Deep Blue vs todays (computer chess programs)

15 years of hardware improvements (= enormosly much!) plus gigant steps in the algorithms for evauating positions.

Just look how Houdini is turing old TOP engines (like Fritz 6) into pieces. Fritz doesn't stand a chance.

Of course, Houdnini would probably win by a large margin.

(What ever happened to Hydra, btw; poor M.Adams ... )

That's somewhat questionable. Deep blue was a computer that contained only 15 million transistors. Today's home computers have billions of transistors, thanks to the progression of Moore's Law. Blue Gene can do much more work than deep blue, being a super computer designed to do one petaflop (or one quadrillion) floating operations per second. Deep Blue could only preform one trillion floating operations per second, much less than Blue Gene. But here's a problem; as much computing power as blue gene has in comparison to deep blue, Blue Gene is not designed to play chess. And once you get a computer to analyze 200 million positions per second (averagely), and analyze rational possibilities for 6-20 future moves? You've pretty much beaten every human on the planet indefinitely. Yes, there are many possible chess games out there, but it's likely that each computer will choose the correct scenario considering they can see so far ahead. And deep blue was specifically made to use ALL computing power to play chess. Chess software only uses a fraction. I'm betting it's more likely that deep blue would end in a stalemate with whatever it faced. Too a computer that can see so far ahead, and has analyzed 700,000 grandmaster games? Chess to it becomes more like Tic-Tac-Toe rather than the complex game it is to us humans. While there are tricks around chess software, deep blue was built to be more of a free thinker out of parameters. Unfortunately, it has been disassembled, so there really isn't any testing to prove my statements. 

No question that if Deep Blue was using the now outdated supercomputer that it used, it would probably get beaten by houdini or some other engine. But put deep blue on todays supercomputers and there would be no contest, simply because todays supercomputers are so much more advanced. Deep blue with old tech=loss for blue. With new supercomputers=win

Haha, I'm not going to retort that . But thanks for giving me some new information. I'm just pointing out that deep blue is still a capable computer. I don't really know very much about todays software, but considering some of the information I just told you (about blue gene and such) I just proved myself wrong. In fact, when I was typing my comment, I was going to a say, "that is somewhat questionable, but I agree".

I find it fascinating that nowadays a ten year difference in computers gives such an advantage. That was not true 100 years ago obviousley. Technology does advance exponentially.

deep blue would be raped easily by todays engines. i  have compared deep blues games on the new engines and the new engines find alot of deep blues moves to be poor. also i was just analyzing on a 64 core rybka  looking at 100,000, 000 moves per second. that would  just wipe the floor with deep blue 6 and 0 for sure

oh and about fritz. deep friz 13 is still a must have even if u have rybka houdini or whatever . fritz can still see moves in certain positions faster then all the other engines. every engine like say the top 10 are all worth having as they have theyr own style. when analyzing u can  add kibitzer engines and i find this to be extreamly usefull.

The Deep Blue that Kasparov played would be crushed by one of today's top engines running on todays desktop machines.

for which part you need time travelling??

Presumably the hardware is no longer in existence?

Kasparov himself said many times a few years after 97' D.B. match, that any computer today is way better and cheaper. But that match was quite questionable...

But you see, computers technology grows exponentialy, and their power double each 2 years (aprox.)... After 15 years, any comparison is absurd (and the combination of hardwre+software make a wolrd of difference).

And now new technologies are comming, and I guess that everything will change a lot in the next 10 years.

Hardware technology has been growing exponentially, for sure. In the case of software generally, I doubt that performance has improved - how else to explain that conventional PCs still take just as long to boot into a useable state and that simple interactions such as opening a window showing a list of files seem to be no faster (and in the case of Window 7 vs Windows XP may even be slower on some machines)? However, chess engine software has improved and will probably continue to improve as evaluation functions are refined. Maybe we will even see developments in terms of chess programs being able to "understand" or recognize certain positional motifs that currently cause them to misevaluate, e.g. the positional queen sacrifice in "Gusev's immortal" http://www.youtube.com/watch?v=1i_4mTeFlDs .

They take longer because they bring up to the memory LOTS of information that older softwares doesn't have.

Compare the complexity (for instance) of a Win7 vs WinXP window. In 7 we have lots of glass effects... More info, more calculation, more consumption of the hardware resources. Of course, this is just a simple example.

WinXP installer fits in a CD. Win7 requires an entire DVD... That's about 6 times more data wich the hardware will be forced to handle CONSTANTLY while managing the user's commands.

But about chess, that's another story...

"They take longer because they bring up to the memory LOTS of information that older softwares doesn't have."

I believe this is called "progress". 

"Compare the complexity (for instance) of a Win7 vs WinXP window. In 7 we have lots of glass effects... More info, more calculation, more consumption of the hardware resources. Of course, this is just a simple example."

Interestingly/annoyingly, I have often found Win7 to be less responsive than XP even when the new special effects have been disabled and I have set my desktop theme to Windows Classic (which I think looks better).

"But about chess, that's another story..."

Agreed.

