URich Sutton ungomunye wamaphayona obuchwepheshe obubalulekile obusekela igagasi lamanje lobuhlakani bokwenziwa, kodwa manje ukholelwa ukuthi indlela izinkampani ezinkulu zobuchwepheshe eziyithathayo ukuze i-AI iqhubeke ikhula inezinkinga ezinkulu.
Lo sosayensi wekhompyutha waseCanada, ongumnqobi woMklomelo weTuring, uthe esiqeshini se-podcast ye-Sequoia Capital esikhishwe ngoLwesibili ngesikhathi sendawo ukuthi imboni ye-AI iya ngokuya ithembela kudatha yokwenziwa yokuqeqesha, futhi lokhu kungase kuholele yonke imboni endleleni engalungile.
Ngesikhathi ebuzwa ukuthi ukusebenzisa idatha yokwenziwa kungasiza yini amamodeli amakhulu olimi ukuthi aqhubeke nokwanda, uSutton washo ngokuqondile: “Leli yiphutha elikhulu impela. Mhlawumbe lokhu kuzoba isifundo esibalulekile esilandelayo.”
Idatha yokwenziwa isho ulwazi olwenziwa ngamabomu ama-algorithm noma amamodeli e-AI, kunokuba luqoqwe kudatha eluhlaza yomhlaba wangempela. Njengoba izinkampani ze-AI ziqhubeka nokufuna ku-inthanethi nakweminye imithombo emisha yedatha yokuqeqesha, idatha yokwenziwa iya ngokuya iheha. Ngokwesibonelo, ukuqeqesha izimoto ezizishayelayo kungase kusetshenziswe izithombe zezimoto ezenziwe ngekhompyutha, kuyilapho amasistimu e-AI okuthola ukukhwabanisa engasebenzisa amarekhodi okwenziwayo okwenziwayo emabhange.
Ngokwezinga elithile, izinkampani ezinkulu zobuchwepheshe nazo ziyawamukela umbono kaSutton.
Izinkampani ezinkulu zobuchwepheshe zenza konke okusemandleni ukufuna idatha yangempela eyengeziwe. Phambilini i-OpenAI yafuna esidlangalaleni amasethi edatha amakhulu, ayimfihlo, inethemba lokuthola idatha engatholakali kalula ku-inthanethi ukuze iqeqeshe amamodeli e-AI. Ekuqaleni kwaleli sonto, i-Google yavuma ukukhokha ama-dollar ayizigidi ezingu-10 (Qaphela: ngezinga lokushintshisana lamanje cishe ama-yuan ayizigidi ezingu-67.569) ukuze ithenge idatha yangaphakathi nesofthiwe ye-Spirit Airlines, eyayisivele ingenamali, okugqamisa ukuthi idatha yangempela eyimfihlo iya ngokuya ibaluleka ekuqeqesheni i-AI.
USutton uthe ku-podcast yangoLwesibili ukuthi idatha eyenziwe ngabantu ayikwazi ukuthatha indawo yedatha yangempela ngokwanele.
Wasebenzisa ukuziphatha kwabantu njengesibonelo. “Ngeke sikwazi ukukhiqiza idatha yokwenziwa yemicabango yabanye abantu,” esho.
USutton uthanda kakhulu ukusebenzisa idatha yokuhlangenwe nakho kwangempela, okusho ulwazi olutholwa ama-agent e-AI ngokuxhumana nemvelo yangempela, ukubona okwenzeka ngempela, nangokuqhubeka nokufunda emiphumeleni yezenzo zawo.
“Awukwazi ukubikezela ngedatha yokwenziwa ukuthi i-drone izoxhumana kanjani nemvelo emhlabeni wangempela ophathekayo,” kusho uSutton. Waphinde waveza ukuthi ezinjini zamarobhothi kunezinto eziguquguqukayo njengokungqubuzana nokuguga. “Lo mhlaba uyinkimbinkimbi ngokwedlulele; noma yikuphi ukuwulingisa kumane kunikeze umbono omncane kakhulu wawo.”
Namuhla, lo mbono usuyisiqondiso sentuthuko yenkampani entsha.
Ngenyanga edlule, uSutton wasungula i-Oak Lab kanye nowayengumfundi wakhe uKhurram Javed. Le nkampani entsha ithuthukisa ama-agent e-AI akwazi ukuqhubeka nokufunda kokuhlangenwe nakho kwawo, kunokuthembela kakhulu kumasethi edatha amakhulu ahlelwe kusenesikhathi.
Okwamanje, i-Oak Lab ayikazimemezeli imijikelezo yoxhaso noma imininingwane yabatshalizimali.