Indlela i-AI esebenzisa ngayo umbala wolwimi ukuqikelela izifo ngokuchaneka okuphezulu

Umzi-mveliso wokhathalelo lwempilo uhlala uthatha iminyaka ukwamkela itekhnoloji entsha kuba kufuneka utsibe kwiihops ezininzi zolawulo. Noko ke, indlela abasabela ngayo kubukrelekrele bokwenziwa yahlukile. Iingcali ezingenakubalwa sele zizama uqikelelo lwe-AI kunye nesakhono sokuhlalutya ukulungelelanisa uxilongo.

Abaphandi kutsha nje babanga ukuba i-AI inokuchonga uhlobo lwesifo umntu anaso-kwaye ukuba uqhubele phambili kangakanani-ngokujonga nje emlonyeni wabo. Le teknoloji iye yahamba indlela ende, kodwa ngaba inokuqikelela izifo zisebenzisa imibala yolwimi kuphela ukuxilongwa?

Uphando lungqina ukuba i-AI ingasebenzisa Umbala wolwimi ukuxilonga

Abaphandi abavela kwiYunivesithi yaseMbindi yobuGcisa eBaghdad, e-Iraq, nakwiYunivesithi yaseMzantsi Australia e-Adelaide, e-Australia, kutshanje ifumene iteknoloji ye-AI inokuhlalutya imibala yolwimi ukuze ixilongwe. Baphuhlise inkqubo yokubona ikhompyutha eyenza kwaye ihlukanise imifanekiso isebenzisa imifuziselo yendawo enemibala, ebonelela ngamaxabiso anokulinganiswa kwiimibala kunye nokukhanya.

Basebenzisa amawaka eemifanekiso zoqeqesho kunye novavanyo, uninzi lwazo zivela kwiSibhedlele sokufundisa i-Al-Hussein e-Iraq kunye neMosul General Hospital eMosul. Baqeqeshe imodeli kubantu bokwenyani abanezifo zokwenyani, hayi iseti yedatha eyenziweyo. Umahluko onje ubalulekile xa uphuhlisa isixhobo seenjongo zokuxilonga.

Abaphandi bahlela imifanekiso kwiindidi ezipinki, ezimhlophe, ezibomvu, eziphuzi, eziluhlaza, eziluhlaza okanye ezingwevu ukuze iimodeli zabo zichonge imibala phantsi kwayo nayiphi na imeko yokukhanya. Baqeqeshe abasixhenxe bebonke. Eyona iphezulu kakhulu yakhiwe nge-Extreme Gradient Boost (XGBoost) - ithala leencwadi elivulekileyo lokufunda koomatshini - apho ifunyenwe ngaphezulu kwe-98% yokuchaneka kwi-avareji. Ngexesha lovavanyo, yafumanisa ngokuchanekileyo i-58 kwimifanekiso ye-60, oku kuthetha ukuba ichanekile i-96.6% yexesha.

Iziphumo zabo zandothusa. Ngeliphandle, bendingalindelanga ukuba inkqubo yabo igqwese iingcali zonyango. Nangona amaqela ophando ahlukeneyo aphuhlise imizekelo efanayo yokuxilonga kwezinye iinjongo, bambalwa abachanekileyo. Uphando lubonisa ukuba noogqirha abaqeqeshiweyo abanamava eminyaka bayifumana ngokufanelekileyo 71.4% yexesha kwi-avareji.

Yintoni etyhilwa ngumbala wolwimi ngempilo yomntu

Ndiya kuvuma, ndacinga ukujonga ulwimi ukuqikelela isifo esingaqhelekanga. Kwakubonakala ngathi yindlela abantu ababeza ngayo ngaphambi kokuba amayeza anamhlanje abekho. Ukuthetha kakuhle, ndandichanile. Ukusebenzisa imibala yolwimi ukuxilongwa kusekwe kunyango lwaseTshayina oneminyaka engama-2,000 ubudala.

