I-AI nekusasa lomsebenzi: Konke sekuzoshintsha


I-New York
CNN

Ezinyangeni ezimbalwa nje, uzokwazi ukubuza a umsizi obonakalayo ukuze abhale amanothi omhlangano phakathi nocingo lomsebenzi, afingqe imicu ye-imeyili emide ukuze alungise ngokushesha izimpendulo eziphakanyisiwe, udale ngokushesha ishadi elithile ku-Excel, futhi uguqule idokhumenti ye-Word ibe isethulo se-PowerPoint ngemizuzwana.

Futhi lokho kukumapulatifomu e-Microsoft angama-365.

Esontweni eledlule, isimo sobuhlakani bokwenziwa esithuthuka ngokushesha sibonakale siphindela phambili futhi. I-Microsoft ne-Google ngayinye kwembuliwe izici ezintsha ezinikwe amandla yi-AI zamathuluzi azo okukhiqiza ayisiginesha futhi i-OpenAI yethula inguqulo yayo yesizukulwane esilandelayo yobuchwepheshe obusekela ithuluzi layo lokuxoxa ngegciwane, i-ChatGPT.

Kungazelelwe, amathuluzi e-AI, osekuyisikhathi eside esebenza ngemuva kwezinsizakalo eziningi, manje anamandla kakhulu futhi abonakala kakhulu kulo lonke uhla olubanzi nolukhulayo lwamathuluzi wendawo yokusebenza.

Izici ezintsha ze-Google, isibonelo, isithembiso ukusiza “ukucabanga ngengqondo” kanye “nokuhlola ukuhlola” umsebenzi obhaliwe kokuthi Amadokhumenti. Okwamanje, uma indawo yakho yokusebenza isebenzisa inkundla yengxoxo edumile i-Slack, uzokwazi ukuba nethuluzi layo le-ChatGPT ukukhulumela nozakwenu, okungenzeka uyicele ukuthi ibhale futhi iphendule imilayezo emisha futhi ifingqa izingxoxo eziteshini.

I-OpenAI, iMicrosoft ne-Google zihamba phambili kulo mkhuba, kodwa aziyedwa. I-IBM, i-Amazon, i-Baidu ne-Tencent zisebenza ngobuchwepheshe obufanayo. Uhlu olude lwabaqalisi nabo bathuthukisa abasizi bokubhala be-AI namajeneretha ezithombe.

I-pitch evela ezinkampanini zobuchwepheshe icacile: I-AI ingakwenza ukhiqize kakhudlwana futhi iqede umsebenzi we-grunt. Njengoba i-CEO ye-Microsoft uSatya Nadella esho ngesikhathi sesethulo NgoLwesine, “Sikholelwa ukuthi lesi sizukulwane esilandelayo se-AI sizovula igagasi elisha lokukhula komkhiqizo: abashayeli bezindiza abanamandla abaklanyelwe ukususa isidina emisebenzini yethu yansuku zonke nemisebenzi, basikhulule ukuze siphinde sithole injabulo yendalo.”

Kepha inani elikhulu lezinketho ezintsha ezifika emakethe kuyaxaka futhi, njengokunye okuningi embonini yezobuchwepheshe kule minyaka eyishumi edlule, kuphakamisa imibuzo yokuthi ngabe izophila ngokuvumelana nodumo noma ibangele imiphumela engahlosiwe, okuhlanganisa ukunika amandla ukukopela nokuqeda isidingo sezindima ezithile (nakuba lokho kungase kube inhloso yabanye abamukelayo).

Ngisho nesithembiso sokukhiqiza okukhulu akucacile. Ukwenyuka kwama-imeyili akhiqizwe nge-AI, ngokwesibonelo, kungase kukhuphule umkhiqizo womthumeli kodwa kunciphise kubamukeli abagcwele imilayezo ekhiqizwa ikhompuyutha ende kunesidingo. Futhi ngenxa yokuthi wonke umuntu unenketho yokusebenzisa i-chatbot ukuze axhumane nozakwabo akusho ukuthi bonke bazokhetha ukwenza kanjalo.

