Learn the game of "Spicy Marshmallow" here: https://scratch.mit.edu/projects/379897022 In this project, instead of playing against a random player, you'll play against a 'machine' made of 12 (gray) numbered cups. On the machine's turn, it will use the cup with the number corresponding to the quantity of things currently in the game bowl. A random color is selected from that cup to decide the machine's choice -- Red for 1, Green for 2, Blue for 3. When the machine begins playing games against you, all cups are started with an equal distribution of the colors: 3 red, 3 green, 3 blue. Reinforcement Machine Learning occurs at the end of each game through encouraging good choices by adding more of the colors to cups that contributed to the machine winning that game, or discouraging bad choices by discarding colors from cups that resulted in losing. Play lots of games and you'll see the machine *learn* to play better and eventually win more games than you!
This was inspired by Dr. Donald Michie's M.E.N.A.C.E.: https://archive.org/details/DonaldMichieGameMechinizationExperiments/ The game of "Last One Loses" is studied by Dr. H.D. Block in https://www.jstor.org/stable/27835902
