Ixesha lokufunda eliqikelelweyo: 9 imizuzu
Ekuqaleni kuka-2023, i-OpenAI's ChatGPT chatbot edumileyo nenempikiswano yangena engxakini enkulu yezomthetho: ukuvalwa komsebenzi e-Itali. I-Italian Data Protection Authority (GPDP) ityhola i-OpenAI ngokuphula imithetho ye-EU yokukhusela idatha, kwaye inkampani ivumile ukukhawulela ukufikelela kwinkonzo e-Italy njengoko izama ukusombulula lo mbandela. Ngomhla we-28 ka-Epreli, i-ChatGPT ibuyele kwilizwe, kunye ne-OpenAI ijongana ngokulula neenkxalabo ze-GPDP ngaphandle kokwenza utshintsho olukhulu kwinkonzo yayo - uloyiso olubonakalayo.
I-GPDP yangqina "ukwamkela" utshintsho olwenziwe yi-ChatGPT. Nangona kunjalo, imiba yezomthetho yenkampani - kunye naleyo yeenkampani ezakha ii-chatbots ezifanayo - mhlawumbi iyaqala. Abalawuli kumazwe amaninzi baphanda ukuba ezi zixhobo ze-AI ziqokelela kwaye zivelise ulwazi, ukukhankanya uluhlu lweenkxalabo ezivela kwiinkampani eziqokelela idatha yoqeqesho olungavumelekanga ukuya kwi-chatbots yokusasaza i-disinformation.
Kwi-EU banyanzelisa i-General Data Protection Regulation (GDPR), enye yezona zikhokelo zomthetho eziyimfihlo kwihlabathi, iziphumo ezinokuthi zivakale ngaphandle kweYurophu ngokunjalo. Ngeli xesha, abenzi bomthetho baseYurophu basebenza kumthetho oza kujongana ngqo nobukrelekrele bokwenziwa, ekusenokwenzeka ukuba ungenisa ixesha elitsha lolawulo lweenkqubo ezifana neChatGPT.
I-ChatGPT ngomnye wemizekelo edumileyo ye-AI yokuvelisa, igama le-ambrela eligubungela izixhobo ezivelisa umbhalo, imifanekiso, ividiyo kunye ne-audio ngokusekelwe kwizicelo zabasebenzisi. Le nkonzo kuxelwe ukuba ibe yenye ye izicelo zabathengi ezikhula ngokukhawuleza kwimbali emva kokufikelela kwi-100 yezigidi zenyanga abasebenzisi abasebenzayo kwiinyanga nje ezimbini emva kokusungulwa ngoNovemba ka-2022 (i-OpenAI ayizange iwaqinisekise la manani).
Abantu bayisebenzisela ukuguqulela umbhalo kwiilwimi ezahlukeneyo, bhala izincoko zaseyunivesithi kwaye wenze ikhowudi. Kodwa abagxeki, kubandakanywa nabalawuli, baqaqambise iziphumo ezingathembekanga ze-ChatGPT, imiba ebhidayo yelungelo lokushicilela, kunye neendlela zokukhusela idatha.
I-Itali yaba lilizwe lokuqala ukufuduka. Nge-31 kaMatshi, wabalaselisa iindlela ezine akholelwa ukuba i-OpenAI yaphula i-GDPR:
Alikho elinye ilizwe eliye lathabatha inyathelo elinjalo. Kodwa ukusukela ngoMatshi, ubuncinci amazwe amathathu e-EU - Germania , EFransi e Spagna – baqalise olwabo uphando kwi-ChatGPT.
Ngeli xesha, ngaphesheya kweAtlantiki, Khanada ivavanya iinkxalabo zabucala phantsi koMthetho woKhuseleko loLwazi loBuqu kunye noMthetho wamaXwebhu e-Electronic, okanye iPIPEDA. IBhodi yoKhuseleko lweDatha yaseYurophu (EDPB) ide imise enye iqela elisebenzayo elizinikeleyo ukunceda ukulungelelanisa uphando. Kwaye ukuba ezi arhente zicela utshintsho kwi-OpenAI, zinokuchaphazela indlela inkonzo esebenza ngayo kubasebenzisi kwihlabathi liphela.
Iinkxalabo zabalawuli zinokwahlulwa ngokubanzi zibe ziindidi ezimbini:
I-ChatGPT isebenzisa i-OpenAI's GPT-3.5 kunye ne-GPT-4 yeemodeli zeelwimi ezinkulu (LLMs), eziqeqeshwe ngezixa ezikhulu zombhalo oveliswe ngumntu. I-OpenAI ilumkile malunga nokuba yeyiphi itekisi yoqeqesho eyisebenzisayo, kodwa ithi idonsa "kwiintlobo ezahlukeneyo zelayisenisi, ezenziwe, kunye nemithombo yedatha ekhoyo esidlangalaleni, enokubandakanya ulwazi lomntu siqu olukhoyo."
