Ngokombiko osanda kukhululwa we-Industrial AI kanye ne-AI Market Report 2021-2026, izinga lokwamukelwa kwe-AI kuzilungiselelo zezimboni likhuphuke lisuka kumaphesenti angu-19 laya kumaphesenti angama-31 esikhathini esingaphezudlwana nje kweminyaka emibili. Ngaphezu kwamaphesenti angama-31 abaphendulile abakhiphe i-AI ngokugcwele noma ngokwengxenye emisebenzini yabo, amanye amaphesenti angama-39 okwamanje ahlola noma ahlola ubuchwepheshe.
I-AI ivela njengobuchwepheshe obubalulekile kubakhiqizi nezinkampani zamandla emhlabeni jikelele, futhi ukuhlaziywa kwe-IoT kubikezela ukuthi imakethe yezixazululo ze-AI yezimboni izokhombisa izinga eliqinile lokukhula konyaka kwangemva kobhubhane (CAGR) lama-35% ukuze lifinyelele ku- $ 102.17 billion ngo-2026.
Inkathi yedijithali izele I-inthanethi Yezinto. Kungabonakala ukuthi ukuvela kobuhlakani bokwenziwa kuye kwasheshisa ijubane lokuthuthukiswa kwe-Intanethi Yezinto.
Ake sibheke ezinye zezinto eziqhuba ukwanda kwe-AI yezimboni ne-AIoT.
Isici 1: Amathuluzi esoftware amaningi ngokwengeziwe we-AIoT yezimboni
Ngo-2019, lapho i-Iot analytics iqala ukumboza i-AI yezimboni, yayimbalwa imikhiqizo yesoftware ye-AI ezinikele evela kubathengisi bobuchwepheshe bokusebenza (OT). Kusukela lapho, abathengisi abaningi be-OT bangene emakethe ye-AI ngokuthuthukisa futhi banikeze izixazululo zesoftware ye-AI ngendlela yamapulatifomu e-AI ephansi lemboni.
Ngokusho kwedatha, cishe abathengisi abangama-400 banikela ngesoftware ye-AIoT. Inani labathengisi be-software abajoyina imakethe ye-AI yezimboni lenyuke kakhulu eminyakeni emibili edlule. Ngesikhathi socwaningo, i-IoT Analytics ikhombe abahlinzeki abangama-634 bobuchwepheshe be-AI kubakhiqizi/amakhasimende ezimboni. Kulezi zinkampani, ezingama-389 (61.4%) zinikeza isoftware ye-AI.
Inkundla entsha yesoftware ye-AI igxile ezindaweni zezimboni. Ngale kwe-Uptake, Braincube, noma i-C3 AI, inani elikhulayo labathengisi bobuchwepheshe bokusebenza (OT) banikela ngezinkundla ezizinikele zesoftware ye-AI. Izibonelo zifaka i-ABB's Genix Industrial analytics kanye ne-AI suite, i-Rockwell Automation's FactoryTalk Innovation suite, inkundla yezokukhiqiza ye-Schneider Electric uqobo, kanye nakamuva nje, izengezo ezithile. Ezinye zalezi zinkundla ziqondise izinhlobonhlobo zezigameko zokusetshenziswa. Isibonelo, inkundla ye-Genix ye-ABB ihlinzeka ngezibalo ezithuthukile, okuhlanganisa izinhlelo zokusebenza ezakhelwe kusengaphambili kanye nezinsizakalo zokuphathwa kokusebenza kokusebenza, ubuqotho bempahla, ukusimama kanye nokusebenza kahle kwe-supply chain.
Izinkampani ezinkulu zibeka amathuluzi azo esoftware ye-ai phansi esitolo.
Ukutholakala kwamathuluzi esoftware ye-ai nakho kuqhutshwa amathuluzi esoftware akhethekile athuthukiswe yi-AWS, izinkampani ezinkulu ezifana neMicrosoft ne-Google. Isibonelo, ngoDisemba 2020, i-AWS yakhipha i-Amazon SageMaker JumpStart, isici se-Amazon SageMaker esihlinzeka ngesethi yezixazululo ezakhelwe ngaphambili nezingenziwa ngokwezifiso zamacala asetshenziswa kakhulu ezimbonini, njenge-PdM, umbono wekhompyutha, nokushayela okuzimele, Sebenzisa ukuchofoza okumbalwa nje.
Izixazululo zesoftware eziqondene necala elithile zithuthukisa ukusebenziseka kalula.
Amasudi esofthiwe e-usse-case-specific, njengalawo agxile ekulungiseni okubikezelwayo, aya ngokuya avama. I-IoT Analytics ibone ukuthi isibalo sabahlinzeki abasebenzisa izixazululo zesoftware ye-AI-based product data management (PdM) senyuke safinyelela ku-73 ekuqaleni kuka-2021 ngenxa yokwanda kwemithombo yedatha ehlukahlukene kanye nokusetshenziswa kwamamodeli okuqeqeshwa kwangaphambili, kanye nokusabalala ukwamukelwa kobuchwepheshe bokuthuthukisa idatha.
