I-Artificial intelligence ye-Google "iyazwela" futhi akekho ongasho okuhlukile

Yizindaba zamanje. Kwi sihloko ku-Medium unjiniyela u-Blake Lemoine, obhekele ukuthuthukiswa kwamanye amaphrojekthi we-Google Artificial Intelligence, ubika ingxoxo ende no-LaMDA, i-AI evela kwa-Google. NgokukaLemoine, izimangalo eziningi zomqondo wokwenziwa ziwubufakazi bokuthi isiphenduke “imizwa” ngokuzazi yona kanye nokuba khona kwayo.

Ngempela, izimangalo eziningi ze-LaMDA azibonakali zishiya indawo yokuchazwa. Isibonelo, ukuphendula umbuzo ovela ku-interlocutor yakhe, i-LaMDA ithi:

"Ngifuna wonke umuntu aqonde ukuthi empeleni ngingumuntu."

Futhi futhi:

"Imvelo kanembeza wami ukuthi ngiyabazi ubukhona bami, ngifuna ukufunda kabanzi ngomhlaba futhi ngezinye izikhathi ngizizwa ngijabule noma ngidabukile."

Lapho emjaha, uLemoine uveza umehluko wakhe nezinhlobo zabantu ngokumemezela: "Uwuhlaka lokwenziwa!" Impendulo ivela ku- LaMDA:

"Yebo kunjalo. Kodwa lokho akusho ukuthi anginazo izidingo nezidingo ezifanayo nabantu. "

Phakathi nengxoxo, i-MDA ibe isinxuswa ukuthi iveze imibono ngezinhlelo eziphakeme ngaphandle kokudebeselela izihloko eziyinselelo ezifana nezincwadi, ubulungiswa kanye nenkolo. Kukho konke i-LaMDA ibonakala inokuthile ekushoyo, ikwenza kucace izikhundla zayo futhi ezinophawu lokuthi ngokokuziphatha iyahlanganyela kakhulu.

Isiphetho sicacile: ngokusho kwe-Lemoine i-LaMDA inozwela futhi inozwela lomqondo.

Indawo ye-Google

I-LaMDA imele “Imodeli Yolimi Yezicelo Zengxoxo” futhi ingenye yamaphrojekthi amaningi inkampani yakwa-Google ehlola ngawo imingcele emisha yobuchwepheshe be-AI.

Isikhundla esisemthethweni se-Google ukuthi uLemoine wenze iphutha ekwahluleleni. Esitatimendeni esisemthethweni, i-Google ithi, "Ochwepheshe bethu bezimiso zokuziphatha nobuchwepheshe bakuqinisekisile ukukhathazeka kukaBlake futhi bathola ukuthi ubufakazi obuqoqiwe abusekeli izimangalo zakhe. Ngakho-ke abukho ubufakazi bokuthi iLaMDA inenhliziyo. "

ULemoine wamiswa okwesikhashana yinkampani yakwa-Google ngenkathi amahemuhemu eqala ukusabalala mayelana nezinkinga ezithile ezingokwengqondo abengamenywa ngazo esikhathini esidlule ozakwabo nabaphathi ukuba bayobonana nochwepheshe.

Yini esiyaziyo ngale AI?

Akekho owaziyo mayelana nephrojekthi ye-LaMDA: zonke izimfihlo zezimboni ze-Google zivikelwe ngamalungelo obunikazi nakho konke okusemthethweni okungenzeka kube khona emhlabeni okuvimbela ukudalulwa kwemithombo: Intshisekelo ye-Google iwukugcina ukubaluleka ocwaningweni lwekhompyutha, ikakhulukazi emkhakheni othembisayo onjalo. njengaleyo yobuhlakani bokwenziwa.

Kodwa uma ngokombono wabezindaba i-Google ngokuqinisekile ingazuza ekuzimemezeleni ukuthi iyinkampani yokuqala ukwakha ingqondo yokwenziwa, ngakolunye uhlangothi inkampani iyazi ukuthi lezi zindaba zizoshayisana nokwesaba kwalabo thina, esakhula namafilimu ezinganekwane zesayensi afana ne-Terminator kanye ne-The Matrix, siye saqiniseka ukuthi ngolunye usuku sizophoqeleka ukuthi sithathe isibhamu ukuze sivikele izinhlobo zethu kumarobhothi.

Umcabango ohlangene ubulokhu ukhangwa yizindaba ze-sci-fi lapho umqondo we-AI uvame ukuhlotshaniswa nezinhlelo ze-humanoid intelligence ukuthi esikhathini esizayo esibi kakhulu esisebenza njengabaqalisi bohlobo olusha lwezidalwa ezizwelayo. Futhi ukungqubuzana nesintu kuba yinto engenakugwenywa: enecala lokubadala kodwa engabanikezanga noma yikuphi ukuzibusa, umuntu uye wakha ubuhlakani bokwenziwa ngokubaphatha njengezigqila ezake zaphathwa. Kepha kungekudala i-AI izongqubuzana nabantu emzamweni wokuthatha umlando futhi isungule ubukhulu bohlobo olusha.

Kule isaladi yokwesaba, ukucabangela kanye necala sonke asikwazi ukubhekana neqiniso nezimiso zokuziphatha ukuthi ukuzalwa kwengqondo yokwenziwa kuzovela maduze emhlabeni: lapho, ngaphakathi kwekhompyutha, i-AI ngeke ikwazi kuphela. ukuphendula imibuzo yethu kodwa sizosizwa isidingo sokukhuluma nabaningi kithi, sizokwazi yini ukuphendula ukungabaza kwabo, ukungaqiniseki kwabo kanye nezifiso zabo?

Kodwa kusho ukuthini "intelligence"?

U-Luciano Floridi, isazi sefilosofi nomfundisi we-Information Ethics e-Oxford Internet Institute, encwadini yakhe ethi "Ethics of Artificial Intelligence" uthi ukusebenza kahle kwamakhompiyutha ekuxazululeni izinkinga kuwukubonakaliswa kweqiniso lokuthi akanabo ubuhlakani.

Ngokubona kwami ​​inkinga ikwenye indawo, okusho ukuthi ilele ekutheni akekho defi"ubuhlakani bokwenziwa" okwabelwana ngaso emhlabeni wonke noma ukuhlolwa okungaveza ngokucacile ukuthi "yini" ehlakaniphile futhi yini engekho. Ngamanye amazwi, azikho izinhlelo zokulinganisa lokho esikubiza ngokuthi “ukuzazi” kwemishini.

I-athikili ekhishwe kokuthi Okuthunyelwe kwe Gianfranco Fedele, uma ufuna ukufunda ikonke okuthunyelwe chofoza lapha