Ko da yake an yi la'akari da bayanan wucin gadi don yuwuwar sa na haɓaka aiki a masana'antu da yawa, fasahar ta ci karo da shingaye da yawa na kanta. Rashin ingantaccen sabar sabar ya sa wasu tsare-tsare su dandana high latency matsaloli ko ma a zahiri hadarurruka. Wannan ya ce, ana iya samun mafita ga waɗannan rashin aiki: ƙididdiga na gefe.
Edge sarrafa kwamfuta yana nufin sarrafa bayanai a cibiyoyin sadarwa da na'urorin da ke kusa da mai amfani, sabanin hanyoyin lissafin gargajiya na sabar sabar da cibiyoyin bayanai. Ana iya ganin lissafin Edge a cikin aiki tare da komai daga tsarin tallace-tallace (POS) zuwa na'urorin Intanet na Abubuwa (IoT) da na'urori masu auna firikwensin - ainihin, na'urorin gefen duk wani abu ne wanda ke ƙididdigewa a cikin gida kuma yana hulɗa tare da girgije. Yanzu, mun fara ganin ƙirar ƙira ta wucin gadi da aka ƙara zuwa wannan nau'in ƙididdiga na gefe.
Haɗin kai na AI da ƙididdigar ƙira
Masanin ilimin wucin gadi Siddhartha Rao yana da kwarewa ta farko da ke aiki tare da AI da fasahar sarrafa girgije. Bayan fiye da shekaru goma na gwaninta aiki tare da manyan kamfanonin fasaha kamar Amazon Web Services, Rao yanzu yana aiki a matsayin co-kafa da Shugaba na. Positron Networks, Kamfanin AI ya mayar da hankali kan mafitacin hankali na wucin gadi ga al'ummar binciken kimiyya. Idan aka ba da buƙatun musamman na wannan al'umma, Rao yana da sha'awar mahadar AI da ƙididdigar ƙira.
"Akwai dalilai da yawa da ya sa lissafin gefen ya zama irin wannan sanannen tsari, ciki har da rage jinkirin hulɗar masu amfani, rage farashin lissafin girgije, da tallafawa ƙwarewar mai amfani da layi," in ji Rao. "Wadannan fa'idodin suna cikin rikici tare da wasu manufofin ƙididdiga, kamar inganta haɓaka ta hanyar rage farashin na'urori, tsawaita rayuwar batir tare da na'urori masu ƙarancin ƙarfi, ko zazzage sabbin samfura a cikin ƙananan mahalli kamar ƙasashe masu tasowa."
Koyaya, wannan yana haifar da tambayar yadda samfuran bayanan sirri na wucin gadi - waɗanda ke buƙatar mahimman ikon sarrafa kwamfuta - za'a iya gudanar da su "a gefen." Rao ya yi bayanin cewa samun nasarar sauya samfuran bayanan sirri na wucin gadi zuwa gefe yana buƙatar sauƙaƙe waɗannan ayyukan.
"Don mahallin, samfuri shine jerin ayyukan algebra na layi (matrix) da aka aiwatar akai-akai don tsinkayar amsa dangane da shigarwa," Rao yayi bayani. “ Injiniyoyin koyon injiniyoyi da masana kimiyya suna rage rikitar lissafi na waɗannan ayyuka ta hanyar amfani da dabaru daban-daban. Sakamakon shi ne ƙananan ƙira waɗanda ke buƙatar ƙarancin kewayon kwamfuta don aiwatarwa, rage buƙatun ƙididdiga da haɓaka tabo yayin inganta rayuwar batir. ”
"Wannan aikin injiniya yana da tasiri mai kyau ga masana'antu," in ji shi. “Misali, samfuran da ke aiwatarwa tare da ƙarancin jinkiri, suna amfani da ƙarancin ƙarfin batir, kuma suna haifar da ƙarancin zafi. Hakanan suna amfani da ƙarancin albarkatun lissafin girgije don aiwatarwa, ƙara rage tasirin muhallinsu, kuma suna rage yawan amfani da bandwidth na sabunta samfurin. Duk waɗannan fa'idodin suna haɓaka ƙwarewar mai amfani tare da haɓaka ragi ta hanyar rage farashin kaya."
