Are computers really better? (argument POV)

I used to think that computers were definitely better than human players.. because that's what appears to be true, but I think there is more to the discussion. 

I think the key element is time. A computer has incredible fast computing power so that it can process thousands of lines in shorter amounts of time. So I started to think what was a fair time increment that a GM could beat a super computer as in: if a computer had 5 minutes on the clock for a GM 1 hour. This would however assert how dominant the computer was.

To further the discussion of time however I think something more important is that again the computer can process lines more quickly. So is it fair that a human opponent would play the game more like a coresspondence game vs a super computer? 

The arguement then is: is a computer really smarter? or does it just analyze lines faster. If a human could un-due the speed element of the processing and match up based only on thinking and strategy would a super-GM out manuever and out-strategize a computer if that person had as much time as he needed like in correspondence chess to go through every line necessary. 

Is a computer really smarter or is it just faster at going through lines? 

That's definitely an interesting thought.

Super computers can beat people without opening book, but I like what you said about computers using an analysis board. 

I think that it would be fair if a Super GM like Magnus played a super computer to be able to use an analysis board (no opening book however) and also make it correspondence style, because people have limitations that computers don't have like short term memory (missing something down a line of 15 moves where-as computers have everything "written" in memory.. it's probably useless for a person who is more fluid and creative), processing speed of lines (computers are at their best at moving through data quickly), and energy (computers have no energy difficiency from sitting at the board too long.. as long as they don't short circuit or a fan doesn't break they can maintain a steady and consistent output) 

I think the best analogy for a chess engine is basically like a math calculator. But not all calculators are designed to do super complex math because it all has to be coded.. and they don't figure out math on their own. It is programmed by equations. 

Chess super computers are programmed to probably solve the problems in a chess game, but they often break chess principles probably because they have no emotional attachment to the game; however, if a super GM had enough time they might be able to disprove some of the random and obscure play of a super computer. 

Why are you using the word "better" which can be rather vague?The word you are looking for is "stronger".And the answer is yes,engines nowadays are stronger than humans.Simple as that.

Computers cheat because they are using opening book and tables.

The reason computers don't play perfect chess is not that they can't calculate, it's that their evaluation is not optimal.

You only need to look at another game, Go, and the attempt to create AI that can beat Go professionals, to realize that computers aren't "better" thinkers than humans. So far no computer was able to beat a pro-Go player without significant handicaps. The strategic elements there are too strong.

In a similar fashion I could imagine that Magnus' strategic understanding together with an analyze board, opening books, and unlimited time could at least pose a serious challenge to Stockfish. All he needs to do is survive intact until the endgame, really.

Here's an article on computer programs and Go: http://www.wired.com/2014/05/the-world-of-computer-go/

Considering the personalities of some Super-GMs and the sometimes significant difference in playstyles, I'm not sure if "more GMs" would actually make the human team smarter.

@gum: 4-dan is an amateur ;) and even there the computer had a stone advantage I believe.

In the case of Go, it merely means Go is not a thought-dependent game, relying more on luck and inter-personal communication than selecting an objectively best place to put your stone.

Sorry I hate Go but you cannot deny that it is perhaps deeper than chess. There is no luck in Go, just like there is no luck in chess. The REAL reason computers can't be programmed for go is there is too much space on the board. Empty space confuses computers like nothing else. The same phenomena happens in chess especially in a certain player's endgames.

*cough* smylov *cough*

How does a GM having more time prove anything? If Nakamura has one hour and Carlsen has thirty minutes and Nakamura wins do we say that Nakamura is stronger?

I'm not sure I completely buy that a comptuer calculating lines is like using an analysis board. Is a GM calculating in thier head not doing the same thing?

It seems clear to me that under equal conditions the 3000+ ELO engines are indeed stronger than any human grandmaster, even when the computer is given a disadvantage. See: Nakamura/Stockfish match.

@Quicksilver: Go is not a thought dependent game that depends on luck? You were clearly either facetious or have no idea about professional Go.

However, that topic already has it's own thread here http://www.chess.com/forum/view/general/go-versus-chess?page=5

@Arawn: A human has imperfect memory and utilizes only a fraction of it's brain's processing power. Meanwhile a computer literally can not accidentally mess up the placement of a piece 15 moves ahead. So essentially a computer does have an analysis board integrated in its program. It doesn't have to "visuallize" the chessboard in the first place.

The reason why the human brain is less powerful in chess is simple: the human brain also has to manage all the functions of the human body, in addition to solving chess problems. 

Thus, I could also turn it around: How is it an equal match, when a computer doesn't have to use processing power to regulate blood pressure, coordinate muscles to move pieces, filter sensory input, and visually recognize the chessboard pieces?

Computers aren't just *faster* at calculation, they're also more *reliable*.

