What’s the hardest video game you’ve ever played? If not QWOP then let me let you know you don’t know how difficult a game can be. The deceptively simple racing game is so hard to master that even one AI trained through machine learning still only gathered a top 10 score instead of shattering the record.
If you have never done so played QWOP before, you owe it to yourself try it and see if you can even get your sprinter off the starting line. Developed by Bennett Foddy in 2008, QWOP was inspired by an 80s arcade game called Track & Field this requires players a crushing buttons to win a race. QWOP it takes a different approach and instead uses players four keys to control the individual movements of a thighs and runner calves—A runner who behaves like a rag doll and is subject tophysics of the world, including the effects of gravity. It may sound simple, but mastering the timing and cadence of the keystrokes needed to get the sprinter to move awkwardly can be incredibly frustrating.
Wesley Liao was curious to know if he has such a tool as AI i have been trained to do things like encourage in a realistic way old photos of the deceased loved ones, he would play QWOP. After first creating a Javascript adapter that allowed an artificial intelligence tool to play and interact with the game, Liao’s first attempt at machine learning simply made AI play the game on its own and learn which actions gave positive results (the sprinter was advancing and increasing their speed) and which resulted in negative results (flexion of the torso of the sprinter too close to the ground.) Through this approach, AI learned a “knee brush” technique that would make it successful through the 100-goal, but not at record speeds.
Liao’s next attempt to train an AI model was to record videos about the game that were trying to succeed in the game, including the use of longer leg steps, which are crucial to increase speed and cross the line. goal with a decent time. The approach was a little more successful, but the AI was not able to master a special technique used by Advanced QWOP players involving an up and forward swing of the legs to generate additional momentum.
G / O Media may receive a commission
Finally, Liao contacted a veteran player known as Kurodo (@cld_el on Twitter), one of the first places QWOP world speed racers, who recorded 50 videos of themselves playing at an expert level. But even with access to the best possible game techniques, Liao found that the best results came from a machine learning training regimen that involved 25 hours of AI playing alone, 15 hours learning to from data obtained from Kurodo’s expert tests and another 25 hours of the game itself.
But even with all that effort, the QWOP-play the best 100 AI-the result of the meter test made him cross the finish line in 1 minute and 8 seconds—a top 10 finish. In accordance with Speedrun.com, the current world record for 100-meter dashboard is just 48 seconds, set just a month ago. Liao is confident of having more training and a different reward system (as AI learns he has done something right), establishing a QWOP the world record could finally be produced, although, since it is a computer that plays the game, the record may never be officially recognized.