Isici sesi-2: Ukuthuthukiswa nokugcinwa kwezixazululo ze-AI kwenziwa lula
Ukufunda ngomshini okuzenzakalelayo (AutoML) kuba umkhiqizo ojwayelekile.
Ngenxa yobunkimbinkimbi bemisebenzi ehlotshaniswa nokufunda komshini (ML), ukukhula okusheshayo kwezinhlelo zokusebenza zokufunda ngomshini kudale isidingo sezindlela zokufunda ngaphandle kweshelufu ezingasetshenziswa ngaphandle kobuchwepheshe. Inkambu yocwaningo, i-automation eqhubekayo yokufunda komshini, ibizwa nge-AutoML. Izinkampani eziningi zisebenzisa lobu buchwepheshe njengengxenye yeminikelo yazo ye-AI ukusiza amakhasimende akhe amamodeli e-ML futhi asebenzise amacala okusetshenziswa kwezimboni ngokushesha. NgoNovemba 2020, isibonelo, i-SKF yamemezela umkhiqizo osuselwa ku-automL ohlanganisa idatha yenqubo yomshini nedatha yokudlidliza nezinga lokushisa ukuze kwehliswe izindleko futhi kunikwe amandla amamodeli ebhizinisi amasha kumakhasimende.
Ukusebenza komshini (ML Ops) kwenza ukuphathwa nokugcinwa kwamamodeli kube lula.
Isiyalo esisha semisebenzi yokufunda komshini sihlose ukwenza lula ukugcinwa kwamamodeli we-AI ezindaweni zokukhiqiza. Ukusebenza kwemodeli ye-AI ngokuvamile kuyehla ngokuhamba kwesikhathi njengoba kuthintwa izici ezimbalwa ngaphakathi kwesitshalo (isibonelo, izinguquko ekusabalaliseni idatha namazinga ekhwalithi). Ngenxa yalokho, ukugcinwa kwamamodeli nokusebenza komshini kuye kwadingeka ukuze kuhlangatshezwane nezidingo zekhwalithi ephezulu yezindawo zezimboni (isibonelo, amamodeli asebenza ngaphansi kuka-99% Angase ahluleke ukuhlonza ukuziphatha okubeka engcupheni ukuphepha kwezisebenzi).
Eminyakeni yamuva nje, iziqalo eziningi zijoyine isikhala se-ML Ops, okuhlanganisa i-DataRobot, Grid.AI, Pinecone/Zilliz, Seldon, ne-Weights & Biases. Izinkampani ezisunguliwe zengeze imisebenzi yokufunda ngomshini ekunikezelweni kwazo kwesoftware ye-AI ekhona, kufaka phakathi iMicrosoft, eyethule ukutholwa kwedatha ku-Azure ML Studio. Lesi sici esisha senza abasebenzisi bakwazi ukubona izinguquko ekusabalaliseni idatha yokokufaka eyehlisa isithunzi sokusebenza kwemodeli.
Isici sesi-3: Ubuhlakani bokwenziwa busetshenziswa ezinhlelweni ezikhona nasemacaleni okusetshenziswa
Abahlinzeki besofthiwe bendabuko bengeza amakhono e-AI.
Ngokungeziwe kumathuluzi esoftware ye-AI amakhulu avundlile akhona njenge-MS Azure ML, i-AWS SageMaker, ne-Google Cloud Vertex AI, amasudi esofthiwe endabuko afana ne-Computerized Maintenance Management Systems (CAMMS), ama-Manufacturing execution systems (MES) noma ukuhlelwa kwensiza yebhizinisi (ERP) manje ingathuthukiswa kakhulu ngokujova amakhono e-AI. Isibonelo, umhlinzeki we-ERP i-Epicor Software yengeza amandla e-AI emikhiqizweni yayo ekhona nge-Epicor Virtual Assistant (EVA) yayo. Ama-ejenti e-EVA ahlakaniphile asetshenziselwa ukwenza ngokuzenzakalelayo izinqubo ze-ERP, njengokuhlela kabusha imisebenzi yokukhiqiza noma ukwenza imibuzo elula (isibonelo, ukuthola imininingwane mayelana nentengo yomkhiqizo noma inombolo yezingxenye ezitholakalayo).
Amacala okusetshenziswa kwezimboni athuthukiswa kusetshenziswa i-AIoT.
