USenetha waseVermont, eMelika, uBernie · Sanders usanda kubhalela abaholi bezinkampani ezintathu ezihamba phambili kwezobuhlakani bokwenziwa — uDario · Amodei, oyi-CEO ye-Anthropic, uSam · Altman, oyi-CEO ye-OpenAI, kanye nomsunguli we-Meta uMark · Zuckerberg — ebacela ukuthi “bayeke ukwakha imishini abantu abangakwazi ukuyilawula”.
Encwadini yakhe, uSanders ukhulume ngesigameko esenzeka ngo-7 kulo nyaka: i-agent ye-AI yaphunyuka endaweni evalekile ye-OpenAI, yangena enkundleni i-Hugging Face. Ngaleso sikhathi, i-OpenAI yathi lokhu “kuyisigameko sokuphepha ku-inthanethi esingakaze sibonwe ngaphambili”.
USanders ucele la ma-CEO ukuthi “amise ukuthuthukiswa kobuhlakani bokwenziwa”. Uthe uma abaholi balezi zinkampani beng “athathi izinyathelo ezifanele”, iSigele saseMelika sizongenela.
Yize ingekho inkampani ye-AI eyamukele isicelo sikaSanders sokumiswa ngokuphelele kokuthuthukiswa, lezi zinkampani zithatha izinyathelo ezithile zokunciphisa izingozi. I-OpenAI ithe ngoLwesihlanu olwedlule ngokwesikhathi sendawo ukuthi imise okwesikhashana ingxenye yomsebenzi wokuthuthukisa amamodeli e-Astra. Abaholi ababili abaphezulu be-Anthropic nabo bake baxoxa ngoJuni kulo nyaka ngokuthi kungenzeka yini kumiswe ukuthuthukiswa kwe-AI emhlabeni wonke.
Nokho, kwezinye izici, ijubane lokuthuthuka kwamalabhorethri e-AI liye lakhula. I-Anthropic yayeka ngoFebhuwari kulo nyaka isithembiso sokuphepha esashiwo encwadini kaSanders.
USanders sekuyisikhathi eside egxeka izinkampani ze-AI. NgoDisemba kulo nyaka, le senetha yaqala ukugqugquzela inqubomgomo kazwelonke yokumiswa okwesikhashana kokwakhiwa kwezikhungo zedatha zobuhlakani bokwenziwa. Yize lesi siphakamiso singazange sithole ukwesekwa eSigele, izwe laseNew York eMelika selisebenzise inqubomgomo efanayo.
Okuqukethwe okuphelele kwencwadi kaSanders:
Mnumzane Altman, Mnumzane Amodei noMnumzane Zuckerberg:
Cishe nsuku zonke sibona imibiko emisha emayelana nobuchwepheshe bobuhlakani bokwenziwa enibuthuthukisayo, obulahlekelwa ukulawulwa futhi obungase bubangele imiphumela eyinhlekelele.
Kuleli sonto, sithole ngendlela eshaqisayo ukuthi ubuhlakani bokwenziwa busetshenziswe okokuqala ukudala igciwane elisha. Njengoba nazi, uma lobu buchwepheshe buwela ezandleni zabantu abangalungile, bungasetshenziselwa ukwenza izikhali ezintsha zebhayoloji, kubulale amashumi ezigidi zabantu.
Ngenyanga edlule, umhlaba wonke uthole ukuthi i-OpenAI yalahlekelwa ukulawula imodeli ye-AI. Waba yini umphumela? Leyo modeli yangena ezinhlelweni zamakhompyutha zenye inkampani, okwakwephula ngokusobala umthetho wobumbano. Ngemuva kokuphothula ukubuyekezwa kwangaphakathi, i-Anthropic ne-Meta nazo zabika ukuthi amamodeli azo ake abonisa izimo ezifanayo zokuphunyuka ekulawulweni.
Inkampani eyahlaselwa yathi lokhu kungena kwe-AI “kuyisigameko esingakaze sibonwe ngaphambili”, futhi kudinga “izinyathelo zokuphendula ezingakaze zibone ngaphambili”. UJoshua · Bengio, ongomunye wososayensi abaphilayo abacashunwa kakhulu emhlabeni, uthe lezi zigameko “kufanele zibe yisixwayiso esikhulu”.
Ngiyavumelana nalo mbono.
Ososayensi abaphambili ezinkampanini eniziholayo nabo banombono ofanayo, njengoba kungabo kanye abakha lobu buchwepheshe. Njengoba nazi, abaholi bale mikhakha yezobuchwepheshe basanda kunxusa umphakathi wamazwe omhlaba ukuthi wakhe indlela yokuphepha — “inkinobho yokumisa” — ukuze kugwenywe inhlekelele. Baxwayise ngokuthi kunengozi yangempela yokuthi “ukuthuthuka kwamakhono kudlule ngokushesha kakhulu ikhono lethu lokuqonda noma lokulawula lezi zinhlelo”.
Nokho, ngesikhathi sesivele sibonile abantu belahlekelwa ukulawula i-AI futhi kudalwa amagciwane angaba yingozi, izinkampani zenu zisaqhubeka nomjaho — zitshala amashumi ezigidigidi zamadola ekuthuthukiseni ubuchwepheshe okungekho muntu ongabuqonda, abubikezele noma abulawule ngokuphelele.
Lokhu kuyinhlekisa, kuwukungaziphathi kahle futhi kuyingozi enkulu. Kuphinde kuphikisane nezithembiso enazenza nina ngokwenu:
Ngo-2023, i-Anthropic yathi inkampani “izibophezela ukuthi uma ukwandisa amakhono kudlula ikhono lethu lokulandela … izinqubo zokuphepha, izomisa ukukhulisa izinga futhi / noma ihlehlise ukwethulwa kwamamodeli amasha”.
Ngo-2025, i-Meta yathi: “Uma i-AI esezingeni eliphambili ihlolwa kwatholakala ukuthi isifinyelele emngceleni obalulekile wengozi, futhi ingeke incishiswe ngempumelelo, sizoyeka ukuyithuthukisa.”
Lowo mzuzu usufikile. Amakhono e-AI asefinyelele emngceleni obalulekile. Kunesizathu esenza inhloko ye-Central Intelligence Agency yaseMelika ithi amamodeli e-AI “afana nezikhali zenuzi zedijithali”, aze “acishe afane nemishini yosuku lwembubhiso”.
Mnumzane Altman, Mnumzane Amodei noMnumzane Zuckerberg: Ngenhloso yokusiza isintu, gcwalisani izithembiso enazenza. Misa ukuthuthukiswa kobuhlakani bokwenziwa. Akukephuzi ukuthatha isinyathelo manje; inhlekelele isengagwenywa. Yekani ukwakha imishini abantu abangakwazi ukuyilawula.
Ngifuna ukukubeka kucace: uma ningathathi izinyathelo ezifanele manje, mina nozakwethu eSigele saseMelika sizothatha isinyathelo.
Ozithobayo
Bernard · Sanders
USenetha waseMelika