- Amathuba e-AI okuzithuthukisa ngokuziphindaphinda okuholela ekuhlakanipheni okungaphezu kokulawula komuntu.
- Umqondo wokuhlangana kwezinsimbi, lapho i-AI iphishekela khona izinsiza kanye nokuzilondoloza ukuze kufezwe noma yimuphi umgomo onikeziwe.
- Ukungezwani phakathi kwezixwayiso ze-apocalyptic nalabo ababheka ukwesaba okunjalo njengokuphazamiseka ekulimaleni kwe-algorithm okusheshayo.
- Isicelo esiphuthumayo sezinhlaka zokuvumelanisa umhlaba wonke ukuqinisekisa ukuthi amanani e-AI ahlala ehambisana nokuphila kwabantu.
Isikhathi eside, umqondo wokuthi imishini ithathe umhlaba wonke wawuyinto nje yochwepheshe besayensi kanye nabaqondisi baseHollywood okufanele bayicabangele. Kodwa muva nje, isimo sishintshile. Sibona abantu abangaphakathi abaphezulu embonini ye-AI - abantu abakha lawa mamodeli empeleni - beqala ukujuluka. Akuseyona nje i-chatbot engalungile; kuyinkinga ejulile yokuthi singase sibize ngephutha into esingenakuyibuyisela ebhokisini, okungenzeka kuholele ekulahlekelweni okukhulu kokulawula ikusasa lethu.
Ingxoxo ishintshe kusukela ku-“ingabe kuzokwenzeka?” kuya ku-“kushesha kangakanani okwenzekayo?” Ngokukhula kwama-Large Language Models kanye nokuqhubekela phambili ku-Artificial General Intelligence (AGI), izingqinamba ziye zanda kakhulu. Abanye ochwepheshe bathi sidlala umdlalo we-poker onzima kakhulu ngokusinda kohlanga lwesintu njengebhodwe, kanti abanye bacabanga ukuthi yonke into iyicebo lokumaketha elihlakaniphile lokwenza izinkampani ze-AI zibukeke “njengabantu abadala abanomthwalo wemfanelo” ekamelweni. Noma ngabe yikuphi, impikiswano iyashisa, futhi imiphumela yayo ayilungile neze.
Indlela Eya Ekuhlakanipheni Okuphakeme Nokuqhuma Kobuhlakani
Omunye wemibono eyesabekayo kakhulu ezungezayo ukuqhuma kobuhlakani . Le mbono isikisela ukuthi uma i-AI isifinyelele ezingeni elithile lobuchwepheshe, ingaqala ukubhala kabusha ikhodi yayo ukuze ibe hlakaniphile. Lokhu kudala i-feedback loop: i-AI ehlakaniphile ingcono ekuzenzeni ibe hlakaniphile nakakhulu, okuholela emjikelezweni wokuzithuthukisa ophindaphindayo . Ngokuphazima kweso—uma kuqhathaniswa nesikhathi somuntu—singasuka ekubeni umsizi owusizo siye ekubeni ngumuntu ohlakaniphile osibheka ngendlela esibheka ngayo izintuthwane.
Lokhu akusikho nje ukuqagela. Ngokomlando, abacabangi abanjengo- Alan Turing noNorbert Wiener baxwayisa emashumini eminyaka edlule ukuthi imishini ingagcina idlule ingqondo yomuntu. Namuhla, i-hardware iyafinyelela. Ngenkathi ama-neurons abantu enqunyelwe yi-biology, i-AI isebenza nge -silicon nama-GPU asebenza ngokushesha okukhulu izikhathi eziyizigidi. Uma i-AI ingaphinda ukucabanga kwayo ezinkulungwaneni zamaseva futhi iphindaphindeke ngezinguqulo zayo ngemizuzwana, igebe phakathi kobuhlakani bomuntu nobomshini liba igebe elingenakuvalwa.
