Umehluko phakathi kwe-AI esezindaweni kanye ne-AI esekelwe efwini: umhlahlandlela ophelele

Informatec Digital » Izinsiza » Umehluko phakathi kwe-AI esezindaweni kanye ne-AI esekelwe efwini: umhlahlandlela ophelele

Isinqumo esikhulu namuhla asisekho ukuthi uzosebenzisa ubuhlakani bokwenziwa noma cha , kodwa ukuthi uzobusebenzisa kuphi futhi kanjani: kumadivayisi akho, efwini, noma inhlanganisela yakho kokubili. Lokhu kukhetha kuthinta izindleko, ukuphepha, ukusebenza, ngisho nemodeli yebhizinisi yanoma iyiphi iphrojekthi yedijithali, kusukela kuhlelo lokusebenza oluncane kuya epulatifomu enabasebenzisi abayizinkulungwane.

Uma sikhuluma ngomehluko phakathi kwe-AI yasendaweni kanye ne-AI esekelwe efwini, empeleni sikhuluma ngokuthi imodeli ihlala kuphi, ukuthi ubani oyilawulayo, nokuthi kwenzekani kudatha . Ukusuka lapho, kuvela izinkinga ezinkulu, njengokuthobela imithetho (i-GDPR, i-HIPAA), ukubambezeleka kwesikhathi sangempela, ukukhuliswa, ukutshalwa kwezimali kwehadiwe, ukukhiya kwabathengisi, kanye namakhono okwenza ngokwezifiso. Ake sikuchaze ngokuthula, kodwa ngaphandle kokushaya indlwabu.

Kuyini i-AI yendawo nokuthi yini i-AI yamafu, kuchazwe ngamagama alula

Umehluko wokuqala ulula kakhulu: i-AI yendawo yileyo esebenza kwihadiwe yakho (i-PC, iseva esendaweni, idivayisi yeselula, izinhlelo zezimboni, njll.), kuyilapho i-AI yamafu iyileyo esebenza ezikhungweni zedatha zomhlinzeki wamafu futhi ivame ukusetshenziswa nge-API noma i-SDK.

Ngasohlangothini lwefu, izinsizakalo ezifana ne -Azure AI Services, i-Azure OpenAI Service, i-AWS, i-Google Cloud, i-OpenAI, i-Anthropic, i-Gemini, kanye ne-Hugging Face ziyasetshenziswa . Lokhu kufaka phakathi amamodeli asezingeni eliphezulu (i-GPT-4, i-GPT-4 Turbo ene-Vision, i-Claude, i-Gemini, i-PaLM, i-DALL-E, amamodeli amakhulu okushumeka, njll.) owafinyelela nge-inthanethi, ukhokha ngokusetshenziswa ngakunye noma ngokubhalisa.

Ngasohlangothini lwendawo, sithola konke kusukela ekusetshenzisweni okuvamile endaweni kuya ekusebenzeni ngqo kudivayisi: ama-PC anama-GPU, izindawo zokusebenza, amaseva esikhungweni sakho sedatha, noma ngisho namadivayisi eselula nawe-IoT. Amathuluzi afana ne -Ollama, i-LM Studio, i-KoboldCpp, i-AnythingLLM, noma izindawo ezifana ne -Windows ML, i-ONNX Runtime, i-Foundry Local, kanye ne-Microsoft Foundry ku-Windows enza kube lula ukulayisha nokusebenzisa amamodeli olimi (i-Llama, i-Mistral, i-Phi, i-Qwen, njll.) noma amamodeli okubona ngqo emshinini wakho.

Phakathi kwalezi zimo ezimbili ezibucayi kukhona indlela yokungena : amamodeli asetshenziswa kumadivayisi asemaphethelweni enethiwekhi (izinzwa, izinto ezigqokwayo, imishini yezimboni, amakhamera, omakhalekhukhwini) acubungula idatha eduze nalapho ikhiqizwa khona futhi athumela idatha ebalulekile efwini kuphela. Lena indlela ethakazelisa kakhulu yokunciphisa ukubambezeleka kanye nethrafikhi yenethiwekhi.

Izinto ezibalulekile okufanele uzicabangele lapho unquma phakathi kwe-AI esezindaweni kanye ne-AI esekelwe efwini

Ukukhetha kahle akukhona nje kuphela ngokuzikhethela komuntu siqu; izici eziningana zenza umehluko. Ubumfihlo, izinsiza, ukusebenza, izindleko, kanye nokukhula ngokuvamile yizona ezinquma kakhulu, kodwa akuzona zodwa.

