Who says disgrace ca n’t be an effective motivator ? Less than a workweek after we sharedWesley Liao ’s experimentsusing machine learning to train an AI to play QWOP , one of the hardest television games of all time , the AI wasre - train with the goal of maximizing its speed , leave in a new world record .
Starting with their previous AI agent named ACER that was trained with a focus on optimum running technique and form , Liao trained a new agentive role with a modify reward system . Previously , behaviour like “ low trunk height , vertical torso motion , and excessive articulatio genus bending ” were discouraged to aid ACER find out a proper stride technique .
https://gizmodo.com/an-ai-was-taught-to-play-the-worlds-hardest-video-game-1846388137
Gif:YouTube - Wesley Liao
But since the new AI federal agent was learning from ACER that had already mastered its stride , the machine learning outgrowth or else solely focused on rewarding improvement made to the sprinter ’s forward velocity . away from a couple of minute of “ pre - training , ” the new AI require just 40 hours of training tofinally beat the dear human QWOP players .
A website calledSpeedrun.comis where you ’ll line up the actively update leaderboard for theQWOP 100 meter - dash , and while the top human player ( Japan’sgunmaneko ) managed to get their sprinter across the ending line in 48.34 secondment , the well register run of Liao ’s fresh groom AI did it in 47.34 seconds . But do n’t expect to see Liao ’s name atop the QWOP leaderboard . Speedrunning is still a contender for human players only and the role of computer software peter , such as an AI , to aid a campaign is strictly forestall .
Do we need separate speedrunning leaderboards for AI players ? Sure , why not ? There ’s good reason to keep a conservative eye on the unbelievable furtherance we ’ve made with artificial intelligence operation , but it ’s also just plain fascinating to see how quickly they can be trained to best a human competitor . Despite being so challenging , QWOP is a very rudimentary video game that focuses on the accurate timing of push button press . It would also be interesting to see an AI tackle a plot like The Legend of Zelda series where fundamental interaction with other AI - powered characters add up into play . In the summons , an AI factor like Liao ’s may even obtain shortcuts , proficiency , or gameplay strategy that could help human speedrunners too . In the lag can we at least get this QWOP - playing AI a participation prize ?
SportsVideo gaming
Daily Newsletter
Get the best technical school , science , and culture news in your inbox day by day .
newsworthiness from the future , delivered to your present .
You May Also Like