Of course. The windows theme/style was just a simple example. There are many other features running in the OSs that we don't even know (thankfully!)... But everything "progresses" in the same ratio. 

You mean the software is getting exponentially slower but in proportion to the exponential speed-up of the hardware? That must take a bit of doing...! 

hehehe!

No, Mr. Burton... I mean the software is getting heavier (not slower), and that forces the hardware evolution to be able to handle all the new features and data implemented in the softwares.

One compensate the other (aproximately), but even if we get one or other detail slower, look at the incredible difference of performance, capacity, interface, features, tools, etc. we have now in our home computers!

It's incomparable what anyone can do with a pc today, and what only trained programmers was able to do with a computer from the 80's for example.

If you try to just get rid of any interface feature (for instance) only to improve performance, you will reach a point where de performance will be incredibly fast, yes, but you will have to use something like the old DOS to deal with your computer (or even worse, write everything in Assembly or Binary code)...

So, just like in some chess situations, we have to sacrifice a little bit of something to get a better situation as a whole.

Also, I must add, chess was (and is) the greatest contributor to computers evolution, because it can force any computer to the limit. See all the effort to build Deep Blue, and the many inovations that came from it and other researches like that.

Despite any apparent obtuseness on my part, I almost entirely agree with what you wrote, Mr BDB. However, I do not see that features, useability and performance have increased to the same extent as e.g. effective CPU speed. Have these things really improved 16-fold over the last 10 years (say)?

Hi everyone, I will jump in here as I had followed the event fairly closely.

Deep Blue was a super computer that had specialized circuitry boards designed to calculate chess moves and chess moves only, it was built by IBM corporation. It lost the 1996 six-games match, beaten by Kasparov 4-2. It could calculate 100 million positions a second.

The following year, 1997, they had a re-match, and Kasparov lost 2.5 to 3.5 in the six-games match. This time Deep Blue's hardware was doubled so it could calculate 200 million positions a second, as well as software improvements. Because of this dramatic increase in strength, it was to be called "Deeper Blue," but apparently the name didn't stick and everyone just called it Deep Blue after 1997.

IBM maintained a website for the matches for a number of years, and text version of the game comments could be downloaded from the website. However, the text files were cut and only contained the comments from the very beginning of each game. I contacted IBM and the answer was something to the effect of they don't know where it is and there is no backup, so can't fix/won't fix. I had the complete version of the comments back in 1997, but didn't keep copies as I assumed they could still be freely downloaded. As a result these comments are permanently lost to me, and I have no idea who out there would still have them, if anybody at all.

Let's look at today's PC performance in 2012, some 15 years later:

Fritz chess benchmark 4.3:

AMD's best 8-core CPU: FX-8150: 11.8 million positions per second

Intel's newest quadcore CPU: Core i7-3770K: 14.0 million positions per second

Intel's really expensive 6-core CPU: Core i7-3960X: 19.3 million positions per second

So the best commercially available PC hardware (before you get into multi-socket workstation and server hardware) today, will get you close to 20 million positions per second, an order of magnitude below Deep Blue's 200 million back in 1997.

Does this mean Deep Blue was stronger than a combination of modern PC hardware and chess software? It's very hard say. Today's chess programs are very smart and trim many unnecessary calculations off its search tree, making them highly efficient compared to the chess programs from the 1990s. With this said, Deep Blue's software was highly sophisticated and it is no slouch even by today's standard as far as the heuristics go, plus it was programmed with all opening and ending chess knowledge available at its time and extremely intelligent. I think most people underestimate how well-programmed Deep Blue was, it was no average PC chess program from the late 1990s.

Deep Blue was de-commissioned after the 1997 match, as it had served its purpose of beating the reigning world chess champion. It is in IBM's storage facility collecting dust nowadays. There is a documentary about the events:

http://www.imdb.com/title/tt0379296/

However the documentary accuses the Deep Blue team of cheating by using a human chess player, an idea Kasparov somehow had due to Deep Blue's humanlike chess moves in 1997, I do not think this accusation was sufficiently founded.

Overall though, I will think today's top PC chess programs such as Rybka are stronger, as its heuristics would more than make up for the raw speed of Deep Blue. I don't think the question will be completely settled though until PCs can in fact calculate 200 million positions per second. Judging from the current rate of computer hardware advancement in the last 15 years, this will most likely be achieved sometime around year 2020 or soon after, probably no later than 2022, if we are talking about top-of-the-line consumer level PC. A multi-CPU workstation level computer can probably be built before 2018 with that capability, the kind of computers people build to win computer chess tournaments, but it will be several times more expensive than your average Bestbuy PCs.

Deep blue was a supercomputer, not a program ;)

We don't have an exact rating for Deep Blue because it did not play enough games against rated players to be rated.

From this website, "In six years, a program like Deep Fritz will again achieve Deep Blue’s ability to analyze 200 million board positions per second. Deep Fritz-like chess programs running on ordinary personal computers will routinely defeat all humans later in this decade."

This article was written in 2002. I would not be surprised that any decent home computer could play as well or better than Deep Blue considering how fast computing speeds grow.