Winning starts with what you know
The new version 18 offers completely new possibilities for chess training and analysis: playing style analysis, search for strategic themes, access to 6 billion Lichess games, player preparation by matching Lichess games, download Chess.com games with built-in API, built-in cloud engine and much more.
I'd like to thank everyone who's sent me e-mails over the last few months. Yes, I do read them all, even though I may not return a response or use them here in ChessBase Workshop. Thanks again for sharing your thoughts and ideas with me.
The last few months have seen some interesting ongoing discussions prompted by some of my previous columns, namely the "intelligent mistakes" and "computers solving chess" topics. We've already closed the "intelligent mistakes" topic (with this week's singulat exception; see below) but that hasn't prevented more e-mails from arriving on that subject. I apologize, but I won't be using them here (there's other topics which must be covered instead). And one e-mailer in particular has sent responses to other e-mailer's printed responses, and then followed up with some responses/addendums to his prior messages.
I'll make a suggestion here: the bext bet would be to start an Internet message board on those two topics (if you'd like). If you do so and send me a link, I'll provide it in a future ChessBase Workshop column; I might even participate in the board myself. But a weekly column (the publication of which is often delayed by more than a month) is a lousy place to try to hold a "give and take" discourse, so I'll again recommend a message board as the way to go.
That having been said, a future column will provide the last of the reader responses to the "computers solving chess" discussion, and then we'll move on.
A curious note: ChessBase has been publishing my columns "out of order" (that is, in an order other than the ones in which I write them), which is why you'll sometimes see me write "Next time around we'll look at..." and then the next column has nothing to do with what I'd indicated. What this signifies is that I'd better stop offering "next week previews"; consider it done.
And a note which made me laugh. I recently saw a chess web blog (or "blog" as common parlance calls it) in which I was referred to as a "National Master". I'm sorry (for more reasons than one) but not only am I not a master, I'm not even a titled player. I'm just an average Joe, down here in the trenches alongside 90%+ of my readers, just trying to get better at this game. Country singer Gretchen Wilson once said that the only difference between her and her legion of female fans is that Gretchen can sing. That's how I feel about this writing gig. As a chessplayer, I'm just like most of you; the only difference between us is that I spend a truly ridiculous amount of time exploring the potential of these software tools and somehow got blessed with the ability to explain what I find in something resembling a clear (and hopefully sometimes entertaining) manner.
Finally, I wish all of my readers a happy, prosperous, healthy, and safe 2006. May you all gain at least another 500 Elo points in the year to come.
As I indicated above, I received a lot of e-mail. One recent message in particular caught my eye; I thought it deserved a column of its own, as well as a response from me.
By way of preamble, let me explain that I'm not a scientist, nor do I have an advanced college degree. I do, however, know a little bit about the way people think. When I started with ChessBase in 1992, I was completely blown away by the potential of this software to help chessplayers learn more, practice, and improve their game. At that time the CB distributor I was working for was releasing a line of CB "add ons": instructional disks designed to take players from beginner to an advanced level. Now as I said above, I'm not a titled chessplayer nor am I a scientist, so what could I contribute to the effort of helping players improve? I quickly discovered a niche that wasn't being filled: that of "interpreter" or "evangelist" or whatever you choose to call it (some have uncharitably termed it "shill" to which I respond with a term of my own that I can't print here). I saw the enormous potential that these software tools contained; my job became to unlock that potential and pass on what I learned. My boss thought I was perfect for the job, since I was an average player struggling to improve. Maybe my position could best be termed "guinea pig" -- if Steve could learn from this software, anybody could.
But to really understand the potential of the software, one of the areas I needed to explore was the process of how people learn new things. I needed to learn how people learn; I was required to learn how I learn, so that I could adequately show the potential to my fellow chessplayers. So I spent considerable time and effort reading up on the subject of human learning, how humans think, and the way it differs from the way computers process information.
Consequently, I consider myself to have a better than average layman's knowledge of the subject -- after all, it was my job to gain this knowledge and it's helped me a great deal in the last fourteen or so years; I still use it all the time when I'm writing these columns.
