Riaan 
U-Edward Olive/Getty Izithombe
I-ChatGPT ingase ibe edume kakhulu, futhi engaba yigugu, i-algorithm yamanje, kodwa amasu obuhlakani bokwenziwa asetshenziswa i-OpenAI ukuze ahlinzeke ngobuhlakani bayo awahlukile futhi awayimfihlo. Amaphrojekthi aqhudelanayo nama-clones omthombo ovulekile angase enze ama-bot esitayela se-ChatGPT atholakale kunoma ubani ukuze awakopishe futhi aphinde asetshenziswe.
Ukuzinza kwe-AI, isiqalisi esesivele sithuthukile futhi ubuchwepheshe obuthuthukisiwe bokukhiqiza izithombe esitholakala kumthombo ovulekile, sisebenzela imbangi evulekile ku-ChatGPT. “Sinezinyanga ezimbalwa sikhishwe,” kusho u-Emad Mostaque, oyi-CEO ye-Stability. Izinombolo zokuqala ezincintisanayo, kufaka phakathi I-Anthropic, Coherefuthi I-AI21basebenza kuma-chatbots obunikazi afana ne-OpenAI’s bot.
Izikhukhula ezizayo zama-chatbots ayinkimbinkimbi zizokwenza ubuchwepheshe buchichime kakhudlwana futhi bubonakale kubathengi, futhi bufinyeleleke kalula emabhizinisini e-AI, onjiniyela, nabacwaningi. Lokho kungasheshisa ukujaha ukwenza imali ngamathuluzi e-AI akhiqiza izithombe, ikhodi, nombhalo.
Izinkampani ezimisiwe ezifana ne-Microsoft ne-Slack zifaka i-ChatGPT emikhiqizweni yabofuthi iziqalo eziningi ziyaphithizela ukwakha phezulu i-ChatGPT API entsha yonjiniyela. Kodwa ukutholakala okubanzi kobuchwepheshe kungase futhi kwenze kube nzima imizamo yokubikezela nokunciphisa ubungozi obuhambisana nakho.
Ikhono elikhohlisayo le-ChatGPT lokuhlinzeka ngezimpendulo ezikholisayo emibuzweni eyahlukene libuye libangele ukuthi ngezinye izikhathi akha amaqiniso noma uthole abantu abanenkinga. Ingasiza ngemisebenzi enonya efana nokukhiqiza ikhodi yohlelo olungayilungele ikhompuyutha, noma imikhankaso yogaxekile kanye ne-disinformation.
Ngenxa yalokho, abanye abacwaningi bacele ukuthi kuncishwe ukuthunyelwa kwamasistimu afana ne-ChatGPT ngenkathi kuhlolwa ubungozi. “Asikho isidingo sokumisa ucwaningo, kodwa ngokuqinisekile singalawula ukuthunyelwa okusabalele,” kusho uGary Marcus, uchwepheshe we-AI ofune ukudonsela ukunaka ezingozini ezifana ne-disinformation ekhiqizwa yi-AI. “Singahle, ngokwesibonelo, sicele izifundo kubantu abayi-100,000 ngaphambi kokudedela lobu buchwepheshe ezigidini eziyi-100 zabantu.”
Ukutholakala okubanzi kwezinhlelo zesitayela se-ChatGPT, nokukhishwa kwezinguqulo zomthombo ovulekile, kungenza kube nzima ukukhawulela ucwaningo noma ukusetshenziswa okubanzi. Futhi ukuncintisana phakathi kwezinkampani ezinkulu nezincane ukwamukela noma ukufanisa i-ChatGPT kuphakamisa isifiso esincane sokunciphisa ijubane, kodwa kubonakala sengathi kugqugquzela ukwanda kobuchwepheshe.
Ngesonto eledlule, i-LLaMA, imodeli ye-AI eyakhiwe yi-Meta—futhi efana naleyo eyisisekelo se-ChatGPT—iputshuzwe ku-inthanethi ngemva kokwabelana nabanye abacwaningi bezemfundo. Uhlelo lungasetshenziswa njengebhlokhi yokwakha ekudalweni kwe-chatbot, kanye nokukhululwa kwayo kwavusa ukukhathazeka phakathi kwalabo abesaba ukuthi amasistimu e-AI aziwa njengamamodeli amakhulu ezilimi, nama-chatbots akhelwe phezu kwawo njenge-ChatGPT, azosetshenziselwa ukukhiqiza ulwazi olungelona iqiniso noma ukwenza ngokuzenzakalelayo ukwephulwa kwe-cybersecurity. Abanye ochwepheshe bayakuphikisa lokho izingozi ezinjalo zingase zeqiwenabanye basikisela ukuthi ukwenza ubuchwepheshe bubonakale kakhudlwana empeleni kuzosiza abanye ukuba bagweme ukusetshenziswa kabi.