Ubumfihlo, ukuthobela imithetho, kanye nokuphepha kuyisiqalo, ikakhulukazi emikhakheni elawulwayo njengebhange, ezempilo, kanye nokuphathwa komphakathi. Umbuzo ucacile: ingabe idatha ingaphuma ekhaya noma cha?

Okubalulekile futhi ukutholakala kwehadiwe kanye nezinsizakusebenza zabantu , ukulula kokugcinwa, isidingo sokusebenzisana kude, inani labasebenzisi abasebenzisana ngesikhathi esisodwa, kanye nobunzima bemodeli ngokwayo (imodeli encane yohlobo lwe-Phi ayifani ne-LLM enamapharamitha angaphezu kwe-100B).

Okokugcina, kubalulekile ukukhumbula uhlelo lwamathuluzi nokuhlanganiswa ne-stack yakho yamanje : uma usuvele usebenza ne-Azure, i-GitHub, i-AWS noma i-Google, noma uma unengqalasizinda ye-Windows ebanzi kakhulu (bheka i-Windows Pro kanye ne-Enterprise ), akufani nokuqala phansi.

Isihluthulelo ukuhlanganisa lezi zici nesimo sakho sangempela sokusetshenziswa: ukuzenzekela kwenqubo yangaphakathi, ama-ejenti okwenza izinqumo azimele, abasizi bokuhlela, ama-chatbot esevisi yamakhasimende, umbono wezimboni, ukuhlaziya okuthuthukisiwe, njll.

Ukuthobela ubumfihlo kanye nokulawula

Lapho kukhulunywa ngedatha ebucayi noma elawulwayo (impilo, ezezimali, idatha yamakhasimende omuntu siqu, impahla yobuhlakani enenani eliphezulu), indawo yokucubungula i-AI ayibe yimininingwane yobuchwepheshe futhi iba yisinqumo esiyingozi.

Ekusetshenzisweni kwe -AI endaweni , idatha iyacutshungulwa futhi igcinwe kwingqalasizinda yakho noma ngisho nakudivayisi yomsebenzisi . Lokhu kunciphisa kakhulu indawo yokuhlasela futhi kwenza kube lula ukuthobela imithetho efana ne -GDPR noma i-HIPAA , uma nje ukuphepha kwakho kwangaphakathi kuklanywe kahle. Kodwa-ke, umthwalo wemfanelo wokubethela, ukulawula ukufinyelela, ukulungisa, kanye nokuhlola uhlala ungowakho ngokuphelele.

Njengoba i-AI isefwini , abahlinzeki abakhulu banikeza izinyathelo zokuphepha eziqinile kakhulu (ukubethela ngesikhathi sokuthutha kanye nokuphumula, izitifiketi ze-ISO 27001, i-SOC 2, njll.), kodwa idatha iya ezikhungweni zabo zedatha . Lokhu kudinga ukubuyekezwa kwezinkontileka, izindawo zesikhungo sedatha, izigaba zokucubungula idatha, kanye nokulungiselela ngokucophelela izimvume kanye nama-API avikelekile. Ngaphezu kwalokho, kubalulekile ukuqinisekisa ukuthi ukuhlanganiswa kunamathela esimisweni selungelo elincane.

Yingakho amakhasimende amaningi ekhetha ukwakheka kwe-hybrid : idatha ebaluleke kakhulu icutshungulwa endaweni, kuyilapho imisebenzi engazweli kakhulu noma enzima (ukuqeqeshwa okukhulu, ukuhlaziywa okuhlanganisiwe) idluliselwa efwini, lapho ingozi ingalawuleka khona.

Ukutholakala kwezinsizakusebenza kanye nobunzima bemodeli

Isihlungi sesibili esikhulu amandla okusebenzisa ikhompyutha onawo. Ukusebenzisa i-AI yesimanje akufani nokusebenzisa uhlelo lokusebenza lwewebhu olulula; i-CPU ( amaprosesa e-Intel ne-AMD ), i-GPU, i-NPU, i-RAM, kanye nesitoreji kubalulekile.

