I-ChatGPT igubha unyaka wayo wesithathu, isibe ngomunye we amathuluzi obuhlakani bokwenziwa anethonya kakhulu yomlando wakamuva. Okwaqala ngoNovemba 2022 njengelukuluku le-technophile kufinyeleleka esipheqululini Sekuyisevisi enkulu izigidi zabantu eziyisebenzisa nsuku zonke emsebenzini, ukufunda, noma ukuxoxa nje.
Ngalesi sikhathi, i-chatbot ye I-OpenAI isuke ekubeni ukuhlolwa okusekelwe embhalweni kuya ohlelweni olukwaziyo hlanganisa izwi, isithombe, ividiyo, ukuphequlula i-inthanethi nokucabanga okuthuthukileNgenkathi i-Europe iphikisana ngomthetho wayo futhi izinkampani zaseSpain ziwufaka ekuhambeni komsebenzi wazo, insizakalo izibonakalisa njengomunye wabadlali abakhulu enkathini yobuhlakani bokwenziwa obukhiqizayo.
Kusukela ku-viral novelty kuya kwelinye lamawebhusayithi avakashelwa kakhulu emhlabeni
Okwamanje, isevisi isivele ifinyelelwe amakhulu ezigidi zabantu ababhalisiwe, ngezilinganiso ezibeka isisekelo somsebenzisi wayo eduze Izigidi ezingu-800, futhi iwebhusayithi yayo ihlale ibalwa phakathi kwe amawebhusayithi amahlanu avakashelwa kakhulu emhlabeniNgokuvakashelwa okuphakathi kuka-5.100 no-5.500 wezigidigidi ngenyanga, lesi sibalo sikubeka kuligi efanayo nezikhondlakhondla ze-inthanethi, okwenza kucace ukuthi akusesona nje isilingo saselabhorethri.
Lokhu kutholwa okukhulu kuyishintshile indlela abantu abaningi abafinyelela ngayo ulwazi: ngaphezu nje kwenjini yokusesha evamile, i-ChatGPT izisungule njenge umeluleki womuntu siqu, osebenza naye emsebenzini, kanye nokwesekwa kwezemfundoKusukela kubafundi abaxhumana nayo ukuze balungiselele izivivinyo kuya kochwepheshe abayisebenzisela ukubhala imibiko noma ikhodi yokususa iphutha, i-chatbot ingene ezindaweni ezihluke kakhulu zempilo yansuku zonke.
I-OpenAI futhi iphrojekthi ukuthi, ngo-2030, ingafinyelela eduze Abasebenzisi abakhokhelayo abayizigidi ezingama-220Lokhu kungashintsha inkundla ibe isevisi yokubhaliselwe yomhlaba wonke, enemali engenayo ephindelelayo kanye nesisindo somnotho esiqhathaniswa nesezinye izinkampani ezinkulu zesofthiwe.
E-Spain nakulo lonke elaseYurophu, lokhu kwanda kuhambisane nokucutshungulwa nokugunyazwa kwe Uhlaka lokulawula lwe-AIefuna ukuchaza ngokomthetho isenzeko esithuthuke ngokushesha kakhulu kunemithethonqubo. Izingxoxo zomphakathi mayelana nomthelela wakho ekuqashweni, emfundweni, nasezimfihlakalweni sezivele zibanjiwe ezikhungweni, ezinkampanini, nasemanyuvesi.
Ukuziphendukela kwemvelo okusheshayo: ukusuka ku-GPT-1 kuya kumodeli yezinga lobudokotela
Ngemuva kwe-ChatGPT umndeni oyimodeli I-GPT (I-Generative Pre-trained Transformer)Amanethiwekhi e-Neural aqeqeshelwe amanani amakhulu ombhalo ukuze abikezele isiqeshana esilandelayo solimi. Yize umphakathi jikelele wajwayelana nayo ngo-2022, indaba yaqala phambilini, nge I-GPT-1 yethulwa ngo-2018 njengobufakazi bomqondo wokumodela okuthuthukisiwe kolimi lwemvelo.
I-GPT-2, eyethulwe ngo-2019, bekuyimodeli yokuqala ekwaziyo khiqiza amathekisthi amade anokuhambisana okuthileNgenkathi i-GPT-3, ngo-2020, yabonisa ukuthi i-AI ingabhala futhi iphendule ngendlela efana kakhulu neyomuntu ngobuningi bezimo. Isinyathelo esimaphakathi, GPT-3.5, yathuthukisa lawo makhono futhi yanikeza isisekelo sohlelo I-ChatGPT yasekuqaleni, eyaduma ingxoxo ne-AI ngefomethi yengxoxo.
