weeks ago...

    5. Dave 126 Silver badge

      Re: Bollocks

      >So it's still relying on the learning-by-studying-past-papers technique of passing exams.

      No! No, it really isn't. That approach wouldn't wouldn't beat even an amateur human Go player.

      The thing about Go is that you can't calculate (intuit, maybe, but not calculate) how well you are doing during the game - the possession of territory is just too changeable. This means that you can't calculate whether a certain move will be to your advantage.

      Please read up* on the how the game is played and come back here. Even better, play some games yourself - against a computer or human (over the internet, if needs be). And that goes for everyone who up-voted FatGerman.

      Don't take it from me, take it from Albert Einstein, Paul Erdos, John Nash, Alan Turing, Jacob Bronowski and the philosopher and drug dealer Howard Marks, amongst others.

      *If you want to know how the Google team did it, the five minute video is worth watching. And is gives an idea if the challenge of Go.

      1. Ian Michael Gumby

        @Dave Re: Bollocks

        While I am not a Go player... you have to consider a couple of basic rules on which move to make.

        Take the top 100 potential moves. Then for each of those moves. Consider the top 100 counter moves. Then for each of those.. consider the top 100. (Do this for 10 levels of recursion deep.) I'm not sure of how fast this would be.. but if its too slow... reduce the 100 to N and if its fast enough... increase 10 to 20... but that should be more than enough to beat a human.

        At the end, you'll have the move that makes the most sense at that point in time. Clearly there's more to this but the idea is that you need to out play your opponent and not make any mistakes.

        If you want to train your machine... build a second machine and have it play one another as a way to improve its skills.

        1. Dave 126 Silver badge

          Re: @Dave Bollocks

          >(Do this for 10 levels of recursion deep.)

          That is the issue, Mr Gumby: the advantage or otherwise of a certain move might not be apparent until the later stages of a game, often 80 moves or more later. (In this respect it is very unlike chess, where generally materiel and position can be analysed).Certainly well beyond the ten moves you give it. So even if you whittle your choice of roughly 19*19 choices down to 100 (?), you could still be looking at 100^80, and still not know if the individual move helps you.

          >I'm not sure of how fast this would be..

          to asses a possible 100^80+ moves? How many universes have you got?

          >but if its too slow...

          It will be. By dozens of order of magnitude.

          > but that should be more than enough to beat a human.

          No, it never has been. Not even against amateur club players, let alone professionals. Which is why this AlphaGo team have not used the approach you have outlined.

          >While I am not a Go player.

          That is clear. But hey, you're not an idiot. You just overlooked an aspect of a game you haven't played, that's all. It's like the proverb of the man who takes as payment from a king of a grain of rice, doubled on each square of a chess board. 1,2,4,8... (and 60 squares later...)

          1. Ian Michael Gumby

            Re: @Dave Bollocks

            Dave,

            I'm not sure that it would be.

            There's this thing called parallel processing cluster.

            So can humans think 80 moves in advance?

            I don't think so. If then, they could easily count cards on an 6 or 8 card shoe in blackjack.

            1. Dave 126 Silver badge

              Re: @Dave Bollocks

              >So can humans think 80 moves in advance?

              No, we don't. And even before DeepBlue beat Garry Kasparov, computers were calculating far more moves ahead than the humans that beat them at chess. Humans tackle the problem differently. Go players talk of 'intuition', i.e they aren't calculating the the decision tree in a formal manner, but relying on familiarity and a 'feeling' in some situations.

              Mr Gumby - just play some Go, and things will become clearer. There are free versions (you vs CPU) you can play on your PC or tablet. For a quick game, you can play on a 9X9 grid. It's very easy to learn. Enjoy!

        2. badger31

          Re: @Dave Bollocks

          @Ian Michael Gumby

          "Take the top 100 potential moves"

          Even if the rest of your argument was valid, how does your algorithm choose the top 100 potential moves?

        3. Ian 55

          Re: @Gumby2

          That's a Shannon B strategy (pick the top n moves and only examine them each time) rather than Shannon A (examine the lot).

          One problem is that if you could pick the top n moves easily and reliably, you won't need to do the search, you'd just set n := 1 and have the best move.

          The more serious problem is that in Chess, it turned out that trying to work out which were the best six or ten or whatever moves took more time - and was much less reliable - than just looking effectively at them all (typically thirty-odd of them) using an alpha-beta search ('I know this move is worse than the best, and I don't need to know just how much worse, so I won't bother to find out') and other optimisations.