Kuzo zonke iimpawu zolwimi, kuquka ukumila, ukumila nokufuma, umbala sesona salathisi sibalulekileyo yezempilo. Ukuba ndithe ndakhupha ulwimi lwam ndajonga esipilini ngoku, ndilindele ukuba kube pink. Nawuphi na omnye umbala unokubonisa ukuba kukho into engalunganga emlonyeni wam, kwinkqubo yokujikeleza kwegazi okanye kumalungu.

Ububomvu bunokuthetha ukuba ndinomkhuhlane ophezulu ngokungaqhelekanga okanye ukunqongophala kwevithamin. Uphando lubonisa ukuba kukho a ikhonkco enokubakho phakathi seswekile nokuba tyheli kolwimi. I-tinge eluhlaza ngokuqhelekileyo ibonisa ukusuleleka kwefungal okanye i-bacterial buildup. Ukuguquka kombala oluhlaza okwesibhakabhaka kunokuba luphawu lwenqanaba eliphantsi le-oksijini yegazi okanye isifo semithambo yegazi. I-Grey inokuba nantoni na ukusuka kwi-fungus ukuya kumhlaza.

Ngelixa iingcali zonyango zineenkqubo ezininzi zokuxilonga, uninzi lusajonga iilwimi kuba zimbalwa iimeko ezichaphazela umbala walo. Ukuba kukho umbala obonakalayo, banokucutha ingcambu yengxaki. Xa kuthelekiswa novavanyo lwegazi olunika ufundo oluchanekileyo kodwa akukho zimpendulo ziqinisekileyo, isoloko ilukhetho olungcono.

Oko kwathiwa, iimpazamo ezenziwa ngabantu zihlala zinciphisa ukuchana kwazo. Ngokwesithethe, oogqirha bazihlola ngesandla iilwimi zabaguli. Nangona iminyaka yamava, nantoni na ukusuka kwizibane ezinemibala engaphezulu ukuya kwixesha lemini inokuchaphazela indlela ababona ngayo umbala. Le ndlela engaqondakaliyo, ezimeleyo ibikho ngaphezulu kweminyaka engama-2,000 - lixesha lokuphucula.

Indlela i-AI echonga kwaye iqikelele ngayo izifo isebenzisa umbala wolwimi

Iingcali zobunjineli kunye nabaphandi abaye baphuhlisa le nkqubo yokujonga ukuxilongwa kwangaphambili baye bathatha inxaxheba kuma-20 eesentimitha kude nomatshini ngexesha lesigaba sokuvavanya. I-AI edibeneyo emva koko yafumanisa umbala wolwimi lwabo kwaye yaxela kwangaphambili imeko yabo yezempilo ngexesha lokwenyani. Yalungisa i-hues kunye ne-luminance isebenzisa imifuziselo yendawo yemibala.

I-algorithm ye-XGBoost ichaze ngokuchanekileyo izifo I-96.6% yexesha ngexesha lovavanyo. Lo mzekelo wokufunda umatshini uchanekile kuba uyaqhubeka nokuziphucula emva kwemveliso yawo yokuqala. Iphindaphinda ukwenza uqikelelo kwaye ibale ireyithi yempazamo ukuze isondele ngokuthe ngcembe kwinjongo yayo, iziqeqeshe ngokufanelekileyo ukwandisa ukuchaneka kwayo.

Uyibona Njani Umatshini Weqela Lophando Imibala Yeelwimi

Iiseli ze-cone - i-photoreceptors kwi-retina ejongene nombono wombala - ivakalelwa ngokubanzi kwimimandla ebomvu-blue-green (RGB). Nangona kunjalo, le modeli yendawo yombala ayibonakalisi kakuhle ulwazi. Inkqubo yombono wekhompyuter endaweni yoko isebenzise iYCbCr, LAB, YIQ kunye neHSV. Ngokungafaniyo nabantu, ayiphelelanga kukukhanya okubonakalayo okumxinwa.

Olunye uphononongo lubonisa inkqubo yombono yekhompyuter enikwe amandla e-AI inokuqaphela ngokuchanekileyo kwaye iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde iphinde ibone imibala, nto leyo ethetha ukuba iyakwazi ukubona imibala esingenakuyibona. Ngelixa iicones zam ze-RGB-sensitive zinako kuphela bona malunga nesigidi semibala , obu bugcisa bunokubona malunga ne-16,8 yezigidi. Ichaneke ngendlela emangalisayo, nayo, itenxa kuphela kwimbonakalo yombala “engabonwayo” malunga ne-1% yexesha ngokomndilili.

Ukuba imodeli yokufunda yomatshini inokuthatha umahluko ofihlakeleyo wokugcwala kunye nokukhanya okungabonakaliyo kum, kutheni ingenakukwazi ukubona imibala endingakwaziyo? Ngokwendalo, iimpembelelo zibalulekile-i-AI inokukwazi ukodlula oogqirha ngokungaguqukiyo.

Kutheni kubalulekile ukusebenzisa imibala yolwimi kuxilongo?

Kuba imodeli enye inokunxibelelana nabantu abaninzi ngaxeshanye, inokunceda izigulana ezahlukeneyo ngaxeshanye. Akukho mfuneko yokuba bandwendwele ikliniki — banokukhuphela usetyenziso kwaye basebenzise ikhamera yeefowuni zabo. Kuba umatshini weqela lophando unokuchonga kwaye uqikelele izifo nokuba kukhanyiswe kangakanani na, mancinci amathuba okuphuma okungachanekanga.

Ndiyakholelwa ekhaya, ukuhlolwa okuqhutywa yi-AI kunokuguqula ukhathalelo lwempilo, kuyenze ifikeleleke kwaye ifikeleleke. Izigidi zabantu ziyafa minyaka le ngenxa yezifo ebebenethuba lokulwa nazo ukuba bebezibambe kwangethuba. Ngokomzekelo, eUnited States, malunga Bangama-12,000 abantu abafunyaniswayo kunye nomhlaza we-gallbladder minyaka yonke, okukhokelela ekufeni kwabantu abaqikelelwa kwi-4,000 ngonyaka. I-AI inokubaxilonga bekude, ibancede bafumane ukhathalelo kwangethuba.

Ukubanakho kwe-AI ukwenza utshintsho kuxilongo kunokuba luncedo kwizibhedlele njengoko bekuya kuba luncedo kwizigulana. Ngaphandle kokusasazeka kwedijithali, inkcitho yezonyango inyuke malunga nama-47% ukusuka kwi-2021 ukuya kwi-2022 - iyonke i-55 yeebhiliyoni zeedola. Eyona ndleko iphambili yayikukuba i-avareji yexesha lababoneleli elichithwe nezigulane linyukile.

Ubuchule bokuzenzela be-AI kunye nendalo yokuzimela inokulungelelanisa ukuqeshwa. Ubungqina bubonisa ukuba obu bugcisa bunokunceda amaziko onyango gcina i-20% -50% yohlahlo lwabiwo-mali lwabo lonyaka , ngoko inamandla apha. Inokwenza izibhedlele zikwazi ukonga iindleko kubathengi okanye zityale imali kwizixhobo ezisindisa ubomi.

Khange ndicinge ukuba ukusebenzisa imibala yolwimi kuxilongo kunokuba nefuthe kangako, kodwa kunamandla angenakulinganiswa nanto. Ngaba i-AI ingathatha indawo yoogqirha? Mhlawumbi akunjalo. Nangona kunjalo, ndiyakholelwa ukuba iya kuba yinto esisiseko kwishishini lezonyango njengoko yongeza ukuchongwa kwesifo, ukuxela kwangaphambili kunye nonyango. Ubuchwephesha babo, obudityaniswe namandla okufunda koomatshini, buya kuba kukubhanqa okungenakuhlulwa.