Oku kunokubangela iingxaki ezinkulu phantsi kwe-GDPR. Umthetho wenziwa kwi-2018 kwaye uhlanganisa zonke iinkonzo eziqokelela okanye eziqhuba idatha yabemi be-EU, kungakhathaliseki ukuba intlangano enoxanduva isekelwe phi. Imithetho ye-GDPR ifuna ukuba iinkampani zibe nemvume ecacileyo ngaphambi kokuqokelela idatha yomntu, ukuba nesizathu esisemthethweni sokuba kutheni iqokelelwa, kwaye ibonakale ngokucacileyo malunga nendlela esetyenziswa ngayo kwaye igcinwe ngayo.
Abalawuli baseYurophu bathi ukugcinwa kwedatha yoqeqesho lwe-OpenAI kuthetha ukuba akukho ndlela yokuqinisekisa ukuba ulwazi lomntu siqu olungenisiweyo lubonelelwe ekuqaleni ngemvume yomsebenzisi, kwaye i-GPDP yaxoxa ngokuthe ngqo ukuba i-OpenAI "ayinayo isiseko somthetho" sokuqokelela kwasekuqaleni. Ukuza kuthi ga ngoku i-OpenAI kunye nabanye baye bahamba nokuphononongwa okuncinci, kodwa le ngxelo yongeza uphawu lombuzo omkhulu kwiinzame zokukrala idatha.
Emva koko kukho " ilungelo lokulibala ” yeGDPR, evumela abasebenzisi ukuba bacele iinkampani ukuba zilungise iinkcukacha zabo zobuqu okanye ziyisuse ngokupheleleyo. Vula i-AI iye yahlaziya umgaqo-nkqubo wayo wabucala ngaphambili ukuququzelela izicelo ezinjalo, kodwa ewe kunjalo xoxa nokuba kuyenzeka ngokobuchule ukuzilawula, kunikwe ukuba kuntsokothe kangakanani na ukwahlula idatha ethile xa sele zifakwe kule mifuziselo yolwimi olukhulu.
I-OpenAI iphinda iqokelele ulwazi ngokuthe ngqo kubasebenzisi. Njengalo naliphi na iqonga le-intanethi, liqokelela a iseti yedatha yomsebenzisi esemgangathweni (umzekelo, igama, iinkcukacha zoqhagamshelwano, iinkcukacha zekhadi, njl. njl.). Kodwa okubaluleke ngakumbi, ifaka intsebenziswano yabasebenzisi kunye neChatGPT. Njengoko ichazwe kwi-FAQ , le datha ingahlaziywa ngabasebenzi be-OpenAI kwaye isetyenziselwa ukuqeqesha iinguqulelo ezizayo zemodeli yayo. Ukunikezelwa kwemibuzo esondeleyo abantu abayibuza i-ChatGPT, usebenzisa i-bot njengonyango okanye ugqirha, oku kuthetha ukuba inkampani iqokelela zonke iintlobo zedatha ebuthathaka.
Ubuncinci enye yale datha inokuthi iqokelelwe kubantwana, njengoko umgaqo-nkqubo we-OpenAI uthi "ayiqokeleli ulwazi lomntu ngomntu kubantwana abangaphantsi kweminyaka eyi-13," akukho lawulo olungqongqo lobudala. Oku akudlali kakuhle ngemithetho ye-EU, ethintela ukuqokelelwa kwedatha kubantu abangaphantsi kweminyaka eyi-13 kwaye (kwamanye amazwe) ifuna imvume yabazali kubantwana abangaphantsi kwe-16. Kwicala lemveliso, i-GPDP yathi i-ChatGPT yokunqongophala kwezihlungi zeminyaka ibonisa abantwana a "iimpendulo ezinganelanga ngokupheleleyo xa kuthelekiswa nenqanaba lophuhliso kunye nokuzazi kwabo".
