Katangar Dokokin EU ta Haɗu da Rarrabuwar Tebura na Meta: Na gida, Tsare-tsare, da Makomar Paradoxical na AI
Yayin da Brussels ta shagaltu da tsara littafin ƙa'ida don makomar AI, Meta cikin nutsuwa ya zame samfuran AI masu ƙarfi na ma'auni mai mahimmanci a cikin nau'i wanda ya dace da teburin ku. Yi tsammani wanne ne ainihin zai sake fasalin duniya da sauri? Dokar AI ta EU, wacce a yanzu tana rayuwa a hukumance, wata dabba ce mai ban mamaki, mai bin doka, shaida ce ga sha'awar nahiyar ta lalata yammacin daji na basirar wucin gadi. Amma kamar yadda tawada ke bushewa akan cikakken tsarin sa, sabbin yunƙurin Meta na barazanar haifar da ɓangarorinsa da yawa, idan ba wanda ya ƙare ba, ta hanyar samar da nagartaccen AI ta hanyoyin da masu mulki ba su yi tsammani ba.
Bari mu bayyana a sarari: Dokar EU AI ba ƙaramin sabuntawa ba ne. Sauyi ne na tushe, wani tsari na Leviathan wanda aka ƙera don yin bayanin yadda aka tsara tsarin AI, haɓakawa, da turawa a duniya [1]. Idan kamfanin ku yana aiki a ciki ko ma * mu'amala * tare da EU, yanzu kuna kan ƙugiya. Wannan ba game da ticking kwalaye ba ne; yana game da gaske sake tsara fayil ɗin AI ɗin ku. Kuna rarraba tsarin ta hanyar haɗari - daga "marasa karɓa" (tunanin ƙima na zamantakewa) zuwa "mafi haɗari" (na'urorin likitanci, muhimman abubuwan more rayuwa) - da aiwatar da ingantaccen tsarin mulki, nuna gaskiya, da matakan ingancin bayanai [2]. Rashin yin biyayya, kuma kuna duban tara har zuwa kashi 7% na canjin shekara ta duniya. Wannan ba bugu ba ne a wuyan hannu; wato yankewar kamfani.
Manufar ita ce daraja, ba shakka. EU tana son haɓaka amintacce, AI mai dogaro da ɗan adam. Suna son hana son zuciya na algorithmic, tabbatar da alhaki, da kare haƙƙoƙin asali. Wannan "Brussels Effect" yana nufin ƙa'idodin su ba makawa za su zama ƙa'idar gaskiya ta duniya, ta tilasta yin la'akari da ɗabi'a cikin jigon ci gaba ga kowane kamfani da ke da burin duniya. Ga CTOs, wannan yana nufin sabon, tsada, da kuma hadadden tsarin mulkin AI da ƙwarewar shari'a yanzu ba za'a iya sasantawa ba. Taswirar hanyoyin samfuran ku kawai sun sami ƙarin rikitarwa, kuma ƙungiyar lauyoyin ku kawai ta sami ƙarfi sosai.
Bam Na Karɓa
Amma sai akwai Meta. Yayin da injina ke yin aiki, Meta kawai ya watsar da samfuran AI masu ƙarfi, kamar jerin Llama, waɗanda za'a iya inganta su don yin aiki yadda yakamata akan kayan masarufi, gami da GPU ɗaya [3]. Bari wannan ya nutse cikin. Samfuran ma'auni masu mahimmanci, suna gudana akan GPU guda ɗaya. Wannan ba nasara ba ce kawai ta fasaha; girgizar kasa ce mai dabara. Tsawon shekaru, hikimar da ta mamaye ita ce cewa AI mai ƙarfi da gaske, musamman manyan samfuran harshe da wakilai, za su kasance keɓancewar yanki na hyperscalers, kulle a cikin manyan cibiyoyin bayanai masu yunwa, ana samun dama ta hanyar kiran API kawai da manyan kuɗaɗen girgije. Meta kawai ya busa wannan ƙirar zuwa smithereens.
Wannan ba batun rage farashi bane kawai, kodayake wannan babban al'amari ne. Yana da game da ba da demokraɗiyya damar samun nagartaccen AI. Ka yi tunanin sabon ƙarni na aikace-aikacen AI akan na'urar waɗanda ba koyaushe suke yin waya gida zuwa gajimare ba. Yi tunanin mataimakan AI na sirri waɗanda ke fahimtar ku da gaske ba tare da aika mafi kusancin bayananku zuwa gonar sabar a Virginia ba. Tsarukan da aka haɗa waɗanda ke yin ainihin-lokaci, yanke shawara masu rikitarwa ba tare da jinkiri ba. Maganganun kasuwancin inda keɓaɓɓen bayanan keɓaɓɓu, kuma AI ba ta taɓa barin wurin ba. Wannan ba kawai "baki AI" ba; shine Sovereign AI ga talakawa.
Abubuwan da ke haifar da ƙirƙira suna da ban mamaki. Masu haɓakawa, masu farawa, har ma da masu tinkerers na yanzu na iya ginawa da tura wakilai masu ƙarfi na AI ba tare da buƙatar ƙirjin yaƙin babban birnin ba ko layin kai tsaye zuwa AWS. Wannan motsi yana rage jinkiri, rage farashin aiki, kuma yana canza ainihin ƙididdiga na sirri. Bayanan yana zama na gida. Sarrafa yana zama na gida. Ƙarfin wutar lantarki yana canzawa daga masu samar da girgije mai tsakiya zuwa na'urar, mai amfani, cibiyar sadarwar gida.
