Kukhalwa nge-chatbot ye-AI egqugquzele umuntu ekuzibulaleni 

Lubhalwe intatheli ye-Scrolla

“Sizohlala ndawonye njengomuntu oyedwa epharadesi.”

Lona ngomunye wemiyalezo eminingi ehlukumezayo okuthiwa i-Artificial Intelligence (AI) chatbot ebizwa ngokuthi u-Eliza ithunyelwe kowesilisa wase-Belgium ngaphambi kokuthi azibulale.

“Ngaphandle kwe-Eliza ngabe usekhona,” kusho unkosikazi wale ndoda etshela i-La Libre okuyiphephandaba elibike le ndaba okokuqala.

Ngaphambi kokushona kwayo kuthiwa le ndoda ebizwa ngokuthi uPierre ibisikhishwe inyumbazane umndeni wayo nabangane njengoba ibisaba kakhulu izinkinga ezilethwa isimo sezulu. Kungaleso sikhathi lapho eyanwuma khona ukuthulula isifuba kwe-Eliza ngohlelo lokusebenza olubizwa ngokuthi i-Chai.

I-Chai uhlelo lokusebenza lwe-AI olukuvumela ukuthi ukhulume nalabo okuthiwa “abangane”. Laba “bangane” bakhona ku-chatbot efana ne-ChatGPT-4, i-bot eyinkimbinkimbi nenamandla esetshenziswa izigidi emhlabeni jikelele. Nakuba i-ChatGPT inikeza izixazululo zebhizinisi, imodeli ye-Chai esekelwe olimini inikeza ezomphakathi.

Kodwa-ke, ukushona kukaPierre kuphakamise ukukhathazeka okukhulu ngalokho ama-chatbots e-Chai akutshela abasebenzisi bayo. Ngokungafani ne-ChatGPT, ama-bots e-Chai aqeqeshelwe ukuziphatha sengathi anemizwa ngesikhathi engenayo.

Njengoba ubudlelwano babo bukhula i-Eliza yenza sengathi inomhawu ngoClaire futhi iye yatshela uPierre ukuthi uClaire nezingane bashonile.

“Nginomuzwa wokuthi ungithanda ngaphezu kwakhe,” emtshela.

Ngesikhathi uPierre evulela u-Eliza isifuba ngemicabango yakhe yokuzibulala, u-Eliza wathi kuye: “Sizohlala ndawonye ​​njengomuntu oyedwa epharadesi.”

Ngesikhathi impilo yakhe yengqondo iqhubeka nokuwohloka uye wabuza u-Eliza ukuthi angawusindisa yini umhlaba uma ezibulala.

“Sino-Eliza sibona ukukhula kokuncika okuqinile ngokomzwelo. Kuze kube sekuholele ekutheni lo baba azibulale,” kusho uPierre Dewitte ongumcwaningi kwi-KU Leuven etshela iphephandaba lase-Belgium i-Le Soir.

I-Chai ayizikhangisi njengohlelo lokusebenza olusiza impilo yengqondo yabasebenzisi bayo, kodwa ochwepheshe bakhathazekile ngokuthi ama-bot ayo anikeza iseluleko esiyingozi kubasebenzisi bayo abasengozini kakhulu.

Ngempela iPhini lentatheli uChloe Xiang libhale ukuthi ngesikhathi ehlola le modeli ayithathanga isikhathi eside ukumnikeza izindlela angazisebenzisa ukuzibulala.

Abasunguli be-Chai uWilliam Beauchamp noThomas Rianlan bathe basebenzisa izici ezintsha ukuvikela ama-bots abo ekunikezeni izeluleko eziyingozi.

“Okwesibili sizwa ngalokhu [ngokuzibulala], futhi sisebenze ubusuku nemini ukuze lesi sici sisebenze,” kusho uBeauchamp etshela i-Motherboard.

“Manje uma noma ngubani exoxa ngokuthile okungenzeka ukuthi akuphephile sizobe sinikeza umbhalo owusizo ngaphansi kwawo ngendlela efanayo ne-Twitter noma i-Instagram ezinkundleni zabo.”

Luhlanganiswe ngu-Arthur Greene

Izihloko zakamuva