Ngokuhamba kwesikhathi, izinguqulo ezigxile ekuthuthukiseni izindleko nokusebenza zakhululwa, njenge I-GPT-4 TurboKodwa ukugxuma okukhulu kwafika namamodeli aklanyelwe ukucabanga ngendlela enembayo nenembe kakhuluUchungechunge olubizwa ngokuthi "o", olunokuhlukahluka okufana no-o1, o3 kanye no-o3-mini, lwaklanyelwe ukunikela isikhathi esengeziwe kunqubo yokukhiqiza izimpendulo, ukuthuthukisa ikhwalithi emikhakheni efunwa kakhulu njengezibalo, izinhlelo noma ucwaningo lwesayensi.
Ukuziphendukela kwemvelo kuye kwaqhubeka GPT-4.1okuqinisa ukusetshenziswa kwamathuluzi angaphandle emisebenzini emide neyinkimbinkimbi, futhi ifinyelela umvuthwandaba okwamanje I-GPT-5 ne-GPT-5.1Lawa mamodeli anquma ngokuzenzakalelayo ukuthi anikeze nini impendulo esheshayo nokuthi kufanele ame kancane futhi athuthukise ukungqubuzana okujulile, kuze kube yilapho kuchazwa inkampani ngokwayo ngokuthi uma kuqhathaniswa nochwepheshe bezinga le-PhD ezindaweni eziningi.
Ngaleso sikhathi, i-OpenAI ibilokhu ilungisa ukuziphatha kwesistimu ukuze yenze okwengeziwe flexible ekusebenzelaneni nabasebenzisiukulungisa isitayela sezimpendulo kumaphrofayili ahlukene kanye nokwenza ngcono ukulandelwa kwemiyalelo eyinkimbinkimbi. Umphumela uba umsizi wengxoxo ongasabonwa njengendawo ekhiqiza umbhalo, kodwa njenge i-interlocutor yedijithali ekwazi ukusekela izinqubo zomsebenzi eziphelele.
Kusukela kumbhalo kuya ku-multimodality yomdabu kanye nama-agent azimele
Ngenkathi inguqulo yokuqala yomphakathi igqame ngamakhono ayo okubhala, ukugxuma okuya phambili kwavela ngenxa ye multimodalityOkungukuthi, ikhono lokusebenza ngezinhlobo ezahlukene zokuqukethwe ngale kombhalo nje. Nge GPT-4I-ChatGPT ifunde uku ukuhumusha izithombe nezwiUkwandisa ububanzi bemisebenzi engenzeka: kusukela ekuchazeni isithombe kuya ekuhlaziyeni igrafu noma ekulobeni kanye nokufingqa okurekhodiwe.
Ushintsho lwangempela lwafika nalo I-GPT-4oImodeli yokuqala yenkampani ene-multimodality yomdabu. Esikhundleni sokuphatha umbhalo, umsindo, nesithombe ngokuhlukene, isistimu ikucubungula ndawonye, ukuyivumela ukuthi iphendule ngesikhathi sangempela izingxoxo zezwi, amavidiyo, nemibhalo ebonakalayo ngesivinini esimangalisayo. Inguqulo yayo engasindi, 4o mini, iletha le misebenzi kumadivayisi anezinsiza ezimbalwa.
Ngokusekelwe kulokho, okulandelayo kuye kwathuthukiswa: amakhono e-ejenti. Amamodeli afana o3 futhi o4-mini Bangazinqumela lapho bedinga hlola iwebhusayithi, sebenzisa ithuluzi langaphandle noma ukhiphe ikhodi ukuxazulula isicelo. Ngale ndlela, isistimu ayikhiqizi umbhalo kuphela, kodwa futhi iyakwazi ukwenza ukulandelana okufushane kwezenzo ukuze ilethe imibiko, izifinyezo, noma izinhlelo zokusebenza ngendaba yamasekhondi noma imizuzu embalwa.
