1978: Runkel on computer chess

41 years ago to the day...

I knew that Wolfram was ill — he discussed his heart problems quite openly with me. But it came as a shock when I learned that he had died last Friday (July 19), not of the cardiac ailment but of cancer, which he hadn't revealed to me.

In my recent eulogy I describe how Wolfram Runkel drew the attention of a rookie science journalist to a new subject for TV documentaries. In 1978, an article in ZEIT magazin (the text is given in full below), caught my attention. It described the progress that had been made in computer chess, how a machine had managed, for the first time, to beat a grandmaster in blitz chess, and how International Master David Levy had successfully bet that no computer would be able to beat him for ten years.

Runkel's article from 41 years ago is a truly remarkable historical document. It tells us what the status was and expectations were at the time. Would you believe that some of the following statements were made in 1978?

• In perhaps five years there will be, on the consumer market, chess computers with about the Chess 4.6 strength

• [Soon] the best chess player will be the one who is the most diligent student on his computer.

• Will players not take their electronic chess masters to tournaments, like students using crib sheets on the toilet, to find the best moves?

Wolfram's article and my subsequent activities dragged me into the world of computer chess – and chess itself. I had only been an avid player who had given up the game and stopped following the news on it (apart from Fischer-Spassky). But after 1978 I was suddenly occupied with the subject all the time, and I was permanently meeting top players — Nigel Short, John Nunn, Garry Kasparov, Vishy Anand, Vladimir Kramnik and many, many others, all of whom have remained friends to this day. I also advanced to be considered a leading expert in computer chess, although I was not a programmer, mostly a story-teller. Fortunately I met Matthias Wüllenweber, a full-blooded programmer, and together we founded ChessBase. That was in 1986, and Wolfgang Runkel watched the growth of our company, where he was a regular visitor, in the years that followed. When he complimented me on the success of the company I used to say: "It's all your fault, Wolfram!"

Wolfram Runkel and Frederic Friedel at the ChessBase Christmas dinner 2010

Fortunately I archived the original article and can show it to you now. I am doing it on Sunday, July 28, 2019 because I am a sucker for auspicious dates: Runkel's article appeared on July 28, 1978, exactly 41 years ago. For German readers I have prepared a PDF where you can read everything in the original form. For our English language readers I have made a full translation, which you can read below. It is a historical piece that is worthy of attention.

Wolfram Runkel

The computer says check

It's about a bet: will a computer, fed with a complicated program, be able to beat a chess master? And it's about more: Does the robot spoil the Royal Game for man?

David Levy's grave face becomes even more serious when asked how he now estimates his chances of winning his chess bet against the computer. A few weeks before Day X, August 31, 1978, when the ten-year betting term expires, the English chess champion, computer expert, and publisher of chess books and magazines, clearly has fears of his cool opponent.

Levy is no longer as sure of himself and his bet as he was ten years ago, when everything started so "very British": at a cocktail party after a game of chess, and of course with a bet — for 500 Pounds Sterling, for starters. Levy had beaten John McCarthy, one of the leading scientists in the field of Artificial Intelligence, leading the annoyed loser to the prophecy: "You can beat me, but in ten years you will be beaten by computer programs. " Levy accepted the challenge. In the meantime three other computer scientists have joined the bet against Levy. It's now for 1250 Pounds, 5,000 Marks.

The shaken ego of mankind

But now it is no longer just a humorous bet; it is about the future of the Royal Game, which has inspired people for centuries as an intellectual competition, as creative art, as a researching science. It is also about the question of whether the ego of man, who has accepted und put to use the physical superiority of machines, now has to acknowledge a mental, even creative superiority of machines.

In August, when the two gentlemen Karpov and Korchnoi fight for the conventional world championship title in human chess in the Philippines, Levy will stand up for humans in America (and maybe in Germany, if a sponsor for 20,000 Marks can be found) in the first official, serious two-game match against the program Chess 5.0 / Cyber 176, a chess computer, or better: a robot. Chess 5.0 will pick up pieces on the board with a robot arm and slam them aggressively on the target  squares (Levy: "I'm afraid that'll annoy me!"), and it will do so with comments ("That was easy"), adopting the psychological tricks of its human predecessors. Of course, without being susceptible to such tricks itself.

