- La IA Ubuchwepheshe bezokwelapha be-Microsoft, obubizwa nge-MAI Diagnostic Orchestrator (MAI-DxO), bufinyelela ukunemba okuphindwe kane kunodokotela abangabantu ekuxilongweni okuyinkimbinkimbi komtholampilo.
- Isekelwe ekuhlanganisweni kwamamodeli amaningana ahamba phambili ukuhlakanipha okufakelwayo, elingisa ukubambisana kwethimba lochwepheshe ukuze lihlaziye izimpawu futhi liphakamise ukuhlolwa ngokuhlanganyela.
- Uhlelo lweMicrosoft aluthuthukisi ukunemba kokuxilonga kuphela kodwa futhi lwehlisa izindleko ezihambisanayo cishe ngamaphesenti angama-20, nakuba lungakalungeli ukusetshenziswa komtholampilo ngobuningi.
- Ochwepheshe kanye neMicrosoft ngokwayo igcizelela ukuthi i-AI kufanele igcwalise, hhayi esikhundleni, umsebenzi wodokotela, igqamisa ukubaluleka kokwahlulela komuntu nokwethemba isiguli.
Insimu yobuhlakani bokwenziwa kwezokwelapha idlula ekuguqukeni kwangempela, futhi iMicrosoft isibe ngomunye wabagqugquzeli abakhulu balolu shintsho. Ucwaningo lwakamuva lwenkampani lubonisa ukuthi uhlelo lwayo, olwaziwa ngokuthi i-MAI Diagnostic Orchestrator (MAI-DxO), luyakwazi ukuxilonga izifo ngokunemba okuphindwe kane kunodokotela abangabantu lapho bebhekene namacala omtholampilo ayinkimbinkimbi.
Lokhu kuthuthuka kubeka inkampani yamazwe ngamazwe yaseMelika phambili ekusungulweni kwezempilo , kodwa ngesikhathi esifanayo kuvula izingxoxo mayelana nendima ezodlalwa ubuhlakani bokwenziwa emtholampilo wesikhathi esizayo nokuthi ingabeka kangakanani, noma kufanele, ukungenelela komuntu endimeni yesibili.
Isebenza kanjani i-MAI Diagnostic Orchestrator
Uhlelo oluthuthukiswe yiMicrosoft lusebenzisa indlela yokuhlanganisa ehlanganisa amamodeli amaningana ahamba phambili obuhlakani bokwenziwa , okuhlanganisa nalawo avela ku -OpenAI , Google , Meta, Anthropic, kanye ne-xAI. Empeleni, lusebenza njengohlobo "lwekomidi lezokwelapha elibonakalayo" lapho amamodeli ahlukene e-AI exoxa futhi ehlaziya izimpawu, ephakamisa futhi ekhetha izivivinyo zokuxilonga , futhi ekugcineni afinyelele esivumelwaneni ngesiphetho somtholampilo.
Ukuze ihlole ukusebenza kwayo, iMicrosoft isebenzise izifundo zezokwelapha zangempela ezingu-304 ezivela ku-New England Journal of Medicine . Ngayinye yalezi zaguqulwa yaba yinselele esebenzisana nayo elingisa inqubo evamile yokuxilonga: ukuhlonza izimpawu, uku-oda izivivinyo , nokwenza izinqumo njengoba icala liqhubeka , isinyathelo ngesinyathelo, njengoba nje udokotela onolwazi angenza ngesiguli sangempela.
Imiphumela ibimangalisa: I-AI ye-Microsoft ibilungile ezimweni ezingaphezu kuka-80% (ngemodeli ye-o3 ye-OpenAI, inani likhuphukela ku-85,5%), kuyilapho iqembu lodokotela abangabantu, abangenalo ukufinyelela ezinsizeni zangaphandle, belilungile cishe ku-20%. Ngakho-ke umehluko ekunembile mkhulu futhi udale injabulo enkulu phakathi kochwepheshe bezempilo nobuchwepheshe.
Ukwehliswa kwezindleko kanye nezindaba zokuziphatha
Esinye sezici ezibalulekile zesistimu yikhono layo lokunciphisa izindleko ezihambisana nenqubo yokuxilonga . Kuyo yonke izilingo, i-AI yakhetha ngokuzimela ukuhlolwa nezinqubo ezingabizi kakhulu ngaphandle kokubeka engcupheni ukunemba kokuxilonga , okwaholela ekongeni izindleko ezingaba ngu-20% uma kuqhathaniswa nezindlela ezivamile. UDominic King, iphini likamongameli we-Microsoft, wagcizelela ukuthi imodeli ayigqamile nje kuphela ngokunemba kwayo kodwa futhi nangokusebenzisa kwayo kahle izindleko ezinhlelweni zokunakekelwa kwempilo.
Kodwa-ke, ukusetshenziswa kobuhlakani bokwenziwa kwezokwelapha kuyaqhubeka nokubhekana nokungaqiniseki kokuziphatha kanye nobungcweti . Nakuba inkampani igcizelela ukuthi ayihlose ukuthatha indawo yodokotela , kodwa kunalokho ukugcwalisa umsebenzi wabo , ukuzenzekela kwemisebenzi yezokwelapha kuphakamisa ukukhathazeka , ikakhulukazi mayelana nokuphathwa kokungacaci, ukusebenzisana kwabantu neziguli, kanye nesibopho samaphutha okuxilonga. UMustafa Suleyman, umholi wephrojekthi, kanye nabanye ochwepheshe bayavuma ukuthi ubuhlakani bokwenziwa busekude kakhulu nokuphindaphinda uzwela kanye nokwahlulela okucacile kwezokwelapha okunikezwa ochwepheshe babantu.