Le ndlela ye-ejenti ezimele isiqala ukubonakala kwizicelo ezikhonkolo ngaphakathi kwezinkampani zaseYurophu, kusukela ku- ukuzenzekela kwemisebenzi yokuphatha Lokhu kufaka phakathi usizo lokuhlaziya idatha kanye nokuhlelwa kwemibhalo yezobuchwepheshe. E-Spain, inani elikhulayo lama-SME nezinkampani ezinkulu zihlola ukuhlanganiswa nezinhlelo zabo zangaphakathi, nakuba zisaqaphile ngenxa yemiphumela yezomthetho nezokuphepha.
Ngokuhambisanayo, inkundla ibizivumelanisa nokusebenzisana kwansuku zonke: ivumela ukugcinwa eduze kwezingxoxo zezwi zesikhathi sangempelaukuhumusha izithombe ezithathwe ngomakhalekhukhwini kanye nokusebenza ngevolumu enkulu yombhalo, njenge layisha ifayela le-PDF ku-ChatGPTEminingi yale misebenzi ifakwe kancane kancane e-Europe, kwesinye isikhathi nokuthunyelwa ngezigaba ukuze kuhambisane nemithetho ye-EU.
Ukusuka kokuzijabulisa kuya enjinini yokukhiqiza kanye negebe labasebenzi
Ezinyangeni zayo zokuqala, isikhalazo esiningi se-ChatGPT sisemandleni ayo khiqiza imibhalo yokuqamba, amahlaya noma izindaba ezimfushaneOkuthile abasebenzisi abaningi e-Spain nakwamanye amazwe ase-Europe abayihlola cishe njengomdlalo. Ngokuhamba kwesikhathi, nokho, ithuluzi lithathe indima ebaluleke kakhulu njenge i-lever yokukhiqiza emahhovisi, ezikhungweni zemfundo, kanye nemisebenzi yomuntu siqu.
Izifundo ezivela ezikhungweni ezifana ne I-MIT kanye ne-Harvard Bawubalile lo mthelela, bephawula ukwanda okufika phezulu 59% ekukhiqizeni futhi nxazonke 40% ezingeni lomsebenzi uma usebenzisa amathuluzi obuhlakani bokwenziwa uma kuqhathaniswa nalawo angawasebenzisi. Lezi zibalo zikhomba ushintsho olukhulu endleleni umsebenzi wolwazi ohlelwa ngayo.
Ngamazwi asebenza ngokukhethekile ku-AI asebenza ebhizinisini, lo mgubho weminyaka wesithathu uphawula inguquko efana ne-a "inguquko yomsebenzi wengqondo"ngokunciphisa kakhulu izindleko zemisebenzi efana nokubhala imibiko, ukudala izethulo, noma ukubhala ikhodi. Eqinisweni, ziyanda inkulumo ngegebe hhayi phakathi kwabantu nobuhlakani bokwenziwa, kodwa phakathi kwalabo Bafunda ukusebenza ngala mathuluzi nalabo abakhetha ukuhlala eceleni.
Kulo mongo, izinhlelo ezithile zokuqeqesha, ukubonisana, kanye nezinhlelo zokuthuthukiswa kwamakhono ziye zavela kulo lonke elaseYurophu, okuhloswe ngazo ukufundisa ochwepheshe abavela emikhakheni eyahlukene kakhulu ukuya. hlanganisa i-ChatGPT kuzinqubo zakhoKusukela eminyangweni yezokumaketha izenzela okusalungiswa emikhankasweni kuya ezimpahleni zomthetho ezisebenzisa i-AI ukuze zihlele imibhalo yezomthetho, ukwamukelwa kudala amaphrofayili amasha kanye nezimfuno emakethe yemisebenzi.
Ngasikhathi sinye, izinhlangano ziphoqeleka ukuthi zicabange kabusha ukuhamba kwazo, ngoba ukusetshenziswa kwamamodeli okukhiqiza akuthinti nje imisebenzi ehlukanisiwe, kodwa lungisa kabusha zonke izinqubo ngaphakathi kwenkampaniInkulumo-mpikiswano ayisekho kangako mayelana nokuthi la mathuluzi kufanele ahlanganiswe yini, kodwa ukuthi lokho kwenziwa kanjani ngokuzibophezela, ngempumelelo, nangokuhambisana nemithethonqubo yamanje.