But that's not the only reason why Levy faces his opponent ("in some ways a Stroke of Destiny") with mixed feelings. For a year computer chess has made sensational progress in playing strength. Already, in 1977 the predecessor of Chess 5.0, Chess 4.6 won "human" tournaments and (albeit in blitz and simultaneous games) out-manoeuvered a number of International Masters, including Levy, and Grandmasters Michael Stean and Robert Hübner. Today already at least 99.5 percent of all (organized) chess players are hopelessly inferior to Chess 5.0.

The development of computer chess, from which telephone and radio technology as well as military strategy had made great use, began in 1948, when the American Claude Shannon created the first chess program, which admittedly at first made small progress.

"I was outraged..."

It was not until two decades later that "MacHack", a program written by a student of the Massachusetts Institute of Technology, defeated a human being in a chess tournament, the first computer to do so. In 1970, (Chess 5.0 predecessor) Chess 3.0, a program of North Western University students Larry Atkin, Keith Gorlen and David Slate, won the first official American Computer Chess Championship.

The inserted photo shows Larry Atkin (front) and David Slate at the 10th ACM North American Computer Chess Championship in Detroit, Michigan 1979. Photo: Monroe Newborn.

The Soviet computer "Kaissa" won the first World Computer Chess Championships in Stockholm in 1974. In 1976, Chess 4.5 took part in the famous Paul Masson tournament in Saratoga, California, beating all six human opponents and becoming the first computer to win a classical tournament. In 1977, Chess 4.6 also won the Computer Chess World Championship, and on the list of prominent humans defeated by the computer were ever more famous opponents.

The victories were, it must be admitted, in blitz or simultaneous games, for example in March 1977 against twelve opponents. While retired Master Edward Lasker classically complained he had lost due to "a mistake in a winning position", the President of the famous Manhattan Chess Club, Walter Goldwater, said about his loss: "I was outraged — for two days ... and the pain is still gnawing. It does not bother us that machines are better at swimming or flying than we are. But it bothers us when we see a machine writing better plays than Shakespeare or a better quartet than Beethoven. I believe it is because we distinguish between physical and mental abilities and creativity."

Do chess computers have creativity? David Levy: "In tactical situations the computer is in fact even more creative than humans; it does not stick to stereotypes and often finds unexpected and therefore very nice solutions." In fact, replaying computer games is more fun than many grandmaster games because of the original and often amusing variations.

When Chess 4.6, which had reached a level or skill that surpassed that of its creators Slate and Atkin, outwitted British Grandmaster Michael Stean in a tactical exchange of blows, the GM said: "This computer is a genius!"

Photo: Michael Stean in 1978 via Wikipedia

The genius, however, does not always play brilliantly. And Levy, who has already lost two quick games against Chess 4.6, "knows how the computer thinks." In short-term, tactical exchanges, which Stean had suicideally undertaken, this electronic thinker that is fast as lightning is able to look ten to twenty moves ahead while analysing all variations to calculate the best move with fatal certainty.

But in fact it does not get much further. In a serious game, both opponents have an average of three minutes of thinking time per move (two hours the for 40 moves). Although the computer can calculate 100,000 variations in this time, it can by no means look at all possible continuations, which in fact come to 25 times 10¹¹⁵. This number surpasses all astronomical numbers. The computer would require 10^90 years, billions of billion of years, per move. Of course, the slower-paced humans can work through even fewer variations than the computer in the same time, but humans possess additional qualities that computers, if at all, have to date only in rudimentary form: positional understanding, experience, chess instinct, concepts that make long-term strategy possible. Levy: "Against Chess 4.6, I can afford to remove a defending piece on the kingside, because 4.6 does not get the idea: Aha, he has weakened his kingside, so I can plan a king attack." The computer wastes a lot of time calculating nonsensical variations.

