Google DeepMind

Google DeepMind
Bayanai
Suna a hukumance	DeepMind Technologies Limited
Iri	ƙaramar kamfani na da ginin daji
Masana'anta	Kirkirar Basira (Artificial Intelligence)
Ƙasa	Birtaniya
Aiki
Ma'aikata	10,000
Kayayyaki	AlphaGo (en) , AlphaFold (mul) , AlphaGo Zero (en) da Gemini (mul)
Mulki
Hedkwata	Landan da Googleplex (mul)
Tsari a hukumance	ƙaramar kamfani na
Mamallaki	Google (mul) da Alphabet Inc. (mul)
Tarihi
Ƙirƙira	2010
Wanda ya samar	Shane Legg (en) Demis Hassabis (mul) Mustafa Süleyman (mul)
Founded in	Landan
deepmind.google

'DeepMind' Technologies Limited, ciniki a matsayin Google DeepMind ko kawai DeepMind, wani dakin gwaje-gwaje ne na bincike na wucin gadi na Burtaniya da Amurka Landan ke aiki a matsayin reshe na Alphabet Inc. An kafa shi a Burtaniya a cikin 2010, Google ne ya saye shi a cikin 2014 kuma ya haɗu da sashen Google Brain na Google AI don zama Google DeepMinda a cikin Afrilu 2023. Kamfanin yana da hedkwata a London, tare da cibiyoyin bincike a Amurka, Kanada, Faransa, Jamus, da Switzerland.

A cikin 2014, DeepMind ya gabatar da na'urorin Turing na jijiyoyi (cibiyoyin sadarwa na jijiyoyi waɗanda zasu iya samun damar ƙwaƙwalwar waje kamar na'urar Turing ta al'ada). Kamfanin ya kirkiro samfuran cibiyar sadarwa da yawa waɗanda aka horar da su tafi da ilmantarwa don kunna Wasannin bidiyo da wasannin jirgi. Ya zama babban labari a cikin 2016 bayan shirin AlphaGo ya doke Lee Sedol, zakaran duniya na Go, a wasan wasanni biyar, wanda daga baya aka nuna shi a cikin shirin AlphaGo . Wani shirin gaba ɗaya, AlphaZero, ya doke shirye-shiryen da suka fi karfi suna wasa go, Chess da shogi (chess na Japan) bayan 'yan kwanaki na wasa da kansa ta amfani da ilmantarwa mai ƙarfafawa. DeepMind tun daga lokacin ya horar da samfuran don wasan kwaikwayo (MuZero, AlphaStar), don lissafi (AlphaGeometry), da kuma binciken algorithm (AlphaEvolve, AlphaDev, AlphaTensor).

Tarihi

Demis Hassabis, Shane Legg da Mustafa Suleyman ne suka kafa farawa a watan Nuwamba na shekara ta 2010. Hassabis da Legg sun fara haduwa a Gatsby Computational Neuroscience Unit a Kwalejin Jami'ar London (UCL).

Demis Hassabis ya ce farawa ya fara aiki a kan fasahar fasaha ta wucin gadi ta hanyar koya masa yadda za a buga tsoffin wasannin daga shekarun saba'in da tamanin, waɗanda suke da mahimmanci idan aka kwatanta da waɗanda suke a yau. Wasu daga cikin waɗannan wasannin sun haɗa da Breakout, Pong, da Space Invaders . An gabatar da AI ga wasa daya a lokaci, ba tare da wani masaniya game da dokokinta ba. Bayan ya kwashe wani lokaci a kan koyon wasan, AI zai zama gwani a ciki. "Hanyoyin fahimta da AI ke wucewa an ce suna kama da na ɗan adam wanda bai taɓa ganin wasan ba zai yi amfani da shi don fahimtar da ƙoƙarin sarrafa shi. " Manufar waɗanda suka kafa shi shine ƙirƙirar AI mai ma'ana wanda zai iya zama da amfani da tasiri ga kusan komai.

Kayayyaki da fasaha

Ya zuwa 2020, DeepMind ya buga takardu sama da dubu, gami da takardu goma sha uku waɗanda Nature ko Science suka karɓa. DeepMind ta sami kulawar kafofin watsa labarai a lokacin AlphaGo; bisa ga binciken LexisNexis, labaran labarai na 1842 da aka ambata DeepMind a cikin 2016, ya ragu zuwa 1363 a cikin 2019.

Wasanni

Ba kamar AIs na baya ba, kamar IBM's Deep Blue ko Watson, waɗanda aka haɓaka don ƙayyadadden manufa kuma kawai aiki a cikin wannan iyaka, algorithms na farko na DeepMind an yi niyya ne don zama gaba ɗaya. Sun yi amfani da ilmantarwa mai ƙarfafawa, algorithm wanda ke koyo daga gogewa ta amfani da pixels kawai a matsayin shigar da bayanai. Hanyarsu ta farko ta yi amfani da zurfin Q-koyon tare da cibiyar sadarwa ta jijiyoyi. Sun gwada tsarin a kan wasannin bidiyo, musamman wasannin arcade na farko, kamar Space Invaders ko Breakout. Ba tare da canza lambar ba, wannan AI ya sami damar yin wasu wasannin da kyau fiye da kowane ɗan adam.