On now to the e-mail. It's from Francisco Rivera and it's in response to the "intelligent mistakes" column (yes, I know I said the subject was closed, but I couldn't pass this e-mail up -- it's that danged interesting). Mr. Rivera starts by quoting from that column:
First we need to sidestep to look at what a personal computer really is. Although I can do a ton of neat stuff with my computer, like play chess, balance personal finances, simulate a historical battle, build an empire, look at pictures, shoot the bad guys, and crank out a weekly chess column of dubious merit, what's really going on in the "guts" of the machine is that it's juggling numbers. And it's really only juggling two numbers: ones and zeros. When you strip away the upper levels of programming languages and look at the core of any digital computer program, it's just bouncing ones and zeros around. Everything is either "one" or "zero", "yes" or "no", "true" or "false", "on" or "off" -- there are no "shades of grey" involved.
Consequently computer programs are, at their heart, a series of "yes" and "no" decisions. Does such-and-such a condition apply? If so, do this. If not, do that. There are no "maybes" in this binary realm, no "fuzzy logic" -- either a condition applies or it doesn't. It's cold and hard as a brick. There's no such thing as "intuition" here, no vague factors that lead to sudden epiphanies.
And that's why I submit to you that computers don't "play" chess. A computer isn't playing chess any more than it's commanding a Civil War army, racing a car, or swinging a lightsaber. It's simulating a person playing chess -- and it's doing it by crunching ones and zeros, not by using intuition or imagination.
Mr. Rivera responded as follows:
Sorry but you are totally wrong. Your brain and mine too are digital computers. The difference is that our brains are massively parallel computers and PCs are serial computers. Each "transistor" in our brains is a digital but slow device named neuron. The neuron is a digital device, he is all or nothing. We have billions and billions of them, operating in parallel. The PCs have only some millions of very quickly transistors.
Anyway, the intuition and imagination are only a kind of "calculation" based in a strong memory support (Do you know some good chess player that forgets the games played?) Of course, is very difficult write a program that can use this type of calculation because the programmers doesn’t know how write such algorithm. The evolution wrote in our brains the required program and embedded in our brains.
He then offered another quote from my column:
That's exactly the reason why it's so hard to get a computer to play "human-like" chess (even beyond the fact that "human" in this case covers a lot of territory, from beginners to world champions). Humans don't crunch numbers the way a computer does. We don't play chess by looking at thousands of board positions, assigning each a numerical value, and then "minimaxing" our way to a decision on the best course of action.
And offered this response of his own:
The humans doesn’t crunch numbers because we have a very good "evaluation" function, therefore, we can cut all moves excepting the bests. Therefore, the human only need rudimentary search capabilities. In the other hand, the computers have very poor evaluation function because the programmers don’t know how to write good evaluation functions, therefore, computers use strong search capabilities for achieve a good level.
Now that's some awesomely thought-provoking stuff. I think that Mr. Rivera and I agree in principle; it's in the details that we differ.
I absolutely agree that the human brain is an advanced digital computer, with neurons acting as the "0" and "1" switches. The big problem here is that we really don't know how the human brain operates, other than the admittedly vague idea that it's a form of parallel processor. And I also agree that programmers don't know how to write an algorithm based on intuition and imagination -- and this is due directly to the fact that the operation of these functions in the human brain aren't understood.
But I still contend that a computer doesn't really play chess, any more than it engages in the other activities I mentioned. A computer doesn't "know" what it's doing; it's just processing information (in the form of the aforementioned zeros and ones). Human beings are very aware of what they're doing when they're playing chess (or any of a zillion other activities). An example from the enormously popular and successful The Sims series of games springs to mind. A human knows that when he's tired it's time to sleep, that when he's hungry it's time to eat, that when he needs to go to the rest room he visits said room and takes care of business. That little person you're controlling on the screen when you're playing The Sims doesn't "know" jack. The program sees that the hunger bar is way up in the red, so the Sim heads for the kitchen. If the energy bar is in the red, he or she sleeps (sometimes, hilariously, right in the middle of the front yard). And let's just pass up the opportunity to discuss the bladder bar (another opportunity in the game for some lowbrow hilarity). The Sim doesn't "know" anything -- it's all a function of the program. And the program doesn't even "know" anything, except that when the scores of the previously mentioned levels dip below a certain point it's time to trigger a Sim's game actions. So the computer itself isn't hungry or tired or uncomfortable -- it's just reacting to a numerical informational input supplied within the context of the software, and then responding within that same context. And that's precisely what a computer is doing when it's "playing chess": reacting to changing mathematical input.
What I was trying to relate in that previous was a philosophical difference between a human and computer ("computer philosophy" -- now there's an oxymoron for you). A computer doesn't know it's "playing chess"; it's just diddling numbers. And that's the crux of the problem, which you discussed in your second paragraph.
Do we really know that intuition and imagination are based on "strong memory support"? If such was the case, we'd never have original works of art or literature -- science-fiction (for example) would never exist. I'll readily admit that art & literature is often based on memory; you start with what you know and extrapolate from it to create something new. But memory can't possibly be the whole of the process; if so, remind me to never have whatever Picasso was having -- I like my women to have all their physical parts in the accustomed places.
So I'm afraid that I'll have to differ with you on that particular detail, even though you might have a degree in that area or may even be working in the A.I. field professionally (I don't know because you didn't say). I don't think that we really have a clue as to the origin of intuition and imagination, and maybe never will. I believe those qualities are among those that separate us from animals and aren't easily explained in scientific terms. (And since a lot of my readers seem to love the subject of theology, I'll skirt that issue as well to prevent a relentless deluge of e-mail on that topic).
In his last paragraph Mr. Rivera made a pretty provocative statement:
In the other hand, the computers have very poor evaluation function because the programmers don’t know how to write good evaluation functions, therefore, computers use strong search capabilities for achieve a good level.
Whoops. I have to disagree. A computer chess program can search all day to see what lies ahead, but it still needs to evaluate what it finds, otherwise the extensive search is pointless. So Mr. Rivera's statement is kind of like putting the cart before the horse. Computer programs by and large do have awesome evaluation capabilities (which is why they excel at short-term tactics); the limits of the search functions (due to the time contraints of not being able to see everything to any great depth) are why they tend to be deficient at long-range strategy.
However, I did love this particular comment of Mr. Rivera's:
The humans doesn’t crunch numbers because we have a very good "evaluation" function, therefore, we can cut all moves excepting the bests. Therefore, the human only need rudimentary search capabilities.
I've been saying this for ages, both in print and in private conversations. Many years ago (while I was in the midst of learning about how people think and learn) I had bit of an epiphany while watching a PBS show on Kasparov vs. Deep Though. GM Yasser Seirawan was expounding on a GM's thought processes while considering candidate moves. Yaz said that GMs reject 90+ of candidate moves instantly based on their intuition and experience, instead focusing their energies on the two or three candidates that "look good" in an effort to see which will bring the biggest reward and/or lead to the most promising position. That got me thinking, and I subsequently came to a realization.
I've said it before and I'll say it again: I think the single biggest difference between a strong chessplayer and an average one is not based on calculative ability (the ability to see farther ahead, as the media stereotypically portrays it), but that the strong player has the ability to choose better candidate moves right from the get-go which then leads to strong play.
And that, friends, is a function of intuition, memory, and experience -- and that's exactly what Mr. Rivera and I are really disussing here.
As some of my correspondents have stated, that may well be the core of the problem of making computers play "intelligent" but weak moves: until we can a) understand the "flawed intuition" of weaker players and b) find a way to mathematically quantify it in terms of a chessplaying algorithm, we might not be able to create a program that plays intermediate-level chess in a human-like manner.
"The devil is in the details" as the old expression goes, and that's really the only point of disagreement between myself and my corrrespondent. In a lot of ways we're thinking along the same lines.
Thank you, Mr. Rivera, for an extremely intelligent and thought-provoking e-mail!
More reader e-mails are on the way in a near-future column (see how nicely I avoided that "next week" thing?). Until next week, no matter what the column's about, have fun!
You can e-mail me with your comments on ChessBase Workshop. All responses will be read, and sending an e-mail to this address grants us permission to use it in a future column. No tech support questions, please.