Despite these strategic weaknesses, Chess 4.6 is one of the 3000 strongest players in the world. This is thanks to its strength in short-term tactical exchanges, because except for great masters most players also lack the same strategic abilities and an infallible sense for long-term positional battles. Judging by its tournament successes, according to the international (ELO) rating system (where for example Bobby Fischer had 2800 points, a grandmaster has 2500-2700, International Masters up to 2300, and strong club players 1800) Chess 4.6 has reached 2270 points.

"In ten years it will beat Bobby Fischer"

Levy has 2325 points and a plan: "Basically, avoid tactical bantering. Build up the position carefully and calmly, but safely, and let it come. Because the computer can not develop a plan it will at some stage start moving its pawns and that, in the long run, will weaken certain important fields. Then you must use those weaknesses calmly and patiently." Of course that sounds easier than it is. Above all, Levy's opponent is not Chess 4.6. but the improved version Chess 5.0. How strong Chess 5.0 now what is, which improvements it will have compared to 4.6, Levy does not know. "In any case Chess 5.0 will be able to play even faster, and calculate several thousand variations more than 4.6. But all of its fundamental weaknesses cannot have been resolved."

Not yet. Levy would restrict a new bet against computers to two, at most three years. What will come after that no one knows exactly, "but I think it's possible that in ten years, when good chess strategists can program computers, they will be able to beat the world champion, even Bobby Fischer."

Apart from a World Champion computer Levy predicts: "In perhaps five years there will be, on the consumer market, chess computers with about the Chess 4.6 strength, but at prices and in formats similar to the current generation of small chess playing gadgets, which are of course still proper idiots!"

These gadgets are called "Chess Challenger" or "Compuchess" [photo Ismenio] or "Boris". As a chess partners they certainly have their advantages, at least in that they are never offended and are good losers. They admit defeat gracefully in their displays: "I loose", and if they win they immediately forget it. The most powerful of these gadgets Boris, which is currently only available in the American market, and which is already familiar with psychological tricks. Suddenly messages will appear in its display, like "Have you ever played before?" or "That's what I expected." It also displays compliments: "Good move," "Interesting," or "Ah, cruel!"

Cruel. Will in five years time a Chess 4.6-strong but only 500 Mark pocket computer spoil the joy of chess?

"Chess will be the game of the future"

When almost every chess player is able to afford a stronger partner, a partner who, once you have accepted its superior strength, spares the weaker human opponent the psycho-stress of a real game and also cures the problem of finding a partner to play against, will chess then lose its interpersonal communicative quality? Levy: "By no means. Once you get to know the computer you want to compete with people. Cars are faster than humans, and yet there are no less hundred metre races then before them."

Levy is convinced: "The computer will make chess the game of the future." He believes that simply more ("many fans can not find a partner today") and better chess will be played. "Computers are new chess players, and addition things that people can learn from." In other words, the best chess player will be the one who is the most diligent student on his computer.

Will players not take their electronic chess masters to tournaments, like students using crib sheets on the toilet, to find the best moves? Levy: "That would cause quite a crush at the toilets. One will have to do security checks at tournament entrances, like they do bomb checks at the airports."

Small chess sets are tested by ZEIT magazine reporter and chess fan Wolfram Runkel. There are two being offered on the German market: "Chess Challenger" for around 800 and "Compuchess" for around 600 Marks (at Friedrich C. Jensen, Spitalerstrasse 12, Hamburg 1). Both look similar to the calculators and work similarly. You enter moves by pressing keys (e.g. e2-e4), and after a while you will get the answer on the display window (e.g. e7-e5). They are always ready to play or to stop, but they are hardly able to win. Chess Challenger does not see any two-move mating threats, and one-move mates only if it is a capture move. It has three different levels for which it needs between 10 and 40 seconds per move. In the first I mated it in nine moves, on the third level in 14 moves.

"Compuchess" has six levels. The first is catastrophic (it keeps losing pieces), the fifth still very simple. It needs more than ten hours per move. I played the Blackmar Gambit, and let the clueless computer capture two pawns, then two bishops, a rook, and mated it in 15 moves (with the queen sacrifice it could have delayed the mate). I could not test the sixth level — the computer needs 24 hours per move. In that time it can a few 100,000 additional variations. But he still remains a short-sighted fellow, a weak training partner.