Amacala amaningana okusetshenziswa kwezimboni athuthukiswa ngokwengeza amandla e-AI kungqalasizinda ekhona yehadiwe/yesoftware. Isibonelo esicacile ukubona komshini ezinhlelweni zokulawula ikhwalithi. Amasistimu okubona omshini ovamile acubungula izithombe ngamakhompuyutha ahlanganisiwe noma ahlukahlukene afakwe isofthiwe ekhethekile ehlola amapharamitha anqunywe kusengaphambili namathreshold (isb., ukugqama okuphezulu) ukuze kunqunywe ukuthi izinto ziyabonisa yini ukukhubazeka. Ezimweni eziningi (isibonelo, izingxenye ze-elekthronikhi ezinomumo wezintambo ezihlukene), inani lezinto ezingamanga liphezulu kakhulu.
Nokho, lezi zinhlelo zivuselelwa ngobuhlakani bokwenziwa. Isibonelo, umhlinzeki wemboni yemboni ye-Vision u-Cognex ukhiphe ithuluzi elisha le-Deep Learning (Vision Pro Deep Learning 2.0) ngoJulayi 2021. Amathuluzi amasha ahlanganisa nezinhlelo zombono zendabuko, okwenza abasebenzisi bokugcina bahlanganise ukufunda okujulile namathuluzi ombono wendabuko ohlelweni olufanayo ukuze ukuhlangabezana nezindawo zezokwelapha neze-elekthronikhi ezidingekayo ezidinga ukulinganiswa okunembile kwemihuzuko, ukungcola nokunye ukukhubazeka.
Isici sesi-4: I-Industrial AIoT hardware iyathuthukiswa
Ama-chip e-AI athuthuka ngokushesha.
Ama-chips e-AI e-Hardware ashumekiwe akhula ngokushesha, anezinketho eziningi ezitholakalayo ukusekela ukuthuthukiswa nokusatshalaliswa kwamamodeli we-AI. Izibonelo zifaka amayunithi akamuva e-NVIDIA acubungula izithombe (i-Gpus), i-A30 ne-A10, ezethulwa ngoMashi 2021 futhi afanele izimo zokusebenzisa i-AI njengezinhlelo zokuncoma nezinhlelo zokubona kwekhompyutha. Esinye isibonelo isizukulwane sesine se-Google Tensors Processing Units (TPus), okungamasekhethi ahlanganisiwe anenhloso ekhethekile (i-ASics) angakwazi ukufeza ukusebenza kahle okuphindwe ka-1,000 kanye nesivinini ekuthuthukisweni kwemodeli kanye nokusatshalaliswa kwemithwalo ethile ye-AI (isb, ukutholwa kwezinto. , ukuhlukaniswa kwezithombe, nezilinganiso zokuncoma). Ukusebenzisa ihadiwe ye-AI ezinikele kunciphisa isikhathi sokubala semodeli ukusuka ezinsukwini kuye emaminithini, futhi kufakazele ukuthi kuwushintsho lomdlalo ezimweni eziningi.
Ihadiwe ye-AI enamandla itholakala ngokushesha ngemodeli ekhokhelwa ukusetshenziswa ngakunye.
Amabhizinisi e-Superscale ahlala ethuthukisa amaseva awo ukuze enze izinsiza zekhompyutha zitholakale emafini ukuze abasebenzisi bokugcina bakwazi ukusebenzisa izinhlelo zokusebenza ze-AI yezimboni. NgoNovemba 2021, ngokwesibonelo, i-AWS yamemezela ukukhishwa okusemthethweni kwezimo zayo zakamuva ezisekelwe kwi-GPU, i-Amazon EC2 G5, enikwa amandla yi-NVIDIA A10G Tensor Core GPU, ngezinhlelo zokusebenza ezihlukene ze-ML, okuhlanganisa umbono wekhompyutha nezinjini zokuncoma. Isibonelo, umhlinzeki wezinhlelo zokuthola i-Nanotronics usebenzisa izibonelo ze-Amazon EC2 zesisombululo sayo sokulawula ikhwalithi esisekelwe ku-AI ukusheshisa imizamo yokucubungula kanye nokuzuza izilinganiso zokutholwa ezinembe kakhudlwana ekwenzeni ama-microchip nama-nanotubes.
Isiphetho kanye neProspect
I-AI iphuma efekthri, futhi izotholakala yonke indawo ezinhlelweni ezintsha, njenge-AI-based PdM, futhi njengezithuthukisi ku-software ekhona kanye namacala okusebenzisa. Amabhizinisi amakhulu akhipha amacala amaningana okusetshenziswa kwe-AI kanye nempumelelo yokubika, futhi amaphrojekthi amaningi anenzuzo ephezulu ekutshalweni kwezimali. Sekukonke, ukukhuphuka kwamafu, amapulatifomu e-iot nama-chip anamandla e-AI kunikeza inkundla yesizukulwane esisha sesoftware nokwenza kahle.
Isikhathi sokuthumela: Jan-12-2022