Ukuhlanganisa lobu buchwepheshe “ezingxenyeni eziyisisekelo zesoftware yokukhiqiza iningi lethu elisebenzisa nsuku zonke kuzoba nomthelela omkhulu endleleni esisebenza ngayo,” kusho uRowan Curran, umhlaziyi wakwaForrester. “Kodwa lolo shintsho ngeke luhlanze yonke into futhi wonke umuntu kusasa – ukufunda ukusebenzisa la makhono kangcono ukuthuthukisa nokulungisa ukuhamba komsebenzi wethu okukhona kuzothatha isikhathi.”

Noma ubani oke wasebenzisa inketho yokuqedela ngokuzenzakalela lapho ebhala i-imeyili noma ethumela umlayezo usevele wazibonela ukuthi i-AI ingasheshisa kanjani imisebenzi. Kodwa amathuluzi amasha athembisa ukwedlula lokho.

Igagasi elivuselelwe lokwethulwa komkhiqizo we-AI liqale cishe ezinyangeni ezine ezedlule lapho i-OpenAI ikhipha inguqulo ye-ChatGPT ngesisekelo esilinganiselwe., abasebenzisi abamangalisayo abakhiqiza izimpendulo ezizwakalayo zomuntu ekwazisweni komsebenzisi, ukuphumelela izivivinyo emanyuvesi ahlonishwayo kanye nokubhala izindatshana ezihehayo ngezihloko eziningi.

Kusukela ngaleso sikhathi, ubuchwepheshe – okuyinto iMicrosoft yenza a Ukutshalwa kwezimali “kwezigidigidi zamadola” ekuqaleni kwalo nyaka – isithuthuke kuphela. Ngasekuqaleni kwaleli sonto, i-OpenAI yethule i-GPT-4, inguqulo enamandla kakhulu yobuchwepheshe obusekela i-ChatGPT, futhi ethembisa ukufuthela iziphindaphindo zangaphambilini ziphume emanzini.

Ezivivinyweni zakuqala kanye nedemo yenkampani, i-GPT-4 yasetshenziselwa ukubhala amacala, ukwakha iwebhusayithi esebenzayo ngomdwebo odwetshwe ngesandla futhi dala kabusha imidlalo eyisakhiwo sodumo njenge-Pong, i-Tetris noma i-Snake enolwazi oluncane kakhulu lokubhala amakhodi lwangaphambilini.

I-GPT-4 iyimodeli yolimi enkulu eqeqeshwe emithonjeni eminingi yedatha ye-inthanethi ukuze ikhiqize izimpendulo ekwazisweni komsebenzisi.

Ubuchwepheshe obufanayo obusekela izici ezimbili ezintsha ze-Microsoft: “I-Co-pilot,” ezosiza ukuhlela, ukufingqa, ukudala nokuqhathanisa amadokhumenti kuzo zonke izinkundla zayo, kanye ne-Business Chat, i-ejenti egibela kanye nomsebenzisi njengoba besebenza futhi bezama. ukuqonda nokwenza umqondo ngedatha yabo ye-Microsoft 365.

I-ejenti izokwazi, ngokwesibonelo, ukuthi yini eku-imeyili yomsebenzisi kanye nekhalenda labo losuku, kanye namadokhumenti abebesebenzela kuwo, izethulo akade bezenza, abantu ahlangana nabo, kanye izingxoxo ezenzeka endaweni yamaQembu abo, ngokusho kwenkampani. Abasebenzisi bangabese becela Ingxoxo Yebhizinisi ukuthi yenze imisebenzi efana nokubhala umbiko wesimo ngokufingqa wonke amadokhumenti kuzo zonke izinkundla kuphrojekthi ethile, bese bebhala i-imeyili engathunyelwa eqenjini labo nesibuyekezo.