Me yasa ƙididdigar gefen shine makomar AI
Babban fa'idar ƙididdiga ta gefe ita ce tana haɓaka ƙwarewar mai amfani sosai. Latency yana ɗaya daga cikin manyan ƙalubalen da kowace sabuwar fasahar ke fuskanta, musamman idan aka yi la'akari da gajeriyar kulawar ɗan adam (da raguwa).
"Idan na'urar gefen dole ne koyaushe ta je gajimare don tsinkaya, tasiri akan latency yana lalata ƙwarewar mai amfani, rage haɗin gwiwar abokin ciniki," in ji Rao. "Ƙasashen abokan ciniki ba su da yuwuwar yin amfani da na'urar, rage amfani da karɓuwa."
Ƙididdigar Edge kuma na iya yuwuwar rage farashin fasaha na fasaha na wucin gadi zuwa maƙasudin da ya fi dacewa ga ƙananan kasuwanci. Bayan haka, kula da sabar da ake buƙata don sarrafa manyan samfura yana da tsada. Ƙididdigar Edge yana rage rikitar waɗannan aikace-aikacen don ba su damar yin aiki a gefen.
Koyaya, Rao ya kuma yi taka tsantsan game da wasu sakamakon da ƙididdige ƙididdigewa a cikin yanayin AI na iya haifarwa, yana mai nuni da su a matsayin "ciniki" don fa'idodin da yake bayarwa. "Mafi girman kuskure da ƙimar hasashe yayin amfani da ƙananan daidaito daga baya yana rinjayar daidaiton amsar," in ji shi. “Tsarin ilimi na iya haɓaka son zuciya da matsalolin adalci a cikin manyan samfuran. A ƙarshe, ƙididdige ƙididdiga na buƙatar hayar tsada, ƙwarewar koyon injin na musamman da kuma samun kayan aikin horar da na'ura masu tsada kamar GPUs, waɗanda ke cikin buƙatu sosai."
A matsayin misali na nasarar haɗin kai na fasaha na wucin gadi da fasaha na lissafin gefe, Rao yana nuna shari'ar amfani daga lokacinsa a AWS wanda ya sami nasarar sauƙaƙa samfurin a sikelin. "An yi amfani da samfurin da na ba da tallafi a AWS a cikin Opus codec," in ji shi. “Wannan codec sama da na’urori biliyan 1 ne ke amfani da shi a duk duniya kuma kwanan nan Amazon ya inganta shi tare da wani fakitin ɓoye bayanan na'ura wanda ya dawo da rafukan sauti har ma a cikin hanyoyin sadarwa masu asara. Ana iya amfani da wannan codec akan na'urori masu iyakacin ikon sarrafawa, kamar Rasberi Pi ko wayar tebur, don tsinkayar samfuran sauti cikin millise seconds."
Har ila yau, Rao ya ambaci shari'ar amfani da ta nuna dama ta musamman a fannin tsaro. Ya kara da cewa "An yi amfani da faifan bidiyo na gaske a kyamarorin da ke kan iyakar bindigar soja don yi musu jagora kan ko motsin mayakan na da shakku ko kuma mai yiyuwa ne don nuna goyon bayan wani mummunan aiki kamar ta'addanci," in ji shi. "Sai sojan zai iya mai da hankali kan sa ido kan wuraren da ke da matukar hadari da kuma hadarin gaske. A cikin misalan biyun, hadaddun, sauti ko bidiyo na ainihin lokaci ana sarrafa su ta hanyar ƙananan injina masu ƙarfi waɗanda ke aiki akan na'urorin IoT. "
Lallai, waɗannan shari'o'in amfani da su cikakkiyar misalan yadda ƙididdige bayanan sirri na wucin gadi a gefen ke ba da kyakkyawan sakamako dangane da inganci da farashi. "Idan ba za a iya inganta samfuran AI na baya-bayan nan don gudana a gefen ba, aikace-aikacen su za a iyakance ga aikace-aikacen girgije da ƙwarewar mai amfani," Rao ya kammala.