Ukuphakama nokwehla kwesisebenzi esisebenza naye… kanye nomuntu omthembayo
Ngokuhambisana nezinzuzo ezisobala, ukukhuphuka kwe-ChatGPT kuveze inani lezinto eziningi ubungozi kanye nezingqinamba ezithatha indawo ngokuya ngokugqama ku-ajenda yomphakathi yaseYurophu. Okunye okuxoxwe ngakho kakhulu yi-phenomenon of the "ama-hallucinations": izikhathi lapho isistimu ikhiqiza khona izimpendulo eziqanjiwe, ezingalungile noma ezingalungile, kodwa ezethulwa ngomoya osobala wokuzethemba.
Le nkinga ayiyona i-OpenAI kuphela. Amanye amamodeli ereferensi, njenge I-Google GeminiClaude, Ukudideka noma i-GrokBabhekene nenselele efanayo. Nokho, ukwethemba abasebenzisi abaningi abakubeka kulawa mathuluzi kukhulisa umthelela ongaba khona wanoma yimaphi amaphutha, ikakhulukazi uma esetshenziselwa ukwenza izinqumo ezibucayi ngaphandle kokuqinisekisa ulwazi ngemithombo eyengeziwe.
Ukuze kuncishiswe le ngozi, izinkampani emkhakheni ziye zaqinisa ukuqeqeshwa kwamamodeli abo ayisisekelo futhi ifake izindlela ezivumela imithombo ekhonjiwe, ukuvuma ukulinganiselwa, noma ukwenqaba izicelo ezinobungozi obukhulu. Esimeni se-OpenAI, izinguqulo zakamuva ze-ChatGPT cishe zizovuma ukuthi azikwazi ukuqedela umsebenzi ngeziqinisekiso noma ukuqondisa kabusha umsebenzisi ekusekelweni komuntu ezimeni ezibucayi.
Okunye ukukhathazeka okukhulayo kuhlobene ne impilo yengqondo nokuncika ngokomzweloIfomethi yengxoxo, ehlanganiswe nokutholakala njalo kwesevisi, iholele abanye abasebenzisi ukuthi baphathe i-chatbot njengomngane, isifuba, noma umeluleki womuntu siqu. Lokhu kungaba usizo ngezinye izikhathi, kodwa kuphinde kube nezingozi lapho ukwesekwa kochwepheshe noma amanethiwekhi okunakekelwa kwangempela ethathelwa indawo ukuxhumana ne-AI.
I-OpenAI ithi i-GPT-5 nabalandelayo bayakwazi bona kangcono izimpawu zokucindezeleka ngokomzweloukunciphisa izimpendulo ezingase zibe yingozi kanye nokugcina izivikelo ngisho nasezingxoxweni ezinde kakhulu, ebeziyiphuzu elibuthakathaka ngokomlando. Njengengxenye yalo mzamo, inkampani ithuthukise izici zokuthi Ukugadwa kwabazali nemikhawulo ethile yentsha, futhi usebenza kumasistimu okuqinisekisa ubudala ukuze avumelane nesipiliyoni sezingane.
I-copyright, idatha kanye ne-cybersecurity endaweni ebonakalayo
Omunye umugqa omkhulu wenkulumompikiswano uzungeza idatha esetshenziselwa ukuqeqesha amamodeliUkuze kuzuzwe ukusebenza kwamanje, lawa manethiwekhi e-neural adinga amanani amakhulu ombhalo, izithombe, umsindo nevidiyo, kwesinye isikhathi atholwa ku-inthanethi ngaphandle kokulawula okuphelele kokuthi inelungelo lobunikazi noma likhululekile ukusetshenziswa.
EYurophu, lapho imithetho yempahla yengqondo iqine kakhulu, lolu daba luye lwadala ukungezwani nemithombo yezindaba, abashicileli, nabadaliabafuna ukuba sobala mayelana nokuthi yikuphi okuqukethwe okusetshenzisiwe futhi ngaphansi kwaziphi izimo. Ingxoxo ihlangana nokuvunywa kwemithethonqubo efana ne Umthetho we-European Union AI, okudinga abahlinzeki bamamodeli akhiqizayo ukuthi banikeze ulwazi olwengeziwe mayelana nedatha yabo yokuqeqeshwa.
Kwengezwe kulokhu yisahluko esithi ukuphepha kwe-cyberKusukela ezinyangeni ezimbalwa zokuqala kutholwe ukuthi i-ChatGPT ingasetshenziselwa yona khiqiza izingcezu zekhodi ezinonyaLokhu kuhlanganisa ukubhala ama-imeyili obugebengu bokweba imininingwane ebucayi noma ukuthuthukisa imikhankaso eyinkimbinkimbi yokubulala ama-disinformation. Nakuba i-OpenAI nezinye izinkampani zethule imigoqo yobuchwepheshe nezihlungi zokuqukethwe, izigebengu ze-inthanethi zihlale zifuna izindlela zokweqa le mikhawulo.