U-Curran uthe nje ukuthi la mathuluzi anikwe amandla yi-AI azoshintsha kangakanani umsebenzi kuncike kuhlelo lokusebenza. Isibonelo, uhlelo lokusebenza lokucubungula amagama lungasiza ukukhiqiza izinhlaka nokusalungiswa, uhlelo lombukiso wesilayidi lungasiza ukusheshisa umklamo kanye nenqubo yokudala okuqukethwe, futhi uhlelo lokusebenza lwesipredishithi kufanele lusize abasebenzisi abaningi basebenzisane futhi benze izinqumo eziqhutshwa idatha. Lesi sakamuva ukholelwa ukuthi sizokwenza umthelela obaluleke kakhulu endaweni yokusebenza esikhathini esifushane nesikhathi eside.

Ingxoxo yokuthi lobu buchwepheshe buzoyithinta kanjani imisebenzi “kufanele igxile emisebenzini esikhundleni semisebenzi iyonke,” kusho yena.

Yize isibuyekezo se-OpenAI’s GPT-4 sithembisa ukulungisa ezinye zezinselelo zayo ezinkulu – kusukela emandleni ayo okuqhubekisela phambili ukuchema, kwesinye isikhathi ukungalungile ngokweqiniso nokuphendula ngendlela enolaka – kusengenzeka ukuthi ezinye zalezi zinkinga ukuze bathole indlela yabo emsebenzini, ikakhulukazi uma kuziwa ekusebenzelaneni nabanye.

U-Arijit Sengupta, oyi-CEO nomsunguli wenkampani yezixazululo ze-AI i-Aible, uthe inkinga nganoma iyiphi imodeli yolimi enkulu ukuthi izama ukujabulisa umsebenzisi futhi ngokuvamile yamukela isisekelo sezitatimende zabasebenzisi.

“Uma abantu beqala ukuhleba ngokuthile, kuzokwamukela njengokujwayelekile bese beqala ukukhiqiza okuqukethwe [related to that],” kusho uSengupta, engeza ngokuthi kungase kubhebhethekise izindaba ezithinta abantu bese kuphenduka ubuxhwanguxhwangu ehhovisi.

Ku-tweet ekuqaleni kwaleli sonto, i-OpenAI CEO uSam Altman wabhala ukuthi ubuchwepheshe balezi zinhlelo “zisanamaphutha, zisalinganiselwe, futhi zisabonakala buhlaba umxhwele kakhulu uma busetshenziswa okokuqala kunangemva kokuchitha isikhathi esiningi nakho.” Inkampani iphinde yasho kokuthunyelwe kwebhulogi ukuthi “kufanele kuqashelwe kakhulu uma kusetshenziswa imiphumela yemodeli yolimi, ikakhulukazi ezimweni eziphakeme kakhulu.”

U-Arun Chandrasekaran, umhlaziyi we-Gartner Research, uthe izinhlangano zizodinga ukufundisa abasebenzisi bazo ukuthi lezi zixazululo zinhle ngani nokuthi imingcele yazo iyini.

“Ukuthembela okungenangqondo kulezi zixazululo kuyingozi njengokuntula ukholo ngokuphelele ekusebenzeni kwazo,” kusho uChandrasekaran. “Izixazululo ze-Generative AI nazo zingakha amaqiniso noma zethule ulwazi olunganembile ngezikhathi ezithile – futhi izinhlangano kufanele zilungele ukunciphisa lo mthelela omubi.”

Ngesikhathi esifanayo, eziningi zalezi zinhlelo zokusebenza azikho esikhathini (idatha ye-GPT-4 eqeqeshwe ngayo inqamule ngoSepthemba 2021). Umthwalo kuzofanele ube kubasebenzisi ukwenza yonke into kusukela ekuhloleni kabili ukunemba ukuze bashintshe ulimi ukuze babonise ithoni abayifunayo. Kuzobaluleka futhi ukuthola ukuthengwa nokusekelwa kuzo zonke izindawo zokusebenza ukuze amathuluzi aqale.

“Ukuqeqeshwa, imfundo kanye nokuphathwa kwezinguquko zenhlangano kubaluleke kakhulu ukuqinisekisa ukuthi abasebenzi bayayisekela imizamo futhi amathuluzi asetshenziswa ngendlela okwakuhloswe ngayo,” kusho uChandrasekaran.