What’s the hardest video game you’ve ever played? If it wasn’t QWOP then let me tell you that, know that you don’t know how difficult a game can be. The deceptively simple running game is so challenging to master that even one AI trained with machine learning still only collected a top 10 score instead of shattering the record.
If you’ve never done that before played QWOP before, you owe it to yourself give it a try and see if you can even get your sprinter off the starting line. Developed by Bennett Foddy back in 2008, QWOP was inspired by an arcade game from the 80’s called Track & Field For that you need players to mindlessly stamping buttons to win a race. QWOP takes a different approach and lets players use it instead four keys to control the individual movements of a runner’s thighs and calves-a runner who acts like a floppy rag doll and is subjected to real-world physics, including the effects of gravity. It may sound simple, but mastering the timing and cadence of the keystrokes needed to just keep the sprinter moving awkwardly can be incredibly frustrating.
Wesley Liao was curious how good a tool like AI has is trained to do things like realistically animate old pictures of deceased loved ones QWOPAfter first creating a Javascript adapter that allowed an AI tool to actually play the game and interact with it, Liao’s first foray into machine learning had simply let the AI play the game itself and learned which actions resulted in positive outcomes (the sprinter went ahead and increasing the speed) and which resulted in negative results (bending of the sprinter’s hull too close to the ground.) This approach taught the AI a ‘knee-scrape’ technique that successfully passed itmeter finish line, but not at record speeds.
Liao’s next attempt at training an AI model involved recording gameplay videos of those trying to succeed in the game, including using longer leg steps critical for increasing speeds and crossing the finish line with a fair amount of time. The approach was slightly more successful, but the AI was unable to master a special technique used by advanced users QWOP players that involve an upward, forward swing of the legs to generate extra momentum.
G / O Media can receive a commission
Finally, Liao contacted an experienced player known as Kurodo (@cld_el on Twitter), one of the top QWOP speedrunners in the world, who recorded 50 videos of themselves playing the game at an expert level. But even with access to the best possible playing techniques, Liao found that the best results came from a machine learning training program that involved 25 hours of AI playing on its own, 15 hours of learning from the data collected from Kurodo’s expert runs, and another 25 hours. of self-play.
But even with all that effort, the QWOPAI’s best 100 gamesmeter line result talked about the finish line in 1 minute and 8 seconds-a top 10 finish. According to Speedrun.comthe current world record of 100 meters is set just 48 seconds ago, just a month ago. Liao is sure of more training and a different reward system (how the AI learns it did something correctly), set one QWOP The world record could eventually happen, although since it’s a computer playing the game, the record may never be officially recognized.