If i try to calculate, say, five moves ahead, and I look at five good candidate moves on each side for each move, well I'm trying to consider about 100,000 possible variations.  That's *way* more than I can actually remember, and far too many for me to even have any confidence about whether I've properly investigated any candidate move.  

So most of the time, I'm not really calculating, at least not more than two or three moves ahead.  I'm just trying to avoid obvious tactical blunders, and when two or more moves seem equally safe, using positional rules of thumb to try to pick the move that leaves me in the best position vis-a-vis my opponent.

A super-GM using long time controls can accurately calculate immensely better than me, and he has much more experience to guide him about which candidate moves to consider (and which not to bother exploring), but GMs still overlook moves and make mistakes of a kind that computers simply do not suffer from.

Back around 1980, the Belle computer was evaluating over 100,000 positions *per second*.  It was "only" rated something like 2200 USCF, so obviously just evaluating lots of positions isn't everything.  You still have to evaluate them correctly.

[quote] I think it is only a matter of time until chess engines are programmed to have more human "shortcuts" ie positional play, which will make them unbeatable.[/quote]

No, the trend is the opposite.  Computer programs have shortcuts and rules of thumb as part of their evaluation function, because the computer isn't fast enough to play out the whole game and give a truly objective answer to whether a position is sound. But ever since computers started getting decent at chess, back in the 1970s, we keep finding out that what we thought we knew about chess was wrong.  For example, endgames that were thought to be dead draws turned out to have winning moves, that no human had ever found because you had to calculate 50 moves ahead before you got to a position that a human could recognize as being better than the starting position.

 The better the computer is at calculating, the less need there is for humans to program it in advance with human hunches about which positions are best.  So, the faster computers get, the *less* positionally they'll play.  They'll still make lots of moves we'll recognize as positionally sound, e.g. fighting for the center during the opening, but we'll see more of the bizarre-seeming intermezzos we already call "computer moves."  

The problem is computers can often still win withput an opening book. I wonder if anyone knows if they perform do good without endgame tablebases also...

No because they evaluate the above positional sac from Petrosian as clearly better for white when it is in fact drawn!  

That is true, but the question is whether they will still play it for Black, even with wrong evaluation.

This doesn't prove anything except that computers can also misevaluate positions. If you research chess engine programming you will find that there are in fact many tactics sets that contain problems that computers do not solve correctly. It does not erase the fact that the 3000+ ELO engines are much stronger than the strongest human grandmaster. Magnus has said himself that he doesn't even see the point in playing computers.

You can research the details of this match. It shows that even with a clear advantage the human cannot hang with the strongest engines. And this particular engine is not even the strongest in the world at the current moment:

http://www.chessgames.com/perl/chess.pl?tid=84631

Even if you call that cheating, it's still only a minor factor in rating performance. I ran a 40 moves/4 min time control match between two copies of Stockfish 6, of which one copy had the opening book and endgame sygyzy bases disabled. The lack of an opening book only hurt it less than 30 rating points, and the lack of sygyzy bases did hurt, but it was barely measurable. (Of course, I only had a 5-piece syzygy set running on a mechanical hard drive.) Other conditions: Engines were only running on a single core (thread) on a not-very-fast PC; Opening book was Sedat's Perfect v10 book.

Computers don't really think, they process.

A computer program is only capable of doing what it was programmed to do. So even if you have a very good chess playing computer, ask yourself, who programmed the computer? 

It was a human of course. 

You could have the fastest supercomputer in the world and allow me to write some super super sucky chess alogrithm for it and it will lose against Grand Masters everytime. The strength of the computer besides the obvious fact that they can process informtion super fast and have loads of memory(RAM) is in the strength of the programming(software).

So we give computers a lot of credit but it is the human mind behind the computer, transferring our human intelligence to the computer circuits. It's like putting our mental fingerprint in electronic form. 

I see far more wonders and awe in God's creation (the human mind) than man's creation(the computer). If it wasn't for the human mind the computer wouldn't exist and if it wasn't for God, neither would we.

**Remember computers didn't create themselves and man didn't either**

"So we give computers a lot of credit but it is the human mind behind the computer, transferring our human intelligence to the computer circuits. It's like putting our mental fingerprint in electronic form. "

I don't think this is a fair description.  If a computer program is just a "transfer of human intelligence", why do the computers beat their programmers at chess (and go, and countless other tasks)?  

 Put another way, is a steam engine just a "transfer of human strength"? Is an atomic bomb?  There's something new in the world, and its properties are different from those of humans.

"A computer program is only capable of doing what it was programmed to do."

That's not a helpful way of thinking about it either.  Computers can be programmed to learn.    That's how the new Go-playing programs work: Go is too hard to calculate even for a computer, so the programs learn as they play, just like humans (but faster).  (Bit of a refutation of my earlier post in this thread.  This new "deep learning" approach is so promising, chess-playing programs are likely to adopt it too, leading to more human-like play.)

As for the Flying Spaghetti Monster, well, let's just agree to disagree.