Alan Turing was the pioneer in establishing a mathematical framework for understanding computers. He demonstrated that computers will always have certain limitations. For instance, a computer cannot consistently determine whether its program will eventually stop running. This realisation led to the understanding that some problems are inherently non-computable; they simply cannot be solved by a computer. Turing was instrumental in the design and programming of some of the earliest computers globally. Alongside his colleague David Champernowne, an economic statistician who provided crucial data to Churchill during the war, he also co-developed one of the very first chess-playing AI programs, named Turochamp.
Turochamp is widely recognised as one of the earliest computer games ever created. However, it had a significant drawback: it was too complicated for the computers of that era to operate. Even though it managed to play a complete game of chess against computer scientist Alick Glennie in 1952, it required Turing to act as the computer and manually execute the program’s functions, with each move taking 30 minutes to process.
Notably, Alick Glennie was the creator of Autocode, the first compiler that translated human-readable programming languages into machine code that the processor could execute.
Turochamp maintained a record of the current chessboard state and used the opponent’s moves as input. It would evaluate all possible moves it could make in response, then consider the responses the opponent could make, along with any significant moves such as capturing a piece worth less than the one that took it.
The program assigned scores to each possible move and used a minimax algorithm to determine the best move. This classic AI heuristic suggests that if you can score your own potential moves and those of your opponent while anticipating several moves ahead, you should choose the moves that yield the highest score for yourself, thereby maximising your chances of winning and minimising your opponent’s chances.
In my earlier blogs, I discussed the so-called credit assignment problem as it relates to biological brains and artificial intelligence. Interestingly, Turing and Champernowne did not mathematically implement this issue in the first chess-playing machine. You may remember another early artificial neural network called SNARC, created by Marvin Minsky. This basic machine was designed to navigate mazes, but its algorithm proved ineffective.
The credit assignment problem arises because we only know the outcome—winning or losing — at the end of a game, which can involve hundreds of moves. If we win, which specific move deserves the credit for being successful? Conversely, if we lose, which moves should be blamed for the defeat?
When Alan Turing played against his machine, he essentially cheated by manually guiding it through its moves. The minimax algorithm alone was inadequate for allowing a machine to learn how to improve its performance in a game.
In 1952, Turochamp was defeated by Glennie in twenty-nine moves. Sixty years later, while the game was being reviewed for a celebration event, the Alan Turing Centenary Conference, researchers unexpectedly found that Turing had cheated during the original match. When the revived program played a chess game against Garry Kasparov in 2012, it lost in just sixteen moves. Although disappointing, considering its minimal “thought process” that only looked a few moves ahead, this early AI performed quite admirably!
Alan Turing held firm beliefs regarding the potential future capabilities of computers. He foresaw that by the end of the 20th century, people would discuss machine thinking. He also anticipated that we would be able to program computers so that it would be difficult to distinguish between humans and machines in conversations. Turing asserted that an average interrogator would have only a 70 percent chance of accurately identifying whether they were conversing with a human or a machine after five minutes of questioning. He referred to this as the imitation game, which we now know as the Turing test. Additionally, the “Imitation Game” became the title of a biographical film about Turing released in 2014.
Turing passed away at the young age of forty-one in 1954. However, his contributions have had a profound impact and have been the subject of ongoing debate. As of November 2025, no AI has successfully passed a Turing test, despite numerous consultants, investors, and others who have received funding asserting that the test has been overcome. I will revisit these assertions later.