Artificial Intelligence Tries the QWOP Running Simulator

QWOP, arguably the most difficult video game ever, was foisted upon us by Bennett Foddy in 2008. The point of the game is to control a sprinter with four keys that move his thigh and calf muscles. A few seconds in, and you’re reduced to trying to control your laughter. Most of us gave up pretty early and decided the value in the game was purely comedy. However, in the years since, some gamers took QWOP as a challenge, and became fairly good at getting the runner through a 100-meter dash. The QWOP record as of now is 48 seconds! Could an algorithm trained to play QWOP do any better? Wesley Liao trained one to find out. While his AI learned to make the QWOP runner perform better than I ever will, it could not outdo the best human gamers. Read more on the experiment at Gizmodo.