Izitholi ze-AI, ngezinye izikhathi ezibizwa ngokuthi yi-AI yokubhala noma izitholi zokuqukethwe ze-AI, zisebenzisa inhloso yokuhlonza ukuthi umbhalo ubhalwe ngokwengxenye noma ngokugcwele ngamathuluzi obuhlakani bokwenziwa afana I-ChatGPT.
Lezi zithungatha ziwusizo ekuhlonzeni izimo lapho ucezu olubhaliwe kungenzeka ludalwe yi-AI. Isicelo sinenzuzo ngezindlela ezilandelayo:
- Ukuqinisekisa umsebenzi wabafundi. Othisha bangayisebenzisa ukuze baqinisekise ubuqiniso bezabelo zangempela zabafundi namaphrojekthi okubhala.
- Ukulwa nokubuyekezwa kwemikhiqizo mbumbulu. Omengameli bangayisebenzisa ukuze bahlonze futhi babhekane nezibuyekezo zomkhiqizo womgunyathi ezihlose ukukhohlisa umbono womthengi.
- Ukubhekana nokuqukethwe kogaxekile. Isiza ekutholeni nasekususeni izinhlobo ezahlukene zokuqukethwe ogaxekile okungahlanekezela ikhwalithi nokwethembeka kwezinkundla eziku-inthanethi.
Lawa mathuluzi asemisha futhi ayahlolwa, ngakho-ke asinaso isiqiniseko esiphelele sokuthi athembeke kangakanani njengamanje. Ezigabeni ezilandelayo, sihlolisisa ukusebenza kwazo, sihlole ukuthi zingathenjwa kangakanani, futhi sihlole uhla lwezinhlelo zokusebenza ezingokoqobo ezinikezayo.
| Izikhungo zemfundo, okuhlanganisa namanyuvesi, ziphezu kwenqubo yokubumba izikhundla zazo mayelana nokusetshenziswa okufanele kwe-ChatGPT namathuluzi afanayo. Kubalulekile ukubeka phambili imihlahlandlela yesikhungo sakho kunanoma yisiphi iseluleko osithola ku-inthanethi. |
Asebenza kanjani ama-AI detectors?
Izitholi ze-AI zivame ukusebenzisa amamodeli olimi afana nalawo akumathuluzi okubhala we-AI ezizama ukuwathola. Ngokuyisisekelo, imodeli yolimi ibheka okokufaka bese ibuza, "Ingabe lokhu kubukeka njengento okungenzeka ngiyenzile?" Uma ithi yebo, imodeli iqagela ukuthi umbhalo cishe udalwe yi-AI.
Ikakhulukazi, lawa mamodeli acinga izici ezimbili ngaphakathi kombhalo: "ukudideka" kanye "nokuqhuma." Uma lezi zici ezimbili ziphansi, maningi amathuba okuthi umbhalo ukhiqizwe yi-AI.
Nokho, asho ukuthini ngempela la magama angajwayelekile?
Ukuxakeka
Ukuxakeka kuma njengemethrikhi ebalulekile esetshenziswa ekuhloleni ubuhlakani bamamodeli olimi. Kubhekiselwa ekutheni imodeli ikwazi kangakanani ukubikezela igama elilandelayo ngokulandelana kwamagama.
Amamodeli olimi lwe-AI asebenzela ukudala imibhalo enokudida okuphansi, okuholela ekwandeni kokuhambisana, ukugeleza okushelelayo, nokubikezela. Ngokuphambene, ukubhala kwabantu kuvame ukukhombisa ukudideka okuphezulu ngenxa yokusebenzisa kwakho izinketho zolimi ezicatshangwayo, nakuba kuhambisana nemvamisa enkulu yamaphutha okubhala.
Izibonelo zolimi zisebenza ngokubikezela ukuthi yiliphi igama elizolandela ngokwemvelo emshweni bese belifaka. Ungabona isibonelo ngezansi.
| Ukuqhubeka kwesibonelo | Ukuxakeka |
| Angikwazanga ukuqeda iphrojekthi ekugcineni ubusuku. | Okuphansi: Mhlawumbe ukuqhubeka okungenzeka kakhulu |
| Angikwazanga ukuqeda iphrojekthi ekugcineni isikhathi angiphuzi ikhofi kusihlwa. | Okuphansi kuya kokuphakathi: Amathuba amancane, kodwa kwenza umqondo wohlelo lolimi nonengqondo |
| Angikwazanga ukuqeda iphrojekthi isemester yokugcina izikhathi eziningi ngenxa yendlela engangingenamdlandla ngayo ngaleso sikhathi. | Phakathi: Umusho uyahambisana kodwa wakheke ngendlela engajwayelekile futhi unomoya omude |
| Angikwazanga ukuqeda iphrojekthi ekugcineni ngijabulela ukukwazi. | Okuphakeme: Uhlelo lolimi alulungile futhi alunangqondo |
Ukudideka okuphansi kuthathwa njengobufakazi bokuthi umbhalo ukhiqizwa nge-AI.
Ukuvutha
“Ukuvutha” kuyindlela yokubona ukuthi imisho ihluke kanjani endleleni ehlanganiswe ngayo nokuthi yinde kangakanani. Kufana nokudideka kodwa kwemisho yonke esikhundleni samagama nje.
Uma umbhalo ngokuvamile unemisho efana nendlela enziwe ngayo nokuthi mide kangakanani, unokuqhuma okuphansi. Lokhu kusho ukuthi ifundeka ngokushelela. Kodwa uma umbhalo unemisho ehluke kakhulu kwenye ngokuthi yakhiwe kanjani nokuthi mide kangakanani, iba nokuqhuma okukhulu. Lokhu kwenza umbhalo uzizwe uzinzile futhi uhluke kakhulu.
Umbhalo okhiqizwe nge-AI uvame ukuhluka kancane emaphethini awo emisho uma uqhathaniswa nombhalo obhalwe ngumuntu. Njengoba amamodeli olimi eqagela igama okungenzeka ukuthi elilandelayo, ngokuvamile enza imisho engamagama ayi-10 kuya kwangama-20 ubude futhi alandele amaphethini avamile. Kungakho ukubhala kwe-AI kwesinye isikhathi kungabonakala kuyisidina.
Ukuqhuma okuphansi kubonisa ukuthi umbhalo kungenzeka ukhiqizwe nge-AI.
Enye Inketho Ongayicabangela: Ama-Watermark
I-OpenAI, umdali we-ChatGPT, kubikwa ukuthi yenza indlela ebizwa ngokuthi “i-watermarking.” Lolu hlelo lubandakanya ukwengeza uphawu olungabonakali embhalweni okhiqizwe ithuluzi, okungase kubonakale kamuva ngolunye uhlelo ukuze kuqinisekiswe umsuka we-AI wombhalo.
Nokho, lolu hlelo lusathuthukiswa, futhi imininingwane eqondile yokuthi luzosebenza kanjani ayikavezwa. Ngaphezu kwalokho, akucaci ukuthi noma imaphi ama-watermark aphakanyisiwe azohlala eqinile uma ukuhlelwa kwenziwa embhalweni okhiqiziwe.
| Yize umqondo wokusebenzisa lo mqondo ukuthola i-AI esikhathini esizayo ubukeka unethemba, kubalulekile ukuqaphela ukuthi imininingwane eqondile neziqinisekiso mayelana nokuyisebenzisa kusalindile. |
Kuyini ukuthembeka kwama-AI detectors?
- Izitholi ze-AI ngokuvamile zisebenza ngempumelelo, ikakhulukazi ngemibhalo emide, kodwa zingase zibe nezinkinga uma umbhalo odalwe yi-AI wenziwa ngamabomu ungalindelekile noma ushintshwa ngemva kokwenziwa.
- Izitholi ze-AI zingase zicabange ngephutha ukuthi umbhalo obhalwe abantu wenziwe yi-AI, ikakhulukazi uma uhlangabezana nezimo zokuba nokudideka okuphansi nokuqhuma.
- Ucwaningo mayelana nezitholi ze-AI lubonisa ukuthi alikho ithuluzi elinganikeza ukunemba okuphelele; ukunemba okuphezulu kakhulu kube ngu-84% ethuluzini leprimiyamu noma u-68% ethuluzini elingcono kakhulu lamahhala.
- Lawa mathuluzi anikeza imininingwane ebalulekile mayelana nokuba nokwenzeka kombhalo okhiqizwe nge-AI, kodwa sincoma ukuthi singanciki kuwo kuphela njengobufakazi. Ngokuqhubeka okuqhubekayo kwamamodeli olimi, amathuluzi awatholayo azodinga ukusebenza kanzima ukuze ahambisane.