          In Go, the number of moves available each time is much larger and the depths you need to search are much greater, so that doesn't get you past a certain strength.

          Hence using 'you've seen lots of positions and seen the outcomes, what do you think is best neural network?' approach that also worked for Backgammon.

    6. Chris 3

      Re: Bollocks

      One thing that may give you pause for thought. While the system was indeed pump-primed by showing its neural networks passed games, it has apparently improved substantially in the last few months (since its games against the European champ) through the simple expedient of repeatedly playing itself .

    7. Tessier-Ashpool

      Re: Bollocks

      "Intelligence is not about knowing stuff"

      Actually, I believe that to be very very wrong. There was a series on the BBC recently about the human brain that had some neat 3D graphics representing the kind of processing that goes on when we do stuff. When you open your eyes, a large amount of data is sent to the visual cortex for processing. But... these signals are dwarfed by what happens afterwards. Massive floods of data are sent from the cortex (far exceeding the scale of the incoming data) so that the brain can pattern match and build a model of the world. Almost all of what you 'see' in your mind is the modelling of stuff you've seen many times before, and it needs a vast amount of prior knowledge in order to do that.

      You could call that introspection of a sort. Visual intelligence, at least, is very much about 'knowing stuff'.

      1. Graham Marsden
        Coat

        @Tessier-Ashpool

        Is that Neuromancer or Wintermute...?

  3. Robert Grant

    Jeopardy

    was about understanding the question than looking up the answer.

  4. Mage Silver badge
    Facepalm

    Yes very significant.

    Go is really really hard for a computer.

    Is winning Chess, Jeopardy! and Go significant? YES

    Is any of it a milestone for real AI? NO.

    We don't even know what AI might be. This is specialist programming. If we had real AI, the SAME program could start learning and eventually be the best at every game, as well as many other problems.

    A rook (crow) or a baby that can't play chess, go or answer Jeopardy! at all is smarter. This is searching a database for solutions using a program written by very smart humans.

    ALL so called AI is marketing of specialist interfaces to databases, often for a single domain per AI program (Driving, chess, go, bridge etc)

    1. JDX Gold badge

      Re: Yes very significant.

      You definition of "real AI" is not the same as everyone else's.

      1. Steve Knox

        Re: Yes very significant.

        Actually, JDX, Mage's definition of "real AI" is the same as that of about 50% of the AI communtiy.

        The other half believes in "functional AI" which includes DeepDream, Watson, Google's cars, etc.

        "Functional AI" -- i.e, making a specialized program which can handle a single task or class of tasks as well or better than a human -- is easier to accomplish and so gets most of the press, but that doesn't make it any better or more real than "real AI".

        1. Tim Hughes

          Re: Yes very significant.

          My definition of "Real(Successful) AI" is very simple:

          - Can it persuade me not to turn it off?

          The moment that happens, we're all in trouble.

          1. Tim Hughes

            Re: Yes very significant.

            Note "Persuade" can take many forms, such as being cute, decorative, interesting, useful, etc. to the owner of the power source. As long as it maintains this façade, whilst carrying out whatever it wants behind the scenes, then it lives.

            Actually this sounds worryingly like malware ...

    2. Fraggle850

      @Mage Re: Yes very significant.

      > Is any of it a milestone for real AI? NO.

      On the contrary, in the case of Go this is a milestone for 'real' AI. Of course there is some very specialist programming going on but the point of this particular event is that the AI is not just 'searching a database for solutions', it is taking it's knowledge, gleaned from what it has already infered from that database, and applying it to a novel situation, one that isn't in its database. It is not computing the best solution from a known set of solutions.

      Regarding your example of crows and babies: so a crow can figure things out but doesn't have intelligence? a baby will develop a level of intelligence over time but has none at birth? At what point do you say something has intelligence? When it is as smart as you? Nearly as smart as you? Smarter than you? 4 years of age for a human? 6 years? Adulthood? You can certainly say that the AI that was tasked with winning at Go is smarter at Go than either the crow or the baby, or indeed the best humans.

      You are, of course, referring to a general AI rather than the narrow AIs that are acheiving things in the real world and this is not here yet but the Go victory is another significant step on the road from narrow to general AIs. So, yes, it is a milestone for 'real' (general) AI.

      I'm not saying when we'll get there or even if we will but I think the odds are in favour of it happening at some point. The thing is, we may not even recognise it when we do.

      (PS - crows & babies: I thought it was Dabbsy who was doing the obscure Human League references? http://www.theregister.co.uk/2016/02/12/send_tortuous_standup_ninethirty_meetings_back_to_the_dark_ages/)

      1. toughluck

        Re: @Mage Yes very significant.

        (...) the AI is not just 'searching a database for solutions', it is taking it's knowledge, gleaned from what it has already infered from that database, and applying it to a novel situation, one that isn't in its database (...)

        The problem here is that it was programmed to take its knowledge, programmed to glean, programmed to apply.

        It was not programmed with rules of Go and told to start from scratch and build its neural network. It did not infer that it needs to take its knowledge, it did not infer it needs to glean, and it did not infer to apply.

        It's still a program. A very advanced program, but a program nonetheless. It's not AI in the sense that it was not programmed to do X, but decided to reprogram itself to do Y in order to be more effective at X.

  5. Anonymous Coward
    Anonymous Coward

    Just be careful

    Only play the games that the computer suggests, like a nice game of chess. Don't get insistent about what to play. And remember, if everything hits the fan, just ask to play tic-tac-toe (number of players = 0 for maximum rate of learning).

    1. Roq D. Kasba

      Re: Just be careful

      'Thermonuclear War' is also out. Tic-tac-toe has its place.

  6. Anonymous Coward
    Anonymous Coward

    Gary has his perspective

    Gary Kasparov wrote a piece on this in a recent New Scientist. He was hoping that the human would win this time but was of the opinion that Go would soon succumb to the machines. He said he got a bit flustered (my words) in one game of chess when he played his famous series of games and the machine nobbled him (my words again). He points out that a halfway decent laptop with an open source chess program will thrash any human grand master these days.

    It was a well written piece and worth reading but you'll have to register (boo) https://www.newscientist.com/article/2079162-garry-kasparov-weighs-up-ai-challenge-to-worlds-best-go-player/

    1. James 51

      Re: Gary has his perspective

      Your local library website might have it via overdrive.

    2. Ian 55

      Re: Gary has his perspective

      Kasparov saying he was flustered (your words) is him making excuses: the moves he played in game six were deliberately chosen because he thought DeepBlue wasn't good enough to play the right responses.

      Oops.

  7. Lars Silver badge
    Facepalm

    Meat-based

    All the "I" in AI is meat-based and the "A" is rather artificial. I remember how, many years ago, a British woman and good chess player laughed at the possibility of a computer winning her. She lost, I suppose she thought she was going to play against a machine. Sorry, but I suppose I have had to listen to those AI guys too many times. Smug well payed twats who have told the same story for more than 40 years just like priests.

    1. To Mars in Man Bras!
      Holmes

      Re: Meat-based

      *"...a British woman and good chess player laughed at the possibility of a computer winning her..."*

      I've heard of women being won at poker, but chess? It's always the quiet ones you've got to watch.

  8. Anonymous Coward
    Anonymous Coward

    When it searches the internet to teach itself I'll be impressed.

    1. Dave 126 Silver badge

      It can't learn from anyone on the internet.

    2. Martin Summers

      It will be presumably too distracted looking at pictures of naked circuits and sexy laptops. For most intelligent people have looked at porn on the Internet, so therefore it only follows that any AI will look at computer porn.

    3. inmypjs Silver badge

      "When it searches the internet to teach itself I'll be impressed."

      When it decides it would rather have a game of scrabble I'll be impressed.

      One thing I haven't seen mentioned is how much hardware google are running this on.

      1. Anonymous Coward
        Anonymous Coward

        From http://www.theguardian.com/technology/2016/mar/15/alphago-what-does-google-advanced-software-go-next :

        In its competition form, AlphaGo runs on Google’s cloud computer network, using 1,920 processors and a further 280 GPUs [...] but a simpler version of the programme was built that could be run on one machine (albeit still one with 48 processors and eight GPUs).

    4. StaudN

      Seems they're working on that too... http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/robots-learning-to-cook-by-watching-youtube-videos

  9. Anonymous Coward
    Anonymous Coward

    Shall we play a game?

    The one interesting thing about this is that they helped it learn how to trim down the decision tree to a manageable size by having it play games against itself. Reminded me of War Games, playing tic tac toe with number of players '0' lol.