Ngaphezu kwalokho, leli thuluzi alikakulungeli ukusetshenziswa kabanzi kwezokwelapha . I-Microsoft igcizelela ukuthi kuzodingeka isikhathi esengeziwe, ukuqinisekiswa okwengeziwe ezimweni zangempela, kanye nokubuyekezwa kwemithetho ngaphambi kokuba i-AI ihlanganiswe njalo ezibhedlela nasemitholampilo.
Ukusebenzisana kwezobuchwepheshe nezinyathelo ezilandelayo
Esinye sezici ezibalulekile ngemuva kokuthuthukiswa okusheshayo kwalesi sixazululo kube ukubambisana phakathi kweMicrosoft nezinye izinkampani ezinkulu zobuchwepheshe, njenge-OpenAI . Le nkampani iqashe abacwaningi abavela ku-Google futhi yaqinisa ubudlelwano nezinhlangano eziholayo kwezobunhloli bokwenziwa, okubonisa ukuncintisana okukhulayo kwethalente kulo mkhakha. UMustafa Suleyman, ohola lo msebenzi kwaMicrosoft, ngaphambilini wayengumsunguli we-DeepMind kanye nenhloko ye-AI kwa-Google.
Ukubambisana phakathi kwamamodeli ahlukahlukene e-AI asetshenziselwa ukuxilongwa kulingisa impikiswano phakathi kochwepheshe abavela kochwepheshe abahlukene, okuvumela ubuhlakani bokwenziwa ukuthi bucabangele izindlela ezahlukahlukene futhi bungene emikhakheni yezokwelapha ebingadinga ukubonisana nochwepheshe abaningana babantu. Umthelela we-AI emkhakheni webhizinisi ubonakala nasekunakekelweni kwezempilo.
Okwamanje, inkampani ayikakanqumi ukuthi izoyithengisa ngqo noma iyihlanganise kumapulatifomu afana ne-Bing ukuze kube lula ukuxilongwa kwabasebenzisi noma isize odokotela neziguli ukuthi benze izinqumo ezinolwazi oluthe xaxa. Kunoma ikuphi, amandla obuhlakani bokwenziwa okuguqula ukunakekelwa kwezempilo awanakuphikwa, kodwa ukusetshenziswa kwayo kuzoqhubeka kancane kancane futhi kuhambisana nokuqapha maqondana nokuphepha, ukuziphatha, kanye nomthelela emphakathini.
Umbono wabasebenzi bezempilo kanye nezinselele zesikhathi esizayo
Ukusabela emphakathini wezokwelapha kuye kwaxubana . Ngenkathi abanye ososayensi nodokotela bekwazisa ukuqina kwezivivinyo kanye nethuba lokunciphisa ingcindezi ezinhlelweni zezempilo, abanye bakhomba ukuthi i-AI akufanele isetshenziselwe ukuthatha indawo yobudlelwano phakathi kodokotela nesiguli , ingasaphathwa eyokwenza izinqumo eziyinkimbinkimbi ngaphandle kokuqondisa komuntu.
Phakathi kwezinzuzo ezingaba khona zokusebenzisa izinhlelo ezifana ne-MAI-DxO ukuncishiswa kwezivivinyo ezingadingekile , ukuphathwa kwezinsizakusebenza okuthuthukisiwe , kanye nekhono lokunikeza ukuxilongwa okunembile ezindaweni ezingenabasebenzi abanele . Kodwa-ke, kusalokhu kukhathazekile mayelana nokufinyelela ngokulinganayo kula mathuluzi , ukucwasa okungenzeka kudatha esetshenzisiwe , kanye nokuqeqeshwa kobuchwepheshe kwabasebenzi bezempilo ukuqinisekisa ukuthi akekho oshiywa ngemuva lapho kuqaliswa amathuluzi amasha.
I-Microsoft kanye nochwepheshe abahlukahlukene bagcizelela isidingo sokuqhuba izifundo ezengeziwe ezimweni zempilo yangempela, ukusebenzisa ukuqeqeshwa ngezimpawu ezivame kakhulu, futhi bahlole ukuhlanganiswa okunesibopho kobuhlakani bokwenziwa nolwazi lomuntu, ukuze ubuchwepheshe busebenze njengozakwethu, hhayi usongo, enhlalakahleni yesiguli.
Ukuthuthuka kwe-Microsoft kubuhlakani bokwenziwa obusetshenziswa ekunakekelweni kwezempilo kuphawula ingqopha-mlando ekuthuthukisweni kwemithi yedijithali.
Umbhali oshisekayo ngomhlaba wamabhayithi nobuchwepheshe ngokujwayelekile. Ngiyathanda ukwabelana ngolwazi lwami ngokubhala, futhi yilokho engizokwenza kule bhulogi, ngikubonise zonke izinto ezithakazelisayo kakhulu ngamagajethi, isofthiwe, ihadiwe, izitayela zobuchwepheshe, nokuningi. Inhloso yami ukukusiza ukuthi uzulazule emhlabeni wedijithali ngendlela elula nejabulisayo.