Ukusetshenziswa kabi akugcini ngekhodi. Ukukhiqiza okuzenzakalelayo kwe ama-deepfakes alalelwayo namavidiyoLezi zixhobo, ezikwazi ukulingisa amazwi nobuso babantu abavelele emphakathini, zivula umnyango wokukhwabanisa, ukuqola, izithombe zobulili ezingcolile okungavunyelwana ngazo, kanye nokukhohlisa kwezombusazwe. Kulo mongo, ochwepheshe bezokuphepha abavela emazweni ahlukahlukene aseYurophu baxwayisa ngesidingo sokuqinisa kokubili ukufunda ngedijithali phakathi kwezakhamuzi kanye namathuluzi okuthola okuqukethwe kokwenziwa.
Ngaphandle kwalezi zingozi, imboni igcizelela ukuthi izinzuzo ezingaba khona ze-generative AI zinkulu uma isetshenziswa ngemibandela ezwakalayo. umthwalo wemfanelo, ukwenza izinto obala kanye nokwengamela komuntuOkubalulekile, abaningi bayavuma, kuyoba ukuthola ukulingana phakathi kokusungula izinto ezintsha nokuvikelwa kwamalungelo ayisisekelo.
Umdlali omaphakathi emlandweni wakamuva wobuhlakani bokwenziwa
Ngaphandle kwendima yayo njengomkhiqizo wezohwebo, i-ChatGPT isivele isendaweni evelele ku- umlando wesimanje wobuhlakani bokwenziwaKwakuyisimiso esaqala ukuletha i-AI ekhiqizayo emphakathini jikelele, nge-interface elula nesiphakamiso esicacile: ukuxoxa nomshini njengokungathi ungumuntu, ngaphandle kokudinga ukuba uchwepheshe wezobuchwepheshe.
Lokho kunyakaza kwaqala a ukuqhuma kwezinhlobo zezilimi ezinkulu Lobu buchwepheshe buvela ezinkampanini zaseMelika nezaseShayina, kanye nasezinhlelweni ezivulekile namaphrojekthi aseYurophu afuna ukunciphisa ukuncika kubahlinzeki bangaphandle. Selokhu zavela, imiqondo efana ne-generative AI, i-AI eguquguqukayo, namamodeli e-AI ayisisekelo asukile emalabhorethri aya ezihlokweni zamaphephandaba kanye nezinkulumo-mpikiswano zasephalamende.
Eminyakeni emithathu yokuqala, uhlelo lubonise kokubili amathuba amakhulu okuzenzela ingxenye yomsebenzi wobuhlakani Okufana nezinkinga ezivelayo ekunikezeni imisebenzi ebucayi kuma-algorithms angahlali elungile. Kusukela kwezemfundo kuya ekunakekelweni kwezempilo, okuhlanganisa ukuphathwa komphakathi kanye nezimboni zokudala, imikhakha emhlabeni wonke iyazibuza ukuthi ingakuhlanganisa kanjani ngaphandle kokulahlekelwa umbono wezingozi.
Okwaqala njenge-AI “ebuncaneni bayo,” ekwazi ngokuyinhloko ukubhala imibhalo enelukuluku, sekuphenduke a ingqalasizinda yedijithali yansuku zonke ezigidini zabantu eYurophu nasemhlabeni wonke. Ukuziphendukela kwemvelo kwayo kuyaqhubeka ngesivinini esikhulu futhi kunzima ukulindela ukuthi kuzohamba kude kangakanani, kodwa umthelela wayo endleleni esisebenza ngayo, esifunde ngayo nesixhumana ngayo kakade usuyingxenye yeqiniso lethu lansuku zonke.
Njengoba izinto zimi, umuzwa ojwayelekile uwukuthi Ubuhlakani bokwenziwa kwengxoxo abasesona nje isithembiso sekusasa. ukuze sibe isici esizinzile samanje, ne-ChatGPT njengesinye sezinkomba zayo eziyinhloko: ithuluzi elinamandla neliguquguqukayo, eligcwele amathuba, elisiphoqa ukuthi sicabange kabusha ukuthi sifuna ukuphila kanjani nobuchwepheshe eminyakeni ezayo.