- Abahlinzeki abazethemba kakhulu ngokuvamile bavuma ukuthi amathuluzi abo awakwazi ukusebenza njengobufakazi obuphelele bombhalo okhiqizwe yi-AI.
- Okwamanje, amanyuvesi awawathembi la mathuluzi.
| Ukuzama ukufihla ukubhala okukhiqizwa yi-AI empeleni kungenza umbhalo ubonakale uyinqaba kakhulu noma ungalungile ngokusetshenziswa kwawo okuhlosiwe. Isibonelo, ukwethula ngamabomu amaphutha esipelingi noma ukusebenzisa ukukhethwa kwamagama angenangqondo embhalweni kungase kunciphise amathuba okuthi akhonjwe yisitholi se-AI. Kodwa-ke, umbhalo ogcwele lawa maphutha nokukhetha okungajwayelekile ngeke ubonakale njengokubhala okuhle kwezemfundo. |
Kusetshenziswa ziphi inhloso izithungatha ze-AI?
Izitholi ze-AI zenzelwe abantu abafuna ukuqinisekisa ukuthi umbhalo ngabe udalwe ngobuhlakani bokwenziwa. Abantu abangayisebenzisa yilaba:
- Othisha nothisha. Ukuqinisekisa ubuqiniso bomsebenzi wabafundi kanye nokuvimbela ukukopela.
- Abafundi bahlola imisebenzi yabo. Ukuhlola ukuqinisekisa ukuthi okuqukethwe kwabo kuhlukile futhi akubukeki ngokungenhloso njengombhalo okhiqizwe i-AI.
- Abashicileli nabahleli babuyekeza okuthunyelwe. Ufuna ukuqinisekisa ukuthi bashicilela okuqukethwe okubhalwe abantu kuphela.
- Abacwaningi. bafuna ukuthola noma yimaphi amaphepha noma izindatshana ezingaba khona zocwaningo ezenziwe nge-AI.
- Ama-Blogger nababhali: Ufisa ukushicilela okuqukethwe okukhiqizwe yi-AI kodwa ukhathazeke ngokuthi kungase kubekwe phansi ezinjinini zokusesha uma kuqashelwa njengokubhala kwe-AI.
- Ochwepheshe ekulinganisweni kokuqukethwe. Ukuhlonza ogaxekile abakhiqizwe yi-AI, ukubuyekezwa okungelona iqiniso, noma okuqukethwe okungalungile.
- Amabhizinisi aqinisekisa okuqukethwe kokumaketha kwangempela. Ukuqinisekisa lokho okubalulekile kokuphromotha akulona iphutha nombhalo okhiqizwe yi-AI, okugcina ukwethembeka komkhiqizo.
| Ngenxa yokukhathazeka ngokwethembeka kwabo, abasebenzisi abaningi bayangabaza ukuncika ngokuphelele kuzitholi ze-AI okwamanje. Kodwa-ke, lezi zitholi sezivele zithandwa kakhulu njengophawu lokuthi umbhalo ungahle ukhiqizwe nge-AI, ikakhulukazi uma umsebenzisi esenokungabaza kwakhe. |
Ukutholwa mathupha kombhalo Okhiqizwe nge-AI
Ngaphandle kokusebenzisa izitholi ze-AI, ungafunda futhi ukukhomba izici ezihlukile zokubhala kwe-AI uwedwa. Akulula ngaso sonke isikhathi ukwenza lokhu ngokwethembeka—ukubhala komuntu ngezinye izikhathi kungase kuzwakale njengerobhothi, futhi ukubhala kwe-AI kuba ngabantu abakholisayo—kodwa ngokuzijwayeza, ungakwazi ukuhlakulela umqondo omuhle ngakho.
