Shin ainun ƙwararren hankali na wucin gadi ya fi makaman nukiliya haɗari? Shin akwai ainihin damar afkuwar bala'i kamar na Chernobyl? Shin za a iya samun fiye da kashi 10% na damar cewa AI na iya "share dukan mutane" cikin shekaru goma masu zuwa?
Waɗannan gargaɗin sun fito a cikin kwanaki biyu da suka gabata daga Birtaniya da Amurka game da yuwuwar tasirin fasahar da yawancin mutane har yanzu suke gani a matsayin injin bincike mai yawan magana.
Wasu daga cikin waɗannan damuwa an taso da su a Westminster, inda 'yan majalisa da manyan jami'ai suka fara bincike kan haɗarin AI na wuce gwanin ɗan adam. Kwatancen nukiliya sun fito daga Des Browne, tsohon sakataren tsaro, da Farfesa Stuart Russell, sanannen masanin kimiyyar kwamfuta a Berkeley.
Beatrice Fihn, wadda ta lashe kyautar zaman lafiya ta Nobel a 2017 saboda jagorancin yakin neman kawar da makaman nukiliya na duniya, ta gaya wa 'yan majalisa: "Wannan ba shine karo na farko da muka fuskanci ƙarfin da zai iya kashe mu duka ba."
Zaman na ranar Litinin an shirya shi ne ta Control AI, ƙungiyar fafutuka da ke turawa don ƙa'idodin duniya kan fasahar. Tana goyon bayan wani doka da aka gabatar a majalisa a wannan makon ta Labour MP Alex Sobel, wanda ke nufin hana ƙirƙirar ƙwararren hankali na wucin gadi (ASI).
A daren Talata, wani babban ma'aikaci a Anthropic ya yarda cewa ya yi imani akwai fiye da kashi 10% na damar cewa fasahar za ta iya "kashe dukan mutane" cikin shekaru goma masu zuwa. Ya kuma yi gargaɗin cewa kamfaninsa na AI ba shi da wani shiri don tabbatar da cewa ASI ya dace, ma'ana ba zai haifar da cutarwa ba. Hasashen lokacin da ASI zai zo ya bambanta daga 'yan shekaru zuwa fiye da shekaru goma.
Evan Hubinger, wanda ke jagorantar kimiyyar daidaitawa a Anthropic, ya sanya wannan sharhi bayan Jacob Coxon, mai bincike mai shekaru 28 a kamfanin San Francisco kuma a baya a OpenAI, ya yi murabus. Coxon ya yi iƙirarin cewa "babu wani kamfani da ke yin abin da ya dace" kuma ya ce suna "caca da rayukanmu."
Coxon ya ce haɗarin a Anthropic an fahimta sosai, amma kamfanin yana "kulle cikin tseren don isa wurin da farko." A matsayin ƙaramin alamar bege, ya ƙara da cewa yana da kyakkyawan fata game da yuwuwar dakunan gwaje-gwaje na Amurka su yi aiki tare don rage ci gabansu a tseren.
An nemi Anthropic ta yi sharhi kan wannan batu. OpenAI ta yi nuni ga wata sanarwa daga babban masanin kimiyyarta, Jakub Pachocki, wanda ya ce a makon da ya gabata: "Haɗin kai na duniya kan ci gaban AI na gaba ya zama babban fifiko ga gwamnatocin duniya."
Yunkurin samun haɗin kai mai faɗi don sarrafa ASI ya kasance muhimmin abu a abubuwan da suka faru a Westminster a wannan makon. A cewar Financial Times, jami'an gwamnati sun nuna damuwa cewa Anthropic ta ƙi mika sabon samfurinta – Mythos 5.1 – ga Cibiyar Tsaro ta AI ta Birtaniya don gwadawa kafin fitarwa.
Ofishin Majalisar ya tabbatar da cewa ƙungiyoyin Amurka kaɗan ne kawai suka sami damar yin amfani da Mythos 5.1. Wani mai magana da yawun ya ƙara da cewa: "Cibiyar Tsaro ta AI tana ci gaba da aiki kafaf da kafaf da abokan hulɗar masana'antu, ciki har da Anthropic."
Labour MP Darren Jones ya rubuta wa shugabannin duniya yana kira ga yarjejeniyar ƙasashe da yawa don tabbatar da cewa an haɓaka AI cikin aminci. Ya rubuta wa firaminista, Andy Burnham, da shugabannin UN da OECD, yana kira ga "yarjejeniyar ƙasashe da yawa don ingantaccen ci gaba da aminci na superintelligence – ba hana ƙirƙira ko aikin kimiyya ba, amma tsarin aminci na farko ga saurin ci gaban wannan fasaha."
Yana nuni ga murabus na Coxon, Jones ya ƙara da cewa: "Muhawarar ta bambanta daga ƙarshen ɗan adam zuwa iƙirarin 'hype na talla.' Ko ta yaya, gwamnatoci dole ne su shiga yanzu."
A gefe guda na Tekun Atlantika, Bernie Sanders ya ƙara yakin neman tsaron AI a wannan makon ta hanyar sake kira ga Majalisa don sarrafa fasahar. Ya yi nuni ga binciken da ya nuna cewa kashi 81% na Amurkawa sun yi imanin cewa 'yan siyasarsu ya kamata su ɗauki mataki.
