Yebo, i-COVID-19 yenze unyaka ka-2020 waba iphupho elibi lokuphila futhi yebo, uselapha… sicabanga - Ingabe kunjalo? Cha? Yebo? Nokho, umkhuhlane awusekho. Usuyazi ukuthi abanye abantu abalinganiselwa kanjani (okusho UKUPHILA), okusetshenziswa njengendlela eyengeziwe yokukugcina unovalo, kepha ngiyaphuma. Manje, ukukhwehlela kwakho okunempilo (kungasolwa) kungembula uma une-virus ye-haulin 'bits engazazi noma cha.
I-Magic AI 8-Ball Ithi… Wena Unezifo
Abaphenyi beMIT uJordi Laguarta, uFerran Hueto, noBrian Subirana banakho ithuthukise imodeli yobuhlakani bokufakelwa leyo yakhelwe ukuhlukanisa abantu abangabonakali nge-COVID-19 kokuqoshwa komsindo. Ngokuvumela i-AI ihlaziye izinkulungwane zokukhwehlela nezindlela zokukhuluma ezahlukahlukene, iqembu likwazile ukulilinganisela ukuze libone ama-98.5% ama-asymptomatics angakhombisanga zimpawu zaleli gciwane kepha ahlola ukuthi linalo. Kungenzeka yini ukuthi babe ngamasu amahle? Kwazi bani ngempela.
Yakha emsebenzini amaqembu amaningi ocwaningo ayewenza ngisho nangaphambi kokuba ubhadane luhlasele; kuphela esikhundleni sokuhlolela i-COVID-19, abantu bebefuna ukuthola iziguli ezinesifuba somoya, inyumoniya, kanye ne-Alzheimer's (izintambo zezwi ezibuthakathaka ezingaba uphawu lwesifo esiguga).
Ngokusho the indawo esemthethweni yezindaba ze-MIT, iqembu lihlanganise amanethiwekhi amathathu we-neural:
Inethiwekhi yokuqala ye-neural iyi-algorithm ejwayelekile yokufunda imishini yokufunda ebizwa nge-ResNet50, ehlelelwe ukuhlukanisa amaphethini enkulumo. Imisindo efana no “mmm” noma “err” ingasho ukuthi ziqine kangakanani noma zibuthakathaka kanjani izintambo zomlomo zomuntu, futhi lokhu kungakhombisa izimpawu zokuqala umuntu ahlangabezana nobunzima bokukhuluma ngenxa yokutheleleka. Noma ingabe i-phlegm nje inamathele emphinjeni wakho noma ukuthi unamabhomu amaningi weSaki? Okungabalulekile.
Inethiwekhi yesibili ye-neural igxila ekutheni abantu bayidlulisa kanjani inkulumo yabo. Imizwelo ibamba iqhaza elikhulu endleleni esikhuluma ngayo, nakule nethiwekhi izama ukuhlukanisa ukuthi uzizwa kanjani lapho ukhuluma. Iminjunju ejwayelekile yokukhungatheka futhi ngaphandle kwemizwelo ngenkathi ukhuluma kungaba yizimpawu ezikhombisa ukuthi umuntu uphethwe yi-COVID-19, futhi le nethiwekhi yesibili yenziwa ukuhogela lokhu. Ngabe ikhona eminye imizwa engaba khona lapha? I-Psshh - Ngiyacela.
Okokugcina, inethiwekhi yesithathu ye-neural iqeqeshwe ku-database ebanzi yokukhwehlela. Ukukhwehlela okukhulu, ukukhwehlela okuncane, ukukhwehlela okuqinile, nokukhwehlela okucindezelweyo - konke lokhu nokuningi kuhlukaniswa ngaphansi kwale nethiwekhi lapho kukhomba khona uma kukhona izinguquko ezibonakalayo ohlelweni lwakho lokuphefumula. Ahhh, impilo nezikhathi zokukhwehlela.
Lezi zinhlelo ezintathu ze-neural, ezihlanganiswe ne-algorithm ethola ukonakala kwemisipha, zakha i-AI yokugcina eyakhelwe ngqo ukuthola abantu abangenazimpawu okungenzeka baphethe i-COVID-19. Ayikwazi ukuxilonga labo asebavele banezimpawu zalesi sifo, kepha ingahlukanisa ukukhwehlela okungaqondakali kokukhwehlela okuvamile okunempilo, ngakho-ke kufanele samukele ukuthi kunembile ngokuphelele.
Isinyathelo esilandelayo seqembu le-MIT manje ukufaka i-AI yabo kuhlelo lokusebenza olusebenziseka kalula olungabiwa ngezinga elikhulu ngokwanele. Icebo ukuhlinzeka ngokuhlolwa kwamahhala kwabantu okungenzeka basolwe ukuthi baphethe i-COVID-19 kepha akubonisi zimpawu zomzimba. Ukuphela kokufanele ukwenze ukulanda uhlelo lokusebenza kufoni yakho, ukukhwehlela kwisipikha, bese uqinisekisa ukuthi ungahle ube yi-asymptomatic noma cha.
Ngubani okulungele ukunikela ngempilo yakhe ku-AI nakuhlelo lokusebenza? Ukuphakamisa izandla? Noma ubani? Noma ubani? "Khwehlelani ku-ze app noma sekuyisikhathi sokubuyiselwa isikhathi kini sonke !!"
Ungathola iphepha likaLaguarta, uHueto, noSubirana ekukhwehleleni kwabo nasenkulumweni yabo ehlaziya i-AI ku- Ikhasi lewebhu le-Institute of Electrical and Electronics Engineers (IEEE) Engineering in Medicine and Biology Society (EMBS).