Imithetho ethize elandelwa izitholi ze-AI, njengokudideka okuphansi kanye nokuqhuma, ingabonakala iyinkimbinkimbi. Nokho, ungazama ukuzitholela lezi zici ngokwakho ngokubheka umbhalo ukuze uthole izimpawu ezithile:
- Lokho kufundeka njengokukodwa, kunokuhluka okuncane ekwakhekeni komusho noma ubude
- Ukusebenzisa amagama alindelekile futhi angahlukile kakhulu, futhi anezakhi ezimbalwa kakhulu ezingalindelekile
Ungasebenzisa futhi izindlela i-AI detectors engazisebenzisi, ngokuqaphela:
| izindlela | incazelo |
| Inhlonipho eyeqile | Ama-Chatbots afana ne-ChatGPT enziwa ukuthi abe abasizi abawusizo, ngakho-ke avame ukusebenzisa ulimi olunesizotha noluhlelekile olungase lungazwakali lukhululekile. |
| Ukungahambisani nezwi | Uma ujwayelene nendlela umuntu avame ukubhala ngayo (njengomfundi), ungakwazi ukubona lapho into ethile ayibhalile ihluke kakhulu kusitayela sakhe esivamile. |
| Ulimi lokuvikela | Naka ukuthi ayikho yini imibono eminingi eqinile nemisha, futhi qaphela uma kunomkhuba wokusebenzisa imishwana ekhombisa ukungaqiniseki kakhulu: “Kubalulekile ukuqaphela ukuthi …” “X ithathwa kabanzi ngokuthi …” “X ithathwa… ” “Abanye abantu bangase baphikise ukuthi …”. |
| Izimangalo ezingenamthombo noma ezicashunwe ngokungalungile | Uma kuziwa ekubhaleni kwezemfundo, kubalulekile ukusho ukuthi ulutholephi ulwazi lwakho. Nokho, amathuluzi okubhala e-AI ngokuvamile awawulandeli lo mthetho noma enze amaphutha (njengokucaphuna imithombo engekho noma engabalulekile). |
| Amaphutha anengqondo | Noma ukubhala kwe-AI kuba ngcono ekukhaleni kwemvelo, kwesinye isikhathi imibono ekuyo ayihlangani kahle. Naka izindawo lapho umbhalo usho izinto ezingahambisani, ezizwakala zingenakwenzeka, noma imibono engaxhumani kahle. |
| Sekukonke, ukuhlola amathuluzi okubhala e-AI ahlukahlukene, ukubuka izinhlobo zemibhalo abangazikhiqiza, nokujwayelana nendlela ebhala ngayo kungakusiza ube ngcono ekuboneni umbhalo okungenzeka ukuthi udalwe yi-AI. |
Izihloli zezithombe ze-AI namavidiyo
Izithombe ze-AI namajeneretha wevidiyo, ikakhulukazi adumile njenge-DALL-E ne-Synthesia, angakha okubonakalayo okungokoqobo nokuguquliwe. Lokhu kwenza kube semqoka ukuhlonza “ama-deepfakes” noma izithombe namavidiyo enziwe nge-AI ukuze kunqandwe ukusabalala kolwazi olungamanga.
Njengamanje, izimpawu eziningi zingadalula izithombe namavidiyo akhiqizwe yi-AI, njenge:
- Izandla ezineminwe eminingi kakhulu
- Ukunyakaza okungajwayelekile
- Umbhalo ongenangqondo esithombeni
- Izici zobuso ezingenangqondo
Nokho, ukubona lezi zimpawu kungase kube nzima njengoba i-AI iba ngcono.
Kunamathuluzi aklanyelwe ukuthola lezi zithombe ezikhiqizwe yi-AI, okuhlanganisa:
- I-Deepware
- I-Intel FakeCatcher
- I-Illuminarty
Akukacaci ukuthi la mathuluzi asebenza futhi athembeke kangakanani, ngakho ukuhlolwa okwengeziwe kuyadingeka.
Ukuvela okuqhubekayo kokukhiqizwa kwesithombe nevidiyo ye-AI kanye nokutholwa kudala isidingo esiqhubekayo sokuthuthukisa izindlela zokuthola eziqinile nezinembile ukuze kubhekwane nezingozi ezingaba khona ezihlobene nama-deepfakes kanye nokubonwayo okukhiqizwa yi-AI.