The Regulatory Paradox
Anan ga inda labaran biyu suka yi karo, suna haifar da ban sha'awa, mai ban tsoro. An ƙera Dokar EU AI da kyau don duniyar da AI ta fi haɓakawa kuma ana tura ta ta manyan, abubuwan da za a iya ganewa, sau da yawa a cikin gajimare, inda bayanai ke gudana ana iya gano su, kuma ana iya aiwatar da bin doka a wuraren shaƙewa. An tsara shi don duniyar sarrafawa ta tsakiya.
Amma menene zai faru lokacin da manyan jami'an AI ke gudana akan miliyoyin GPUs guda ɗaya, akan kwamfyutocin kwamfyutoci, tsarin da aka haɗa, da sabar gida, mai yuwuwar gyare-gyare, daidaitawa, ko ma masu amfani da buɗe ido? Ta yaya kuke tantance ƙaƙƙarfan ƙira mai gudana akan GPU ɗaya a cikin ƙaramin ofis ɗin bayan kasuwanci? Ta yaya kuke aiwatar da buƙatun bayyana gaskiya yayin da layin ƙirar zai iya zama faci na abubuwan buɗaɗɗen tushen tushe da daidaitawa na gida? Rukunin "haɗari mai girma" ba zato ba tsammani ya zama mai wahala sosai don saka idanu da sarrafawa lokacin da AI ba ta zaune a cikin yanayin girgije mai kayyade.
Tsarin Tarayyar Turai, yayin da yake tashe-tashen hankula, yana jin kamar yana ƙoƙarin daidaita kogi ta hanyar gina madatsun ruwa a maɓuɓɓugarsa, a daidai lokacin da kogin ke daf da fashe gaɓoɓinsa tare da mamaye duk faɗin ƙasar. Har ila yau "Tasirin Brussels" na iya yin tasiri kan manyan ayyukan kamfanoni, amma "Tasirin Meta" na ko'ina, mai ƙarfi AI na iya ƙirƙirar ƙaƙƙarfan tattalin arziƙin inuwa na aikace-aikacen AI.
Tattakin Tightrope na CTO
Ga CTOs da jagoranci na injiniya, wannan yana ba da kyakkyawar tafiya ta igiya. A gefe ɗaya, Dokar EU AI tana buƙatar tsauri, tsada, kuma galibi jinkirin yarda. Dole ne ku * saka hannun jari a cikin shugabancin AI, kimanta haɗari, da tsarin ɗabi'a [4]. Ƙungiyoyin doka da bin bin doka yanzu suna da alaƙa da dabarun ku na AI. Yin watsi da wannan shine kashe kansa na kuɗi.
A gefe guda, Meta ya buɗe ƙofofin ambaliya zuwa sabbin abubuwan da ba a taɓa gani ba a gefen. Damar gina sauri, mai rahusa, ƙarin sirri, da ƙarin ƙarfin aikace-aikacen AI yana da girma. Fa'idar fa'idar yin amfani da AI mai ƙarfi na gida na iya zama mai canzawa [5]. Ka yi tunanin duniyar da mutum-mutumin masana'antar ku ke koyo da daidaitawa a filin masana'anta ba tare da aika mahimman bayanan aiki ga wani ɓangare na uku ba, ko kuma inda sabis na abokin ciniki AI ke aiwatar da hadaddun tambayoyin gaba ɗaya a kan-gida, yana kare sirrin mai amfani ta ƙira.
Kalubalen shine a sasanta wadannan runduna guda biyu. Ta yaya kuke rungumar ikon da aka rarraba na gefen AI yayin da lokaci guda ke tabbatar da bin ka'idoji masu tsauri, kuma galibi gajimare? Wannan ba matsalar fasaha ba ce kawai; dabara ce. Yana buƙatar sabon nau'in gine-ginen AI - wanda ya fahimci ba kawai haɓaka ƙirar ƙira da turawa ba, har ma da tsarin tsari, ikon mallakar bayanai, da abubuwan da suka shafi ɗabi'a a babban matakin na'urar.
Makomar AI ba kawai game da manyan samfura ko guntu masu sauri ba. Yana da game da asali tashin hankali tsakanin tsakiya iko da karkatacciyar iko. EU na son gina amintaccen lambun da za a iya tsinkaya don AI. Meta kawai ya mika wa kowa iri don daji. Kewaya wannan sabon yanayin zai ayyana shekaru goma masu zuwa na jagorancin fasaha.
Sources
[1] Majalisar Turai, Majalisar Tarayyar Turai - Doka (EU) 2024/1689 kan Ilimin Artificial (Dokar AI) - https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R1689 [2] PwC - Dokar EU AI: Jagora mai amfani don yarda - https://www.pwc.com/gx/en/issues/data-privacy/eu-ai-act.html [3] Meta AI - Llama 2: Buɗe Foundation da Samfuran Taɗi Mai Kyau - https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/ [4] Gartner - Manyan Dabarun Fasaha na Fasaha don 2024: Amintaccen AI, Haɗari da Gudanar da Tsaro (AI TRISM) - https://www.gartner.com/en/articles/top-strategic-technology-trends-for-2024 [5] McKinsey & Kamfanin - Ƙimar tattalin arziƙin AI na haɓakawa: Iyakar yawan aiki na gaba - https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier