Thoughtwave was the second of the 10 games I have had marketed. Sigma File was the first. The game was inspired by Piet Hein's game Hex but I wanted to introduce structure to the elements placed on the board.

1973: Invented, tested, and rapidly submitted to Intellect UK who had offices in Marlborough Street above Polly Peck. The name I first chose was Way-Lay (`ways' are laid, and players waylay their opponent in a sense). The pieces had no wave pattern and the board was a beautiful 18 inch square of black perspex which I still have, though the perspex pieces have been cannibalised for other games. The game was accepted for marketing almost straight away. Intellect thought up the name and introduced the wavy pieces.

1975: I think the game was launched at the Earls Court Toy Fair in January. A review appeared in Games and Puzzles magazine in May (Issue 36), and an article written by me followed in July (Issue 38). Incidentally, the review (in which I had no part) awarded the game the overall rating of 6 out of 6. 1978: Probably the year in which Parkers marketed the game in Germany under the name Ultra.

1979: I think this is when Games Centre commissioned a magnetic version from Stratabord. Games Centre was liquidated shortly afterwards and the production had a very short life.

The very first working game-playing machine was the King-Rook-King Chess end game solver, created by Leonardo Torres y Quevedo in 1890. It played the side of the king and the rook whist a human player attempted to defend with just a king. The next was the "Nimotron" created by Codon et al., in 1940, which would play a perfect game of "Nim", although it is significant to both these events that both these mini-games have solutions which can be memorised or solved algorithmically by human players.

The first working chess playing program was written by Alan Turing in 1951, but it was never run on a computer, instead it was tested by hand against a very poor human, and lost.

Its it thought that a program that could play Draughts (aka Checkers) was written as early as 1947, but research wasn’t published until 1952. Schaeffer’s "Chinook" took the world championship from the champion of 40 years, Marion Tinsley’s, in 1994.

In 1980 The Fredkin Prize was established. It offered:

$5000 to the first Chess program to achieve Master rating,

$10,00 to the first program to a USCF rating of 2500, and

$100,000 to the to defeat the human world champion.

The first two prizes where claimed by Belle in 1982 and Deep Thought in 1989 respectively. The $100,000 prize remains unclaimed, after the convincing victory’s of Gary Kasparov in extended play with Deep Thought.

(Note: This is currently in dispute, since Deep Thought beat Kasparov using tactics modified by the programmers during play)

Computer AI Chess Rankings

1965 MacHack 1400

1968 Chess 3.0 1500

1975 Chess 4.6 1900

1982 Belle 2200

1988 Deep Thought 2551

1994 Deep Thought 2 2600

And for reference, Human Grand Master

1992 Kasparov 2805

A human beginner is given an initial ranking of 1000.

Deep Blue, the planned successor to Deep Thought, will use 1024 VLSI chips in parallel, and will search 1 billion moves a second, allowing it to achieve a look ahead of 14 moves for each move.

Othello is a very popular choice for keen young AI programmers, because it only has a small search space compared to chess, with only 5-15 different moves available at any one time during normal play, compared to Chess’ 22 or more. Human champions normally refuse to play computers in tournaments.

Go is a very popular challenge for AI programmers who like their work cut out for them, and is as, if not more, popular in Japan as Chess is in the western world. It has up to 360 different moves available at any one time, so normal search methods such as the ones normally used for chess, draughts and othello are next to useless on current hardware. Knowledge based systems are currently in development, but currently play very poorly. Research in this area is very competitive, possibly due to the $2,000,000 prize offered to the first program to beat a top player.

As far as I know I am the only person to have investigated developing AI for Thoughtwave, which is unsurprising given its insane opening branch factor of 1500. Despite this I have developed an AI using specialised search algorithms based on A*, and a progressive route invalidation system that allows the AI to retain portions of the search tree created in previous turns. At the moment the AI plays at beginner level, and can play its self from start to finish in approximately 4 minutes on a standard Intel PII266 with 64MB of RAM under Windows 95. I plan to add special case opening moves, blocking tactics and a learning system based on Genetic Algorithms.