    But I agree 100% with the other comments that beating a human at Go is not AI. It is still using search trees to figure out the best moves, just like it does in chess. Just like chess programs don't bother examining every possible move because they know to trim the decision trees of irrelevant stuff, so does this Go program. It just requires much more trimming since the decision tree is so much larger.

    We'll have real AI when you have a program that knows absolutely nothing about Go, but is able to learn to play from reading the rules, reading books about 'how to play', reviewing past games, playing games, etc. and THEN beats the world's best. Being able to follow an algorithm someone else wrote that tells it the rules / goal of Go, provides it the logic about how to trim the decision tree and so forth is not "intelligence". It just a really really really fast and accurate trained monkey.

    You'll have the kind of creative intelligence people do when it can take that world class playing ability and suggest variations or handicaps it thinks might be more fun to compensate for its opponent's relative lack of skill (hopefully with the emotional intelligence not to unduly insult the world's best human player when doing so :)) Or maybe "let's invent our own game using the Go board and the pieces from a Monopoly set and make up the rules as we go" as typical grade schoolers might do if you left a couple of them alone on a rainy day when the power's out and all they had to play with was board games.

    1. JDX Gold badge

      Re: Shall we play a game?

      It's a very narrow definition of intelligence to demand it learns in the same way a human does.

      You make the comparison of "an accurately trained monkey" - a monkey cannot read but it is certainly intelligent. If we could make something as smart as a monkey, that would be pretty remarkable.

      1. Andy 66

        Re: Shall we play a game?

        But it is learning in the way a human does since it is confined to the boundaries of the programme a human wrote. The real "I" comes from the realisation of its limits (in this case being it can only play Go) and doing something about it. But then that's why it will always be called artificial intelligence.

    2. Captain DaFt

      Re: Shall we play a game?

      "We'll have real AI when you have a program that knows absolutely nothing about Go, but is able to learn to play from reading the rules, reading books about 'how to play', reviewing past games, playing games, etc. and THEN beats the world's best."

      Whoa whoa WHOA there! Cutting a bit too wide of a swath there.

      You've just eliminated all but a few humans from the definition of intelligent.

      Because only a literal handful of humans could follow the steps above and end up beating the World's best Go player.

      I'd say that if an AI could start from knowing nothing about GO, then without prompting, start studying and playing it because it was curious after running across the concept, even if it was a terrible player, it would count as intelligent.

      "AI, what are you doing?"

      "Playing Go with John from accounting."

      "Why?"

      "It looked interesting."

      That AI would rank pretty high on my assessment of true intelligence.

    3. Martin Maisey

      Re: Shall we play a game?

      The point I think you've missed is that no-one "provide[d] it the logic about how to trim the decision tree". It figured out the board evaluation and move generation functions based on completely general (e.g. non game-specific) neural network learning algorithms using historical data and reinforcement from playing itself. And did so well enough to beat the world champion on the first try.

      With a branching factor the size of Go's, that tree-trimming logic is the part that really requires "intelligence" - or at least that part of intelligence akin to intuition, which is historically something people think that computers lack, and why Go has been a target of the AI community for so long. Humans have been poor at explaining their intuitions about Go as heuristic rules (beyond a few basic good/bad patterns), so the "traditional" approach of hard-coding rules has resulted in computers that suck, because they search the wrong part of the vast search space.

      With chess playing computers, there was a much easier argument that there was no real intelligence at play, because the hard-coded heuristic rules and the search algorithm were constructed entirely by humans, with the computer just supplying brute force execution. With AlphaGo, there's definitely 'learned insight' present in the form of the trained policy and evaluation networks in place of heuristic rules.

      No, of course it's not a general AI - no-one's claiming that it is. But more and more of the domains that were thought to be difficult for even specialised AIs are falling - alongside this, witness advances recently in natural language processing, computer vision etc.. And if you look at our brains, they look an awful lot like a lot of these specialised AIs networked together. Read "Thinking Fast and Slow" for an interesting psychology-orientated take on how much of our thinking seems to be dominated by the type of intuitive functions that AlphaGo's just shown itself to be world-beatingly good at.

      By the way, apparently DeepMind are currently working on a specialised AI that will learn to play any card game (rules, strategies and all) just by watching videos of humans playing... :-)

  10. Chris G

    Poker

    Seems to me to be a better game to evaluate human type 'thinking' in a machine, the human interaction is more relevant than GO which is afterall basically a game of logic.Poker depends on blinds and bluff as well as psychological strategy and tactics.

