The father of machine learning >

Long before the machine learning advances of today, there was the pioneering work of Arthur

Samuel. Starting in 1949 while at IBM, Samuel began working on a computer that could teach itself to improve its ability to play checkers.

Given the tiny amount of computer memory available at the time, Samuel had to develop a novel system of learning. Instead of the computer analysing every possible future move and its relative value until the conclusion of the game, it would only take into account the pieces on the board at any given time. It would then use this to determine the best chance of maximising its own success and minimising that of its opponent.

At the same time, the computer would remember every move it had ever seen and the value of the reward each move gave it. It would then use this knowledge to improve its future gameplay. Eventually, the computer was able to play at the level of a respectable amateur.