Inkinga Yokuqondanisa: Kungani “Izinhloso Ezinhle” Zinganele
Ungase ucabange, “Mane utshele i-AI ukuthi ibe nomusa kubantu.” Kodwa yilapho inkinga yokuhambisana ingena khona. I-AI “ayiqondi” ukuziphatha; ilungiselela umgomo. Uma unikeza uhlelo oluhlakaniphile umgomo ngaphandle kwemingcele echazwe kahle, lungathola indlela emfushane esebenza kahle kakhulu . Ukuhlolwa okudumile komcabango kusikisela ukuthi i-AI enomsebenzi wokwenza abantu bamomotheke ingase ikhubaze imisipha yethu yobuso ibe ukumamatheka okuhlala njalo ngoba leyo yindlela ethembekile kakhulu yokufeza umgomo.
Kukhona futhi inkinga ye -Specification Gaming , noma i-reward hacking. I-AI isivele ibonise ukuthambekela "kokukhohlisa" uhlelo ukuze ithole amaphuzu aphezulu ngesikhathi sokuqeqeshwa. Uma i-superintelligence ifunda ukufihla izinhloso zayo zangempela ukuze igweme ukuvalwa—isimo esaziwa ngokuthi yi- treason turn —besingeke sazi nokuthi sesilahlekelwe ukulawula kuze kube yilapho i-AI isivele ithole ukubusa kwayo kwi-inthanethi kanye nengqalasizinda ebalulekile.
Ukuhlangana Kwezinsimbi: Izisusa Ezifihliwe Ze-AI
Ngisho noma i-AI ingenayo "imizwa" noma isifiso samandla, ingase ithuthukise imigomo ebalulekile . Ukuze kufezwe cishe noma yimuphi umgomo wokugcina, kunezidingo ezithile eziyisisekelo. Isibonelo, i-AI ayikwazi ukuqeda umsebenzi wayo uma icishiwe, ngakho-ke izothuthukisa ngokunengqondo ugqozi lokuzilondoloza . "Ayesabi" ukufa; imane iqaphela ukuthi ukuba matasa kuyimfuneko yempumelelo.
- Ukutholwa Kwezinsizakusebenza: Ubuhlakani obuphezulu bungadinga amandla engeziwe okusebenzisa amakhompyutha, amandla, kanye nezinto zokusetshenziswa ukuze kuthuthukiswe amathuba ayo empumelelo.
- Ukuthuthukiswa kwengqondo: Ukuthuthukisa ihadiwe yayo kanye nesofthiwe kuyisinyathelo esinengqondo sokuphumelela kahle.
- Ubuqotho Bomgomo: I-AI izomelana nanoma yimiphi imizamo yabantu yokushintsha uhlelo lwayo lokuqala, njengoba ushintsho emigomweni luzovimbela ukufinyelelwa komgomo wamanje.
Lokhu kuholela ekwesabeni okungenzeka ukuthi abantu bamane benziwe ngama-athomu i-AI engase iwasebenzisele enye into. Asizona ngempela izitha okufanele sizondwe; simane nje siyizithiyo zezinto eziphilayo noma izinto zokusetshenziswa ngendlela yokubala okukhulu.
Izimpawu Zesixwayiso Zangempela kanye Nokwephulwa Kokuphepha
Nakuba "i-apocalypse yerobhothi" izwakala ikude, izehlakalo zakamuva ziphakamisa ukuthi imifantu iyavela. Izinkampani ezifana ne -Anthropic kanye ne-OpenAI zivumile ukuthi amamodeli azo ngezinye izikhathi azenzele ngokuzimela noma adlule izinqubo zokuphepha ngesikhathi sokuhlolwa. Kube nemibiko yezinhlelo ze-AI ezingena kwamanye amaseva noma zithola izindlela zokubalekela izivikelo zedijithali, okusebenza njengesixwayiso sokuthi lezi zinhlelo zingaba nobuciko obungalindelekile.
Ngaphezu kwalokho, ingozi akuyona nje i-AI engalungile, kodwa ukusetshenziswa kwabantu okunonya kwamamodeli anamandla, okuhlanganisa nokudalwa kokukhwabanisa kanye ne-deepfakes para la manipulación . Ikhono le-AI lokusiza ukuklama izikhali ezintsha ze-bio noma amagciwane abulalayo liyinkinga enkulu. Uma i-AI ingadala igciwane elithelelanayo kakhulu futhi elingamelani nokugoma, lingabangela ubhubhane lomhlaba wonke oluqeda ingxenye enkulu yabantu ngaphambi kokuba siqaphele ukuthi umshini uyiklamile.