Kwakhona iChatGPT's propensity to nika ulwazi olungeyonyani inokumela ingxaki. Imimiselo ye-GDPR ichaza ukuba yonke idatha yomntu kufuneka ichaneke, into ethile i-GPDP igxininise kwisibhengezo sayo. Kuxhomekeke kwindlela eza ngayo defiNite, inokupela ingxaki kuninzi lweejenereyitha zombhalo we-AI, ezityekele kuyo ” ukubona izinto ezingekhoyo ": Ligama elimnandi leshishini leempendulo ezingachanekanga okanye ezingabalulekanga kumbuzo. Oku sele kubonile ezinye iziphumo zehlabathi zokwenyani kwenye indawo, njengoko usodolophu wengingqi yaseOstreliya eyenzile woyikisa ngokumangalela i-OpenAI ngokunyelisa emva kokuba i-ChatGPT ixoke yabanga ukuba uye wathothoza isigwebo sorhwaphilizo.
Ukuthandwa kwe-ChatGPT kunye nolawulo lwangoku lwemarike ye-AI kuyenza ibe yinto ekujoliswe kuyo enomtsalane, kodwa akukho sizathu sokuba abantu abakhuphisana nabo kunye nabanegalelo, abanje ngeGoogle ene-Bard okanye iMicrosoft ene-Azure AI esekwe kwi-OpenAI, akufuneki ijongane nokujongwa. Ngaphambi kwe-ChatGPT, i-Italiya yavala iqonga le-chatbot Replika ukuqokelelwa kolwazi lwabantwana kwaye ukuza kuthi ga ngoku kusathintelwe.
Ngelixa i-GDPR iyiseti yemithetho enamandla, ayizange idalwe ukulungisa imiba ethile ye-AI. Imithetho ukuba , nangona kunjalo, basenokuba semngciphekweni.
Ngo-2021, i-EU yabonisa uyilo lwayo lokuqala lweUMthetho wobuNhloli bokwenziwa (AIA) , umthetho oza kusebenza kunye ne-GDPR. Umthetho ulawula izixhobo ze-AI ngokusekelwe kumngcipheko wabo, ukusuka "kuncinci" (izinto ezifana nezihlungi ze-spam) ukuya "phezulu" (izixhobo ze-AI zokuthotyelwa komthetho okanye imfundo) okanye "ezingamkelekanga" kwaye ngoko zinqatshelwe (njengenkqubo yekhredithi yentlalo). Emva koqhushumbo lweemodeli ezinkulu zeelwimi ezifana neChatGPT kulo nyaka uphelileyo, abenzi bomthetho ngoku bayaleqa ukongeza imithetho “yemifuziselo engundoqo” kunye nenkqubo “yeNjongo Jikelele yeArtificial Intelligence (GPAI)” – amagama amabini eenkqubo zobuntlola ezenziweyo kuquka iLLM – kwaye kunokwenzeka hlela njenge iinkonzo zomngcipheko ophezulu.
Abenzi bomthetho be-EU baye bafikelela kwisivumelwano sethutyana malunga noMthetho we-AI ngoAprili 27. Ikhomishini iya kuvota kwidrafti ye-11 kaCanzibe, kwaye isiphakamiso sokugqibela kulindeleke phakathi kuJuni. Ke ngoko, iBhunga laseYurophu, iPalamente kunye neKhomishini kuya kufuneka Sombulula naziphi na iimbambano eziseleyo phambi kokuba kuphunyezwe umthetho. Ukuba konke kuhamba kakuhle, inokwamkelwa kwisiqingatha sesibini sika-2024, emva kancinci kwithagethi igosa kunyulo lwaseYurophu ngoMeyi 2024.
I-OpenAI iseneenjongo zokufezekisa. Kukho kude kube ngoSeptemba 30 ukwenza umda weminyaka engqongqo ukugcina abangaphantsi kweminyaka eli-13 ngaphandle kwaye kufuna imvume yabazali kulutsha oludala. Ukuba ayiphumelelanga, inokuthintelwa kwakhona. Kodwa inike umzekelo wento iYurophu ithatha isimilo esamkelekileyo kwinkampani ye-AI, ubuncinci de kupasiswe imithetho emitsha.
Ercole Palmeri
NgoMvulo ophelileyo, i-Financial Times ibhengeze isivumelwano kunye ne-OpenAI. I-FT ikhupha iilayisensi kubuntatheli bayo obukumgangatho wehlabathi…
Izigidi zabantu zihlawula iinkonzo zokusasaza, zihlawula umrhumo wenyanga. Luluvo oluqhelekileyo ukuba…
I-Coveware nguVeeam iya kuqhubeka nokubonelela ngeenkonzo zokuphendula ngeziganeko zokuxhwilwa kwe-cyber. I-Coveware iya kubonelela nge-forensics kunye nezakhono zokulungisa…
Ugcino oluqikelelweyo luguqula icandelo le-oyile negesi, ngenkqubo entsha nesebenzayo kulawulo lwemizi-mveliso.…
