We'relookingatinstanceswhere a computerizedopponentbested a humanbeingor a teamofpeopleat a competitivecontestorgame.
Therobotsinquestiondon't needtohavecompletelyswept a seriesofgamesagainsthumans, butthereneedstohavebeenatleastsomesortofdefinitivewinningstreakintheirfavor.
Octoberof 2016 markedthefirsttimewhenMicrosoftAiAIprogramequaledtheerrorratioof a professionalteamofhumantranscriptionexperts.
Weshouldmentionthattheprogram's successratewasonlyslightlybetterthanitshumancounterparts, withtheaveragebeingmoreonthelevelofhumanparodythananysortof a blowout.
Still, thisayeayesuccessstoryservedas a breakthroughwhenitcametoimprovinghowtoolslikeourmobilephonesunderstanduson a dailybasis.
EnterGoogle's Deepmindtechnologyanditsabsolutedominanceover a scoreoficonicAtarifranchises.
A 2015 reportboastedhowDeepMindcleaneduponeventhehardestAtarititles, oftentimesperformingat a levelsuperiortonearly 3/4 ofprofessionalhumangametesters.
I want a hardfigureforproof.
Deepmindscored a whopping 999,990 pointsatMsPacManbeatthatone.
NumbersevenRubik's CubeRobotssolving a Rubik's Cubecantakepeoplehours.
Thehumanrecord, however, actuallystandsatjust 3.47 seconds, heldbyChina's DoYouShun.
Well, theOpenaye, ayefiveProjectansweredthatquestionwiththeirfinalpublicdemonstrationin 2019 whentheyabsolutelyslaughtered a humanteamonthebattlefieldofDotatwo, alsoknownasdefenseoftheAncients.
Thecreepsareon.
TheagentDJiscalledonOpen A I takeGameone.
Thehumanexpertsneverhad a chanceasopen A.
I wonmatchaftermatchwithrelativeease.
Sodefinitivewastheprogram's winthatthedevelopmentteamdecidedtochangetheirfocuson a humancooperation, presumablyknowingfullwellthatnorealcompetitionlayouttherefortheirvictoriousboughtmonsters.