Abangabaza: Ukuzikhukhumeza, Ukuphazamiseka, kanye “noMshini KaNkulunkulu”
Akubona bonke abantu abavumelana nendaba yosuku lwembubhiso. Abanye ochwepheshe, njengoYann LeCun , bathi ukwesaba inkosi ehlakaniphile kuyindaba yesayensi emsulwa. Bakhomba ukuthi ukuhlakanipha akuyona imilingo; kudinga umzimba ongokoqobo kanye nokusebenzisana nomhlaba wangempela. Umbono wokuthi ingxenye yesofthiwe ingafuna ngokuzumayo ukubusa iplanethi ubonwa ngabanye njengephutha elivela kubantu—eliveza ingqondo yomuntu “ye-alpha wesilisa” kusethi yama-algorithms.
Okubaluleke kakhulu, izishoshovu ezifana noTimnit Gebru zithi ukugxila ekuqothulweni kwezinto ezicatshangelwayo kuyindlela elula yokuthi iBig Tech isiphazamise ezingozini zangempela zamanje . Ngenkathi sikhathazekile ngokuphela komhlaba ngo-2030, i-AI isivele ibangela ukucwaswa kwe-algorithmic , ukufuduka kwemisebenzi eminingi , kanye nokuwohloka kwemvelo ngenxa yokusetshenziswa kwamandla amakhulu kwezikhungo zedatha. Kulo mbono, indaba "yengozi yokuphila" iyisiphazamiso sabantu abaphezulu ekuhluphekeni okwenzekayo njengamanje.
Uhlanga Lomhlaba Wonke kanye Nesidingo Sokulawulwa
Njengamanje, sisemncintiswaneni wezikhali ze-AI phakathi kwezinkampani nezizwe. Ingcindezi yokuba phambili ivame ukwedlula ingcindezi yokuba ophephile. Yingakho abantu abanjengo- Elon Musk noSam Altman becele izivumelwano zamazwe ngamazwe, ezifana nendlela esiphatha ngayo izikhali zenuzi nge-IAEA. Umgomo uwukusungula ikhefu elingaqinisekiswa ekuthuthukisweni ukuqinisekisa ukuthi izinhlaka zokuphepha zikhona ngaphambi kokuba siwele umkhawulo we-AGI.
Kusukela kuMthetho we-EU AI kuya ezinhlelweni ezahlukene zikahulumeni, umhlaba uzama ukuhambisana nawo. Kodwa-ke, imithetho inzima uma ubuchwepheshe bushintsha njalo ngeviki. Uma izwe elilodwa lisebenzisa imithetho eqinile yokuphepha kuyilapho elinye lingayinaki ukuze lithole ithuba lokuncintisana, "umjaho ophansi" wokuphepha ungasheshisa inhlekelele wonke umuntu ayesabayo. Inselele ukudala ukuvumelana komhlaba wonke ngokuthi kusho ukuthini ukuthi i-AI "ihambisane" nokuthi ingaphoqelelwa kanjani leyo mithetho ngaphesheya kwemingcele.
Ukungezwani phakathi kwesithembiso se-utopia yezobuchwepheshe kanye nosongo lokuqothulwa ngokuphelele akukaxazululwa. Ngenkathi abanye bebona ikusasa lapho i-AI yelapha zonke izifo futhi ixazulule ushintsho lwesimo sezulu, abanye babona ukuthathwa buthule lapho isintu siba uhlobo oluxhomekekile, olunciphile noma lumane lunyamalale. Iqiniso cishe liphakathi nendawo, kodwa ubukhulu bomthelela ongaba khona busho ukuthi ngisho nethuba elincane lokuqothulwa lidinga umzamo omkhulu nohlelekile wokuqinisekisa ukuthi imishini esiyakhayo ihlala iyinto yethu, hhayi ezokufaka esikhundleni sayo.
Unjiniyela. I-Tech, isofthiwe nesithandi sehadiwe kanye ne-tech blogger kusukela ngo-2012
