Ubuhlakani bokwenziwa bufinyelela ingqopha-mlando entsha nosuku ngalunye oludlulayo. Ukwengeza kulokhu kulandelana, i-AI ye-Google manje imvamisa idlula ochwepheshe, ithola umdlavuza webele emzimbeni womuntu.
Ukuhlolelwa umdlavuza webele ngokuvamile kwenziwa ngokuskena ama-mammogram (noma izithombe ze-X-ray zamabele), kodwa anemingcele. Banganikeza umphumela ongemuhle noma ongemuhle ongelona iqiniso.
I-mammogram engalungile ibukeka ijwayelekile ngisho nalapho ukhona umdlavuza webele. Ngokufanayo, i-mammogram ene-fake-positive ibonisa umdlavuza webele ngisho noma ungekho.
I-AI yakwaGoogle ingathola umdlavuza webele, kodwa...
I-Google AI uqeqeshiwe ukufunda nokuhlaziya ama-mammograms ukuze uthole ukuthi kukhona amangqamuzana omdlavuza. Ngalokho, abacwaningi basebenzise cishe ama-mammograms angaziwa angama-76,000 kwabesifazane base-UK kanye nama-mammograms ayi-15,000 kwabesifazane base-US.
Basebenzise isizindalwazi se-mammogram esingaziwa sabesifazane abangu-25,000 base-UK kanye nabesifazane abangu-3,000 base-US ukuze bahlole i-AI.
Imiphumela ebuyiswe yi-AI yaqhathaniswa nemibiko yangempela yezokwelapha ukuze kuqinisekiswe ukunemba. Ikwazile ukwehlisa inani lama-negative angamanga ngo-9,4% kwabesifazane baseMelika kanye no-2,7% kwabesifazane base-UK.
Ngakolunye uhlangothi, yehlise amaphuzu angamanga ngo-5,7% e-US kanye ne-1,2% e-UK.
Okutholakele kushicilelwe kujenali Nature yiqembu elihlanganisa abacwaningi abavela e-DeepMind, Royal Surrey County Hospital, Northwestern University kanye neCancer Research UK Imperial Centre.
namanje akaphelele
Ngenkathi i-AI ye-Google ikwazile ukushaya ochwepheshe babantu ezimweni eziningi, kube nezikhathi lapho ochwepheshe bethola khona izimpawu zomdlavuza ezimeni lapho i-AI iphuthelwe khona.
Kodwa-ke, lapha, i-Google iveza ukuthi i-AI ibikwazi ukufinyelela olwazini oluncane kunochwepheshe babantu, njengemilando yesiguli kanye nama-mammograms adlule. Noma kunjalo, wakwazi ukwenza kahle kakhulu.
Ngakho-ke kusho ukuthi i-AI ayikwazi ukufaka esikhundleni abantu kulo mkhakha okwamanje. Kodwa-ke, inhloso ukuthi ubuchwepheshe kufanele bugcwalise ama-radiologists ngesikhathi sokuhlola futhi bakhulise ukunemba nokusebenza kahle kwemiphumela.
Njengoba intuthuko iqhubeka, into eyodwa abacwaningi abazama ukuyithola ukuthi bangayenza kanjani imiphumela ye-AI ibe jikelele.
Okokuqala, benze isivivinyo esihlukile lapho i-AI yaqeqeshwa khona ngedatha evela kwabesifazane base-UK futhi yahlola isethi yedatha yabesifazane base-US. Yehlise ukunegethivu okungamanga ngo-8.1% kanye nemibono engamanga ngo-3.5%.
NgeThe Verge