Ukuqonda kunye nokucubungula iireferensi zentetho ngumceli mngeni omkhulu kulwalathiso lwenkqubo yolwimi lwendalo kumxholo wabancedisi be-AI njenge-Amazon Alexa. Le ngxaki idla ngokubandakanya ukudibanisa ngokuchanekileyo izimelabizo kwimibuzo yabasebenzisi kunye neengqikelelo ezichaziweyo, umzekelo, ukuthelekisa isimelabizo esithi "bona" ββkwingxelo "dlala icwecwe labo lamva nje" kunye negcisa lomculo elithile. Iingcali ze-AI e-Amazon zisebenza ngenkuthalo kubuchwephesha obunokunceda i-AI iqhube izicelo ezinjalo ngohlaziyo oluzenzekelayo kunye nokutshintshwa. Ke, isicelo "Dlala icwecwe labo lamva nje" liza kutshintshwa ngokuzenzekelayo ngokuthi "Dlala icwecwe lamva nje le-Imagine Dragons." Kule meko, igama elifunekayo lokutshintshwa likhethwe ngokuhambelana nendlela enokwenzeka yokubalwa kusetyenziswa umatshini wokufunda.
Izazinzulu Isiphumo sokuqala somsebenzi wakhe kushicilelo lwangaphambili olunesihloko esinzima kakhulu- "Ukunyusa umkhondo wemeko yengxoxo ye-multi-domain kusetyenziswa uhlaziyo lwemibuzo." Kwixesha elizayo elingekude, kucetywa ukunikezela olu phando kwisebe laseMntla Melika loMbutho weeLwimi zeKhompyutha.
"Ngenxa yokuba i-injini yethu yokuhlaziya imibuzo isebenzisa imigaqo jikelele yokusebenzisa amakhonkco entetho, ayixhomekeke kulo naluphi na ulwazi oluthile malunga nesicelo apho iya kusetyenziswa khona, ngoko ayifuni ukuphinda iqeqeshwe xa siyisebenzisa ukwandisa amandla e-Alexa," kuchaza. Arit Gupta (Arit Gupta), ingcaphephe yeelwimi eAmazon Alexa AI. Uqaphele ukuba itekhnoloji yabo entsha, ebizwa ngokuba yi-CQR (umbuzo wokubhala kwakhona umxholo), ikhulula ngokupheleleyo ikhowudi yomncedisi welizwi kuyo nayiphi na inkxalabo malunga neembekiselo zentetho kwimibuzo.
Okokuqala, i-AI imisela umxholo jikelele wesicelo: loluphi ulwazi umsebenzisi afuna ukulifumana okanye isenzo amele asenze. Ngexesha lencoko yababini nomsebenzisi, i-AI ihlela amagama angundoqo, iwagcine kwizinto eziguquguqukayo ezikhethekileyo ukuze zisetyenziswe ngakumbi. Ukuba isicelo esilandelayo siqulethe nasiphi na ireferensi, i-AI iya kuzama ukuyibuyisela ngeyona ndlela inokwenzeka yamagama agciniweyo nafanelekileyo ngokwesemantiki, kwaye ukuba oku akukho kwinkumbulo, iya kuguqukela kwisichazi-magama sangaphakathi sawona maxabiso asetyenziswa rhoqo. , kwaye emva koko uphinde wakhe isicelo ngokutshintshwa okufakiweyo, ukugqithisela kumncedisi welizwi ukuze aphunyezwe.
Njengoko uGupta kunye noogxa besitsho, i-CQR isebenza njengomaleko wokuqhubekela phambili kwimiyalelo yelizwi kwaye igxile kuphela kwintsingiselo yesyntactic kunye nesemantic yamagama. Kuvavanyo lwedatha eqeqeshwe ngokukodwa, i-CQR iphucule ukuchaneka kombuzo nge-22% xa ikhonkco kumbuzo wangoku ibhekisa kwigama elisetyenziswe kwimpendulo yamva nje, kwaye nge-25% xa ikhonkco kwintetho yangoku ibhekisa kwigama. kwintetho yangaphambili.
umthombo: 3dnews.ru