Isiphetho
| Izitholi ze-AI zisiza ukukhomba imibhalo ekhiqizwe amathuluzi afana ne-ChatGPT. Ngokuyinhloko babheka "ukudideka" kanye "nokuqhuma" ukuze babone okuqukethwe okudalwe yi-AI. Ukunemba kwazo kusalokhu kukhathaza, ngisho nezingcono kakhulu zibonisa amaphutha. Njengoba ubuchwepheshe be-AI buthuthuka, ukwehlukanisa abantu kokuqukethwe okukhiqizwe yi-AI, okuhlanganisa izithombe namavidiyo, kuba nzima, okugqamisa isidingo sokuhlala ucophelele ku-inthanethi. |
Imibuzo evame ukubuzwa
| 1. Uyini umehluko phakathi I-AI Detectors futhi Ama-Plagiarism Checkers? A: Kokubili izitholi ze-AI kanye nabahloli bokukopela bathola ukusetshenziswa emanyuvesi ukuze banqande ukungathembeki ezifundweni, nokho bayehluka ngezindlela nezinjongo zabo: • Izitholi ze-AI zihlose ukuhlonza umbhalo ofana nokuphumayo kumathuluzi okubhala e-AI. Lokhu kuhilela ukuhlaziya izici zombhalo ezifana nokudideka nokuqhuma, kunokuba uziqhathanise nesizindalwazi. • Izihloli zokukopa zihlose ukuthola umbhalo okopishiwe kweminye imithombo. Bafinyelela lokhu ngokuqhathanisa umbhalo nesizindalwazi esibanzi sokuqukethwe okushicilelwe ngaphambilini kanye nemibono yomfundi, bekhomba ukufana—ngaphandle kokuncika ekuhlaziyeni izici ezithile zombhalo. 2. Ngingayisebenzisa kanjani i-ChatGPT? A: Ukuze usebenzise i-ChatGPT, mane udale i-akhawunti yamahhala: • Landela lesi isixhumanisi kuwebhusayithi ye-ChatGPT. • Khetha okuthi “Bhalisa” bese unikeza ulwazi oludingekayo (noma sebenzisa i-akhawunti yakho ye-Google). Ukubhalisa nokusebenzisa ithuluzi kumahhala. • Thayipha ukwaziswa ebhokisini lengxoxo ukuze uqalise! Inguqulo ye-iOS yohlelo lokusebenza lwe-ChatGPT iyafinyeleleka okwamanje, futhi kunezinhlelo zohlelo lokusebenza lwe-Android ezisendleleni. Uhlelo lokusebenza lusebenza ngendlela efanayo kuwebhusayithi, futhi ungasebenzisa i-akhawunti efanayo ukuze ungene ngemvume kuzo zombili izinkundla. 3. I-ChatGPT izohlala imahhala kuze kube nini? A: Ukutholakala kwesikhathi esizayo kwe-ChatGPT mahhala kuhlala kungaqiniseki, ngaphandle komugqa wesikhathi othize omenyezelwe. Ithuluzi lethulwa okokuqala ngoNovemba 2022 “njengokubuka kuqala kocwaningo” okuzohlolwa isizinda somsebenzisi esibanzi ngaphandle kwenkokhiso. Igama elithi "ukuhlola kuqala" liphakamisa izindleko zesikhathi esizayo ezingaba khona, kodwa asikho isiqinisekiso esisemthethweni sokuqeda ukufinyelela kwamahhala esikhona. Inketho ethuthukisiwe, i-ChatGPT Plus, ibiza u-$20/ngenyanga futhi ihlanganisa izici ezithuthukisiwe njenge-GPT-4. Akukacaci ukuthi le nguqulo ye-premium izothatha indawo yamahhala noma uma eyokugcina izoqhubeka. Izinto ezifana nezindleko zeseva zingaba nomthelela kulesi sinqumo. Inkambo yesikhathi esizayo ayiqinisekile. 4. Ingabe kulungile ukufaka i-ChatGPT kwezingcaphuno zami? A: Kwezinye izimo, kufanelekile ukubhekisela ku-ChatGPT emsebenzini wakho, ikakhulukazi uma isebenza njengomthombo obalulekile wokufunda amamodeli olimi lwe-AI. Amanyuvesi athile angase adinge ukucashunwa noma ukugunyazwa uma i-ChatGPT isizile ucwaningo lwakho noma inqubo yokubhala, njengokwenza imibuzo yocwaningo; kuyatuseka ukuthi ubheke imihlahlandlela yesikhungo sakho. Kodwa-ke, ngenxa yokwethenjwa okuhlukahlukene kwe-ChatGPT kanye nokuntula ukwethembeka njengomthombo, kuhle ukuthi ungayishiyi ukuze uthole ulwazi oluyiqiniso. Ngesitayela se-APA, ungaphatha impendulo ye-ChatGPT njengokuxhumana komuntu siqu njengoba izimpendulo zayo zingafinyeleleki kwabanye. Embhalweni, yisho kanje: (ChatGPT, ukuxhumana komuntu siqu, February 11, 2023). |
