Batsho ukuba kwizinto ezintsha zakudala. Umzekeliso obhalwe kwiinkulungwane ezine zexesha lethu ufanelekile ukuqonda ukuba yeyiphi imida yezicelo ezitsha zeArtificial Intelligence. Ndibhekisa "kumqolomba we-ChatGPT" ongekho ngaphezulu okanye ngaphantsi kolungelelwaniso lomzekeliso odumileyo womqolomba kaPlato.
Andichasanga nokusetyenziswa kwezixhobo zobuntlola ezenziweyo. Enyanisweni, ndifumanisa ukuba bawenza lula umsebenzi. Kodwa okoko nje zisetyenziswe ngabantu abanolwazi olwaneleyo lokuvavanya umsebenzi wakho.
Umzekelo; umntu unokucela i-ChatGPT ukuba ibhale iplagi ye-WordPress, kodwa ukuba umntu akanalo ulwazi lwe-PHP loo plugin inokubangela iingxaki ezinzulu zokhuseleko.
Umzekeliso womqolomba
UPlato wayesisithandi sobulumko esingumGrike esasiphila phakathi kwenkulungwane yesi-XNUMX neyesine ngaphambi kweXesha Eliqhelekileyo. Iingcamango zakhe wayezivakalisa ngeentsomi nangemizekeliso. Eyona nto yayisaziwa kakhulu yile yomqolomba.
Ithunyelwe kwi La República, umzekeliso uyaqikelela iqela labantu libotshelelwe emqolombeni, emva kwabo linomlilo obeka izithunzi eludongeni phambi kwabo. Izithunzi kuphela kwento abayibonayo kwaye bacinge ukuba yiyo yodwa into ekhoyo, ukutyeshela oko kulele ngaphaya.
Xa elinye lamabanjwa likhululwa, liyakwazi ukubona ihlabathi ukuba likuko ngokwenene yaye liyayiqonda indlela awayelinganiselwe ngayo amava alo emqolombeni.
Ngokutsho kwabaphengululi bakaPlato, lo mzekeliso ugxininisa ukuba sonke siphila ubomi bethu ngokusekelwe kulwazi kunye namava ethu. Ulwazi kunye namava alingana nezithunzi zomqolomba. Njengamabanjwa, kukho inyaniso eyinyani kwaye ingaphaya kokuqonda kwethu.
Umqolomba we-ChatGPT
I-ChatGPT kunye nabakhuphisana nabo banabo bobabini abathandayo kunye abagxeki. Kodwa, akukho mntu wayenike inkcazo yobugcisa malunga nokungaphumeleli kwayo kude kube nenqaku iposwe eNew Yorker ngumbhali wentsomi yesayensi uTed Chang
Ukucacisa iziphene kwiimodeli zolwimi, i-Chang yenza isifaniso nento eyenzekayo ngemifanekiso kunye neefayile ezivakalayo.
Ukurekhoda kunye nokuveliswa kwakhona kwefayile yedijithali kufuna amanyathelo amabini: eyokuqala yi i-encoding, ngelo xesha ifayile iguqulelwa kwifomati ehlangeneyo ngakumbi, ilandelwa yi-decoding, eyinkqubo ebuyela umva.. Inkqubo yokuguqula ibizwa ngokuba yilahleko (ifayile ebuyiselweyo iyafana neyokuqala) okanye ilahleko (Olunye ulwazi lulahlekile ngonaphakade). Uxinzelelo olulahlekileyo lusetyenziswa kumfanekiso, ividiyo, okanye iifayile zeaudio kwaye uninzi lwexesha alibonakali. Xa kunjalo, ibizwa ngokuba yi-compression artifact. Izinto zakudala ezicinezelayo zivela ngendlela yokufiphaza kwimifanekiso okanye ukubetheka kwisandi.
UChang usebenzisa umzekeliso weJPG entsonkothileyo esuka kwiwebhu ukubhekisa kwiimodeli zolwimi. Kwaye, oku kuchanekile kakhulu. Zombini zicinezela ulwazi olugcina kuphela "Into ebalulekileyo". LImifuziselo yolwimi ivelisa, ukusuka kumthamo omkhulu wedatha yokubhaliweyo, ukubonakaliswa okuhlangeneyo kweepateni kunye nobudlelwane phakathi kwamagama namabinzana.
Ukusuka kuyo, isicatshulwa esitsha senziwa sizama kangangoko ukwenza ukuba sifane nomxholo kunye nentsingiselo kumbhalo wokuqala. Ingxaki kuxa kungekho lwazi lwaneleyo kwiwebhu ukwenza umbhalo omtsha. Oku kuguqulela kwi-ChatGPT ikwazi ukubhala isincoko senqanaba lekholeji, kodwa ingenzi imisebenzi elula enedijithi ezi-5.
UChang ugqiba ngelithi:
Nokuba kunokwenzeka ukunqanda iimodeli ezinkulu zolwimi ekuthatheni inxaxheba ekugunyaziseni, ngaba kufuneka sizisebenzise ukwenza umxholo wewebhu? Oku kuya kuba sengqiqweni kuphela ukuba injongo yethu kukupakisha kwakhona ulwazi esele lufumaneka kwiWebhu. Ezinye iinkampani zikhona ukwenza loo nto kanye; sizibiza ngokuba ziifektri zomxholo. Mhlawumbi i-fuzziness yeemodeli zolwimi luncedo kubo, njengendlela yokuphepha ukuphulwa kwelungelo lokushicilela. Ngokubanzi, nangona kunjalo, ndingatsho ukuba nantoni na elungileyo kwiifektri zomxholo ayilunganga kubantu abafuna ulwazi. Ukunyuka kolu hlobo lokupakisha kwakhona yinto eyenza kube nzima kuthi ukufumana into esiyifunayo kwi-Intanethi ngoku.; Okukhona isicatshulwa esiveliswa ziimodeli zolwimi ezinkulu zipapashwa kwiWebhu, kokukhona iWebhu iba luguqulelo olumfiliba ngokwayo.
Kwaye, njengamabanjwa emqolombeni, amava ethu aya kuba mancinane kakhulu kunoko kunikelwa yinyani.