Sanatan mai zaman kansa daga Vermont ya ce: "Ba za mu iya barin ƴan mutane masu kwadayi su yi wasa da Allah su yanke shawarar makomar ɗan adam – tattalin arzikinmu, muhalli, dimokuradiyya, sirrinmu da sauransu – ba tare da ra'ayin jama'a ba."
Bayan tallan jarida
View image in fullscreen
Sanatan Amurka Bernie Sanders yana son Majalisa ta sarrafa masana'antar AI mai 'kwadayi'. Hoton: Linda Nylind/The Guardian
ControlAI na samun kuɗaɗe daga Jaan Tallinn, hamshakin attajirin wanda ya kafa Skype, wanda ke kiran kansa "mai adawa da ɓacewa." Ya saka wani ɓangare na dukiyarsa wajen yakin neman tsaron AI kuma ya ce wasu manyan jami'an AI za su yarda da share ɗan adam.
Duk da kasancewarsa mai saka hannun jari na farko a Anthropic da Google DeepMind, biyu daga cikin manyan dakunan gwaje-gwaje na AI, Tallinn ya ƙiyasta cewa kashi 10-15% na ma'aikatan AI sun yi imanin cewa fasahar za ta zama magajin ɗan adam mai cancanta.
Ya ce a wata hira a farkon wannan bazara: "Wani sanannen mai binciken AI ya gaya mani, 'Jaan, kada ka damu da wannan. Mutane jinsin da za a iya jefar da su ne.' Daga abin da na fahimta, ya yarda da zama ɓatacce."
Ƙungiyar fafutuka ta Tallinn ta gudanar da zaman ga 'yan majalisa a Westminster ranar Litinin, tana roƙonsu su goyi bayan matakan da za su hana mafi ƙarfin samfuran AI. A kan kowace kujera akwai kwafin littafin Idan Wani Ya Gina Shi, Kowa Ya Mutu: Hujjar Adawa da Superintelligent AI.
'Yan siyasa sun ji daga Russell, wanda ya yi gargaɗin Chernobyl kuma ya ce: "Wata yuwuwar ita ce bala'i mafi girma, inda ɗan adam ya rasa iko ba tare da juyawa ba. Amma ba mu da ikon yanke shawara kan ko za mu ci gaba da wanzuwa."
Browne ya ce lokacin da yake sakataren tsaro, yana tunanin yakin nukiliya shine mafi girman barazanar ɗan adam. Yanzu ya yi imanin "AI mai ƙarfin gaske yana da barazana a wannan girman, watakila ma mafi girma."
OpenAI models sun tafi ba tare da izini ba. Muna buƙatar ingantaccen bincike na 'hugging face' | Mackenzie Arnold da Stephan Llerena
Karanta ƙarin
Sauran masu ba da gudummawa sun fi taka tsantsan. Dr Andrew Rogoyski na Cibiyar Surrey don AI mai Dogaro da Mutane ya ce: "A gaskiya, waɗannan tsarin ba su da yawa kamar mutane, balle mutane su yi aiki tare. Ina tsammanin muna kan hanyar zuwa 'babban rashin jin daɗi,' inda AI mai ci gaba ya zama mai tsada sosai kuma ba shi da amfani don ci gaba a yanayin da yake yanzu."
David Barber, darektan Sofair, wani dakin binciken AI na gwamnati wanda ya haɗa malamai daga Oxford, Cambridge, Edinburgh, da UCL, ya ce yana damuwa game da mutane "suna jefa jariri tare da ruwan wanka."
"AI ba zai tafi ba," in ji farfesan. "Yana da matuƙar amfani, ko kun bar shi ya shiga intanet da tsarin daban-daban ba tare da wani ƙuntatawa ba, wanda ke da yuwuwar matsala. Muna iya buƙatar koyon yadda za mu sarrafa waɗannan abubuwa da kyau. Akwai raunuka a cikin tsarin software waɗanda ke buƙatar gyara. Amma hakan yana yiwuwa."
"Abin da muke buƙata a matsayin ƙasa shine mu fahimci duality: yana da matuƙar muhimmanci da amfani, kuma a lokaci guda, wani abu ne da ke buƙatar tunani a hankali da sarrafawa a hankali."
Sandra Wachter, farfesa a Cibiyar Oxford Internet, ta ce ba ta yi imani da "yanayin Terminator" ba amma ta lura cewa AI na haifar da barazana na gaske, ciki har da tasirin muhalli, yada bayanan karya, da maye gurbin ayyuka.
"Waɗannan matsalolin na gaske ne kuma suna gaggawa kuma suna buƙatar magance su yanzu," in ji ta. "Yanayin Terminator babban abin raba hankali ne daga matsalolin gaske."
Gary Marcus, mai sharhi kan masana'antar AI kuma masani, ya ce: "Akwai bambanci tsakanin superintelligence wanda ya dace (idan irin wannan abu zai yiwu) da superintelligence wanda bai dace ba."
"Yana da yuwuwa aƙalla cewa na farko zai iya zama mai amfani. Ya zuwa yanzu ba mu da ko ɗaya, amma yawan hype tare da ƙarancin taka tsantsan daga OpenAI ya kai mu inda muke, tare da rashin amana mai tsanani a ko'ina."