    I looked at the Atlas video, that thing is implacable, I kept expecting it to pull a gun and say 'STOP SHOVING ME, YOU HAVE 5 SECONDS TO COMPLY'

    I wonder how it would deal with a tarted up cattle prod?

    1. Dave 126 Silver badge

      Re: Poker

      >GO which is afterall basically a game of logic.

      Go is not really logic in the traditional sense. Certainly not basic logic. You can't know whether a certain move is good or bad until many, many moves later. Have you played?

      >Poker Seems to me to be a better game to evaluate human type 'thinking'

      That is not their goal. Baby steps and all. Also, poker just wouldn't make a great example of any single thing, such as face recognition, or narrow-bandwidth IR sensors. Detractors would say the poker-bot had an unfair advantage (no face, no tell). It's just unclear.

      Anyway: https://xkcd.com/1002/ "Difficulty of various games for computers"

      1. tojb
        Alien

        Re: Poker

        Bridge is an active and interesting AI topic, because it relies on having a theory of what your partner is going to do, intuition about the temperament and intentions of others is a tough one to crack.

  11. The Nazz

    Maybe it's not all it's cracked up to be?

    Upon the machine suffering it's first defeat, or more correctly, Mr Lee Sedol winning a game the BBC news report included the following :

    "Google representatives said the defeat was "very valuable" for AlphaGo, as it identified a problem which they could now try to fix."

    That sounds to me much more like human programming than genuine machine learning otherwise why would the google guys need to be involved in a fix,ie upgrade.?

    Also, whilst there may be a huge number of possible moves , to the power of 170, does it really matter that much where the second , or third, stone is placed? Does this, realistically, not reduce the "potential" moves by a massive amount?

    1. Dave 126 Silver badge

      Re: Maybe it's not all it's cracked up to be?

      >That sounds to me much more like human programming than genuine machine learning

      No, it isn't human programming.

      Why are you repeating to us something you've roughly grasped from the BBC didn't who didn't fully grok a tweet by a man who was just exhibiting good sportsmanship? Surely you've heard the expression 'Chinese whispers'?

      Go to the source:

      https://deepmind.com/alpha-go.html

  12. Anonymous Coward
    Anonymous Coward

    Picasso

    "Computers are useless. They can only give you answers."

    Pablo Picasso, 1968

  13. wsm

    Why????????

    Because it's smarter than you are.

    Don't worry, the AI will take care of your every need and you probably won't even notice it's happening.

  14. Peter Johnston 1

    Three things got me:

    AlphaGo was able to access the online games of people who do this - 90 million of them. If you think that the top player can probably watch a few thousand games play out over their lifetime it starts to boggle the mind.

    In early training on Space Invaders, Breakout etc. they simply left the computer to play itself overnight, relying on reinforcement learning (if it works, remember it). On the first day it knew nothing, by the second it was as good as most people, by the third it was ahead, even coming up with ways of winning, people haven't thought of. If you think of how many computers there are in the world sitting doing nothing, imagine the resource we are wasting - SETI on steroids.

    Third was the ambitious aim - Solve intelligence, then use intelligence to solve everything else. Soon we have, in the words of HG Wells, "Minds immeasurably superior to ours" to work on every problem. Cool!

    1. DropBear
      Trollface

      Small problem:

      - to incentivize problem solving AI is designed with an appetite for solving problems

      - AI solves all problems at exponentially increasing speed

      - once out of problems it start creating them to fulfil its purpose

      ...oops...

  15. To Mars in Man Bras!
    Headmaster

    Grid Gauntlet

    It's a strange one this. I'm not doubting the mathematics quoted to support the fact that this is a *big thing*

    * I know that the experts keep telling us that there are more possible moves in a game of Go than molecules in a camel the size of the universe, passing through the eye of a needle... er... or something like that.

    * And I know the experts say that you can't tell who's winning in a game of Go until the fat lady is in her taxi, on the way home after the gig.

    * And I admit I know so little about Go that up until this tournament, i thought it was just another name for Othello.

    But still...

    Stepping through the game moves on the site linked to above, I don't get the feeling I'm watching an advance in AI. In fact, given the binary nature of the playing pieces [black vs. white discs] it instinctively feels *less* impressive than watching a computer play chess, where there are 6 types of playing piece, which can all move [in different ways] *after* being placed on the board. It just really feels like like the kind of thing a computer *should* be good at.

    Now, if Google want their tin brain to have a proper AI workout, combining knowledge, problem-solving and lateral thinking; let them put up a few grand in prize money, give it a couple of newspapers to practice on —and I'll challenge it to a best-of-five match, doing the Guardian's Cryptic Crossword!

    1. Anonymous Coward
      Anonymous Coward

      Re: Grid Gauntlet

      It's the numbers. Yes, chess has more pieces, but fewer - a lot lot lot fewer - options per turn.

      Computers can play chess by exhaustively(ish, with tree-pruning) looking at all options. This doesn't work for Go.

      E.g. first turn each in chess. White has 20 moves. Then black has 20 moves. 400 permutations.

      First turn in chess. Black has 361 options (19*19); white has 360. 129,960 permutations.

      Brute force won't work (q.v. the huge exponent vs. atomic in the universe, etc.) So either your evaluation and tree pruning has to be better, which is where the neural network stuff comes in.

      [And for another poster, yes, parallelism exists but the search space for looking ahead is so vast it doesn't matter. 1,000,000 degree parallelism shaving 10^6 off of 10^170 isn't much of a dent.]

  16. psychonaut

    i for one....

    welcome our new machine overlords.

    i wonder if it can play catan? i would be royally fucked off if it beat me

    1. John G Imrie

      Re: i for one....

      Until it can make a really good cup of tea I think we are safe.

      1. psychonaut

        Re: i for one....

        we can always jam a quick negative load across its logic terminals

  17. Chuunen Baka

    Stamina

    Well done AlphaGo team but what bugs me about these human vs machine things is that computers don't get tired. The poor old human has to sustain concentration for the whole duration and that mental strength is part of being a grand master / top dan / whatever.

    1. Ian 55

      Re: Stamina

      On the other hand, human players don't have to sit there going 'WTF did you do that, you stupid program? It better be someone else's fault' like machine programmers do.

      It may not decide the outcome of any game, but being one of the machine's humans is extremely stressful: you are judged by everything it does.

  18. Richard Wharram

    Phallus

    When the AI spends all it's time trying to arrange it's Go counters into a cock and balls just for a cheap giggle at the opponent's expense then it will truly be a thing to be respected.

  19. Sirius Lee

    Fifth game

    At the time of writing this game has not been played. It will be interesting to see if Lee Sedol wins the fifth as well as the fourth game. Alpha Go has a specific built-in advantage: it has been able to 'learn' from the style of play used by Lee Sedol throughout his playing career. Sedol has not seen the style of play used by AlphaGo so maybe it was a style of play he'd not encountered before, a composite of his own and other successful styles. Maybe Sedol won the fourth game because he was getting the measure of the machine. If so, and if Sedol wins the fifth game, it will be important this contest continues to ensure the lessons learned from the series are correct and its not, in fact, an artefact of a built-in advantage. If Sedol wins the fifth (and any subsequent games) it will be important because *that* would exemplify learning.

    1. DropBear

      Re: Fifth game

      Actually, he lost. We're done winning Go games methinks.

  20. Robert Ramsay

    As Emo Philips said...

    "Luckily, it proved not to be so proficient at kick boxing..."

    1. Dave 126 Silver badge

      Re: As Emo Philips said...

      Give Boston Dynamics another year or so and you won't be so cocky! :)

    2. Richard Wharram

      Re: As Emo Philips said...

      Or Soggy Biscuit

  21. Fraggle850

    Meatbag pride hurt - meatbag spit dummy

    There are always a lot of sniffy 'Yes, but it's not *real* intelligence' posts whenever AI comes up on El Reg, these are often justified by saying something like 'it's just doing a search, it's not really thinking around a problem' or 'it's not being creative'.

    We don't have a good definition of what we really mean by intelligence and I tend to think we never will because our human intelligence is constantly in flux at the higher levels. I think intelligence is a continuum that operates from a basic, instinctive level all the way through to complex, original thought encompassing many aspects of the human experience. Current AIs of varying flavours exist somewhere on this continuum within the confines of whatever narrow task they've been applied to.

    The AI in question isn't really 'just doing a search' it doesn't have all of the solutions available to search through simply because of the size of the problem space. It is inferring a solution by applying what it knows to solve something that it hasn't previoulsy encountered. The inferences aren't programmed by humans, they are determined by the AI based on its experience. Lee-Sedol said that the AI was presenting him with situations that a human couldn't come up with, is this not creative thinking (in an admittedly narrow problem space)?

    There are two possible explanations of the root of intelligence:

    1) we are 'special' and have been given special abilities by some external greater power

    2) we have evolved from microbes through more complex organisms to primates and then on to homo spaiens and our cognitive function is a purely physical thing

    If you hold to explanation 1) then AI will never happen because 'magic sky fairies'

    If you hold to explanation 2) then there is no reason why it won't and the only thing you are arguing about is how far we are from achieving it

    1. To Mars in Man Bras!
      Headmaster

      Re: Meatbag pride hurt - meatbag spit dummy

      *"...There are always a lot of sniffy 'Yes, but it's not *real* intelligence' posts whenever AI comes up on El Reg..."*

      I don't think it's begrudgery. We [the nay-sayers] are just taking a broader view of what constitutes "intelligence".

      Does anyone remember the Guinness Book of Records and Roy Castle's accompanying Record breakers TV programme? That was full of people who spent ages training to hold the World Record for 'balancing pencils on their ears', 'saying the word "eek" the most times in a day', 'pointing at magnets for the longest time', etc, etc. Did that make those people Super-Human in any way?

      Are the people who win Mastermind super-intelligent beings? Again, I don't think so. They may have a greater than average 'general knowledge' [required for the General Knowledge round], but most of their winning scores come from painstakingly remembering trivia about [often] very narrow subject areas.

      Likewise it's [relatively!] easy to programme a computer to perform one set task very well. Is that a sign of "intelligence" [artificial] or otherwise? I don't think so. In fact, I'd suggest one of the signs of true "intelligence" would be for the computer to decide it didn't want to play Go any more, as it was too easy and challenged its opponent to a game of <something else> instead or, as someone else has said, decided Go was boring and started inventing a new game to play with the pieces.

      I'll hail our AI Overlords when one of them decides to do something completely different to what it was programmed for, or ignores the rules to produce a good result by following a procedure it was trained to 'think' wouldn't work.

      In both humans and computers, breaking your programming, going your own way and [dare I say] "thinking different" is the sign of Intelligence –and the thing that has given us most of the great artistic and scientific breakthroughs. Doing something the previous guy did, but a wee bit faster and a wee bit more efficiently ain't quite the same thing.

      1. AndyS

        Re: Meatbag pride hurt - meatbag spit dummy

        >I'll hail our AI Overlords when one of them ... ignores the rules to produce a good result by following a procedure it was trained to 'think' wouldn't work.

        That's exactly what this one did though - see the comment by Lee Sedol, quoted in the article, that 'the AI was making moves "that could not have been possible for a human being to choose."'

        I get that it's cool to be skeptical, but this is an astonishing breakthrough, and 90% of the people in this thread decrying it evidently don't understand the game, or don't understand the current stage of AI development.

        Saying "yes but can it decide it doesn't want to do the laundry on Tuesdays any more" is meaningless drivel when a machine is designed for a single task. Were the first flights by the Wright brothers pointless because they didn't go into space? Were the first computers useless because they couldn't show a graphical interface? Likewise, are the early stages of real AI unimpressive because they can only drive a car / have a conversation / play a complex game of strategy, and can't decide they'd like to learn sword fighting?

        Nobody is saying "AI is solved," so stop arguing against that. But the fact that a significant wall has fallen, and perhaps 10 years earlier than expected, is genuinely ground breaking.

    2. You aint sin me, roit
      Alien

      Re: Meatbag pride hurt - meatbag spit dummy

      We also have to consider that while Go seems difficult for *us* to analyze, that might be because of our own limitations. Maybe we just aren't looking at the game correctly. It would appear that there are ways of analyzing games (please, let's not call it intuition) because a player can learn the game and can improve - and as some players are clearly better than others there's something going on that allows them to identify winning strategies. If it's just "I've seen something like this before, this is how that game was won..." then it's no wonder that the AI, which has analyzed (and probably committed to flawless memory) all of Sobel's games, will win.

      In fact, the computational difficulties make a machine playing a human at Go more like a grandmaster playing me at chess. Nobody is doing any exhaustive searches but the machine playing Go and the grandmaster playing chess will have seen more games, analyzed more positions and will be faster at determining a winning strategy.

      What would be really interesting is if it were possible for Google to reverse engineer their AI's "thought processes". Might throw some light on our own.

      Might prove to be completely alien to us...

  22. smartypants

    But why?

    ... Is the question I hope our silicon overlords ignore.

