If you’re a human and you’ve never played ‘Super Mario World’ before, it would probably take you a while to get the hang of things. However, if you happen to be a computer, here’s a way for you learn the ropes and easily beat the game.
Seth Bling, who shot to viral fame for beating ‘Super Mario World’ in six minutes by figuring out a computer glitch, created a program where a computer learned to master ‘Super Mario World’ within one day.
Called MarI/O, the program knew absolutely nothing about ‘Super Mario World’ at the start, not even how to jump or figure out that you have to move right to play the game. However, Bling incorporated a process called neural evolution where a computer mimics how the brain works but at a much faster pace.
The computer used neural networks and genetic algorithms to learn the intricacies of ‘Super Mario World’ and how to quickly get through each level. Now, it did go through a trial and error process like most of us did when we first started playing at a friend’s house because our parents wouldn’t buy us a Super Nintendo. However this process was accelerated.
MarI/O doesn’t necessarily play for points and mainly just avoids monsters, but it still is pretty impressive nonetheless. Take a look at the video Bling made below, which further explains how the computer quickly put together a winning strategy.