Indlela ukhathalelo lwempilo olunokuthi lusebenzise ngayo obu buchwepheshe bungcono

Umzi-mveliso wokhathalelo lwempilo uyayazi i-AI kwaye unomdla wokuyamkela, ngoko ke kusenokwenzeka ukuba ngumcimbi wexesha ngaphambi kokuba ibe yinto eqhelekileyo. Iingcali zeprojekthi yexabiso lentengiso yethekhinoloji kweli candelo liya kukhula kwi Izinga lokukhula ngonyaka elihlanganisiweyo lama-38.4% ukusuka kwi-2020 ukuya kwi-2030 - ukufikelela kuqikelelo lwe-208.2 yeebhiliyoni zeedola - igxininisa ukwanda kwayo ngokukhawuleza kwintsimi.

Nangona kunjalo, ngelixa i-__ i-72% yoogqirha ivuma __babona isithembiso esikhulu se-AI kwi-diagnostics, kuphela i-38% iyisebenzisa ekusebenzeni. Ngokwenyani, kungathatha iminyaka ngaphambi kokuba basebenzise imibala yolwimi ehlalutywe imodeli yokuxilongwa kunye nokuqikelela. Uqikelelo lwam olufundileyo lwelokuba izibhedlele ziyakuthatha ishumi leminyaka ukutsiba kwiihupula endaweni yokunqumla iteyiphu ebomvu.

Ngethamsanqa, ihlabathi lezempilo leselula likhula ngokukhawuleza kwaye lifikeleleka nanini na. Ngelixa ndingenako ukukhankanya i-mHealth ngaphandle kokubonisa ukunqongophala kwayo kulawulo lolawulo kunye nokhuseleko lwabucala, kuya kuba bubudenge ukuyihoya xa uxoxa ngekamva le-AI yezonyango.

Kutheni ukhathalelo lwempilo kufuneka lulumke malunga nale teknoloji

Ngaxa lithile, ndiye ndazibamba ndicinga ukuba le mpumelelo ilunge kakhulu ukuba ibe yinyaniso. Yintoni oyibambileyo? Ziziphi izinto ezingalunganga zokuxilongwa kwe-AI? Ndenze isabelo sam esifanelekileyo sophando kulo mmandla, ngoko bendisazi ukuba ubumfihlo, imiba yeenqobo ezisesikweni kunye nolawulo iya kubakho. Nangona kunjalo, uphononongo olutyhile le teknoloji yempumelelo kwakhona kufanelekile ukuphononongwa phantsi kwe-microscope.

Ndiqaphele ukuba umsebenzi wabaphandi ekusebenziseni i-AI ukuhlalutya imibala yolwimi ukuze uxilongwe awukaphononongwa ntanga. Oko yapapashwa ngoJuni ka-2024, ayimangalisi loo nto. Nangona kunjalo, oko kuthetha izinga lokuchaneka kwabo ngama-98%. ithe phezulu emoyeni. Inkqubo yabo ilungile, kodwa yintoni isayensi ngaphandle kokuphindaphinda? Ngaphandle kokuba umntu uphinda iziphumo zabo, imboni yokhathalelo lwempilo inokungaluthathi ulwakhiwo lwabo.

Ingaba oku kuPhumelela kuya kuchaphazela njani iTekhnoloji yezoNyango ye-AI?

Ukusebenzisa isixokelelwano sokucinga ukuhlalutya imibala yolwimi kuxilongo kunokubonakala kuyinto eqhelekileyo kunye nokuphumelela okungephi, kodwa kunokuguqula ukhathalelo lwempilo. Ukuqikelela ngokuchanekileyo izifo usebenzisa i-app kunokusindisa amawaka obomi. Ngapha koko, abanye abaphandi banokusebenzisa le teknoloji ukukhuthaza impumelelo yabo yoxilongo oluqhutywa yi-AI.