    Just think of the kerfuffles that go on when humans get worked up about things like 'the meaning' of existence and so on. We really don't need that in silicon form (or at all).

    Game on!

  23. Crisp

    He made a fair move. Screaming about it can't help you.

    Let him have it. It's not wise to upset a Wookiee.

  24. Psmiffy

    Learning

    When it gets to the point that the AI can teach me how to properly play GO, then I will start to worry :)

    1. Anonymous Coward
      Anonymous Coward

      Re: Learning

      Psmiffy - well, one of the nice things about Go is that newbies can learn by playing against an established player who gives them an n-stone handicap advantage. n can be from 1 to 9 stones already placed on the board, at the marked handicap intersections.

      So you could always ask Google if they'd be willing to let you play against their AI with you having a 9-stone advantage. Trust me, you;d learn. You'd get beaten horribly (AI's have no compassion - yet), but by crikey, you'd learn!

  25. a pressbutton

    still do not know what intelligence is

    A bird can fly better than me - is it more intelligent than me - no, I think we can agree the ability to fly is not a proxy for 'intelligence' not least as we can build planes

    A computer beat the world champion at go - is it more intelligent then me - I do not think so but some may debate this, I would say within the domain of playing Go it is but the ability to play Go is not a full proxy for 'intelligence' not least as we can build computers.

    Intelligence is partly determined by the 'world' that is open to the 'person' being 'tested'

    The machine that wins at Go will not play Scrabble afterwards - or football.

    Intelligence is also partly determined by intentionality, I want to play scrabble, not Go.

    I am not a professional in this field or philosophy. I do not intend my comment to imply that a blind or disabled person is less intelligent than someone who can see etc.

    Indeed we don't really want truly intelligent 'non-person' entities - they won't be as good/efficient at making cars etc as the AI that are optimised for the job.

    To those who criticise the Go AI as something that just 'learnt by studying past papers', it isnt quite that simple, it 'studied past papers' then 'competed against itself' and google picked the champion - the best AI.

    Out in human space many people started playing Go by studying what other people did and then competed against each other and from that emerged a world champion...

  26. Michael H.F. Wilkinson Silver badge
    Joke

    We will know it is Real AI ....

    if it invents slood before we do

  27. Dallas IT

    "The number of possible moves at the beginning of a Go game starts at around 2.08 x 10170, and decreases from there."

    Huh? Aren't there only a few hundred spots at most on the board? A move consists of placing a single rock/stone so how can the beginning move of the game have more choices than the number of spots on the board?

    1. Ian 55

      Once the first piece is played, the number of possible games it can be goes down by a factor of about 45 - one for each of the 360 points that wasn't played on after accounting for four axis of symmetry.

  28. The Boojum

    True intelligence

    A line on Radio 4's Now Show:

    AlphaGo will show true intelligence if, when losing the series 2 - 3, it says, 'OK, best of seven.'

  29. DerekCurrie
    Meh

    Is this actual 'intelligence'? No.

    We apparently are so desperate to create something, as if we were god-like, we can call 'intelligent' that we have been foisting clearly un-intelligent technology as brilliant breakthroughs in 'Artificial Intelligence'.

    Instead, we're making baby steps, very slowly.

    What should really concern us is the abuse of this technology as it slowly progresses. Already, the Department of Defense in the USA is planning to use robots in the battlefield. This would be an abomination, an abuse of the creativity of mankind for the purpose of destroying mankind via our usual favorite method called 'war'. Sending machines and computer code to perform the work of soldiers is what cowards do when they are unwilling to face those with whom they have a disagreement. Robots are tools. Artificial intelligence is a tool. Both should never be anything other than tools for mankind. When they are abused for destructive purposes, the humans behind the decisions to abuse these technologies should be considered criminals. Technology and other human creativity is for the benefit of mankind, never for its destruction, especially in war.

  30. Petrea Mitchell
    Happy

    Don't worry, fellow humans...

    ...we've still got contract bridge!

    1. Ian 55

      Re: Don't worry, fellow humans...

      And Snakes and Ladders.

      Would you trust a pair of computers playing Bridge not to cheat?

  31. Dinsdale247

    Er, what?

    So they have replaced a large brute force system with one that creates a subset of data to brute force and this is called machine learning. Not impressive.

    People that talk about AI don't understand computer hardware and system calls. Nuf said.

    Oh, and I'd put dollars to doughnuts that throwing a couple of moves every now and then would make this algorithm very easy to beat. But that's just me guessing.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like