An article by Martin Gardner in his Mathematical Games column in Scientific American is titled A Matchbox Game-learning Machine. Following is a quote from that article:
“There is every reason to expect that a chess-learning machine, after playing thousands of games with experts, will someday develop the skill of a master. It is even possible to program a chess machine to play continuously and furiously against itself. Its speed would enable it to acquire in a short time an experience far beyond that of any human player.”
This description is similar to how AlphaZero was programmed to play chess by competing against itself c.2017. However, the Scientific American article appeared in the March 1962 issue, more than 50 years before AlphaZero. I think it’s interesting that the idea behind programming AlphaZero was floating around so long beforehand.
The entire article in question may be found here. The article also appears in the book The Unexpected Hanging, and Other Mathematical Diversions.