Anyone who plays video games know that game bot , by artificial means intelligent virtual gamers , can be spot a mile away on account of their unmindful predictability and utter lack of behavioural pragmatism . Looking to change this,2 K Gamesrecently launch theBotPrize competition , a kind of Turing trial run for nonplayer characters ( Nonproliferation Center ) . And remarkably , this yr ’s carbon monoxide gas - winner , a team from The University of Texas at Austin , make a nonplayer character ( NPC ) that was so realistic that it appeared to be more human than the human actor — which is kind of a problem when you think about it .
Neuroevolution
To produce their super - realistic game bots , the software developer , a squad precede by Risto Miikkulainen , programmed their NPCs with pre - existent models of human behavior and fed them through a Darwinian weeding - out process call neuroevolution . Essentially , the only bots that live into sequent propagation were the ones that appeared to be the most human — what the developers and rival organizer compare to passing the classic Turing Test ( an attempt to distinguish AIs from actual humans ) .
With each passing coevals , the developers re - inserted precise copy of the surviving NPCs , along with slightly modified ( or mutated ) versions , thus take into account for on-going variation and excerpt . The simulation was run over and over again until the developers were satisfied that their game bot had evolved the desired characteristics and behaviour . And in fact , Miikkulainen and his squad have been refining their virtual player over the past five years .
humanity
The final manifestation of their try was dub UT^2 — and it was this NPC that die head - to - heading against human adversary and other game bot at the 2 K Games tournament .
And the game of alternative ? Unreal Tournament 2004 , of course . The secret plan was selected on account of its complex gameplay and 3D environments — a challenge that would require mankind and bots to move around in 3D space , engage in helter-skelter combat against multiple opponents , and reason about the beneficial strategy at crucial moments . Moreover , the biz is also capable of bring about some telltale human behaviors , including unreason , anger , and impulsivity .
As each histrion ( human or otherwise ) work to eliminate their opponents , they were afterwards assessed for their “ humanness . ” By the end of the tournament , there were two clear succeeder , UT^2 and MirrorBot ( developed by Romanian computer scientist Mihai Polceanu ) . Both NPCs nock a humanity rating of 52 % , which is all okay and well except for the fact that the human players score only 40 % .
In other words , the secret plan bots appear to be more human than human .
Limits of the Turing Test
Now , this is a serious trouble . Human instrumentalist should have been assessed with a humanness rating of 100 % , not 40 % . Clearly , the judges utterly failed to describe true human characteristics among the human players . So by upshot , UT^2 and MirrorBot essentially achieved a rating well than 100 % — which is impossible . How can something be more than something you ’re trying to emulate ?
And indeed , this experimentation is a good display case for the limits of the Turing Test . avowedly , the 2 K Games tournament was n’t meant to be a true Turing Test , but one that measure the humanness of NPC in a very specific gaming setting . That said , the results demonstrated that human behavior is much more complex and hard to quantify than we tend to retrieve . Human idiosyncrasies , plus the ability to adapt and counter - adapt to attempts to name it , will likely forever put it beyond the reaching of a simple Turing Test .
For exemplar , given the entailment of the 2 K Games tournament , how are we think to measure something like a chatbot for its humanness now that we know something can seemingly appear to be more human - like than humans ? Moreover , given all the subjectivity regard in the rating , how accurate is any of this ?
Perhaps its metre to retire the Turing Test and come up with something a bit more….scientific .
Top imagevia . Inset image via Jacob Schrum / University of Texas at Austin .
FuturismRoboticsRobotsTechnologytelepresence
Daily Newsletter
Get the best tech , science , and culture news in your inbox daily .
News from the future , delivered to your present .
You May Also Like