Esimweni sendawo noma esisezindaweni , umkhawulo unqunywa yihadiwe yakho: ama-PC, amaseva, izindawo zokusebenza ze-GPU, amadivayisi eselula, njll. Amamodeli amancane, asebenzayo (njengomndeni wakwaPhi, amamodeli amancane e-Qwen, amamodeli epharamitha alinganisiwe angu-3-7B, kanye namamodeli e-NPU-optimized) asebenza kahle kakhulu kule ndawo, futhi amadivayisi afana ne- Copilot+ PC asevele ehlanganisa ama-NPU aklanyelwe le njongo, namamodeli afakwe ngaphambilini alungele ukusetshenziswa ne-Windows.

Amamodeli amakhulu e-GPT-4, Claude, kanye ne-Gemini , noma ama-LLM anamapharamitha angaphezu kwezigidigidi ezingu-70, kunzima ukuwaphatha endaweni yakho ngomuzwa omuhle womsebenzisi ngaphandle kokuthi une-GPU ephezulu kakhulu (i-RTX 4090 kanye nenye efanayo) kanye nendawo elungiselelwe kakhulu. Ezinhlanganweni eziningi, kunengqondo kakhulu ukusebenzisa lawa mamodeli kusuka efwini , lapho esevele elungiselelwe futhi elinganisiwe.

Amapulatifomu e-AI asekelwe efwini njenge-Azure AI Services, i-Azure OpenAI Service, i-AWS, kanye ne-Google Cloud akuvumela ukuthi usebenzise amandla amaningi ngangokunokwenzeka , ukhokhele kuphela lokho okusebenzisayo. Lapha unokufinyelela kuma-LLM, amamodeli okubona, izwi, izinhlelo zokuncoma, nokunye okuningi, ngaphandle kokuthenga i-GPU eyodwa.

Ukusebenza, ukubambezeleka, kanye nokuthembeka

Esinye isici esibalulekile ukuthi udinga i-AI ngokushesha nangokuthembekile kangakanani ukuze uphendule. Ukubambezeleka kungaba yinkinga ezimweni eziningi: ukuhweba nge-algorithmic, ukulungiswa kokubikezela ezitshalweni, ukutholwa kwamaphutha ngesikhathi sangempela, ukushayela okusizwayo, amarobhothi, noma nje ulwazi oluhle lomsebenzisi ku-chatbot.

Nge -AI yendawo noma enqenqemeni , idatha icutshungulwa kudivayisi uqobo noma kunethiwekhi yakho yangaphakathi, ngakho-ke awuxhomekile kwikhwalithi yokuxhumeka kwakho kwe-inthanethi (hlola izintambo zakho kanye nezintambo ze-Cat5, Cat6, kanye ne-Cat7 ) . Lokhu kunciphisa kakhulu ukubambezeleka futhi kugwema ukuzulazula, okubaluleke kakhulu ekubonweni kwezimboni, i-IoT, noma izinhlelo zokusebenza ezidinga ukuqhubeka nokusebenza noma ngabe inethiwekhi iyehla. Umehluko ukuthi ukusebenza okuphezulu kunqunyelwe yihadiwe yakho.

I-AI esekelwe efwini ingathembela kuhadiwe enamandla kakhulu futhi ekhethekile, kodwa uzohlala une -latency yenethiwekhi eyengeziwe . Ngokuxhumeka okuhle, ngokuvamile kuyamukeleka kakhulu kubasizi ababonakalayo, ukukhiqizwa kombhalo, noma ukuhlaziywa kwe-batch, kodwa ingasilela ekulawuleni kwesikhathi sangempela. Ukuze kuncishiswe lokhu, abahlinzeki banikeza izindawo ze-latency ephansi kanye ne-edge computing , okusondeza i-computing kumsebenzisi wokugcina.

Ngokuphathelene nokuthembeka, ukuthunyelwa kahle endaweni kugwema ukuthembela ekunqamukeni kwangaphandle, yize kudinga ukuthi ube nesu lakho lokutholakala okuphezulu kanye nelokusekelayo . Ifu, ngakolunye uhlangothi, linikeza ama-SLA, ukuphindaphinda kwendawo, kanye nokubuyiselwa kwenhlekelele, kodwa futhi lisengozini yokwehluleka kwabahlinzeki ngezikhathi ezithile.

Izindleko, ukukhuliswa kanye nemodeli yezomnotho

Impikiswano ngezindleko ayilula njengokuthi "indawo ishibhile, ifu libiza kakhulu" noma okuphambene nalokho. Kuyadingeka ukuhlukanisa phakathi kokutshalwa kwezimali kokuqala (i-CapEx) kanye nezindleko zokusebenza (i-OpEx) , kanye nokucabangela ukusetshenziswa kwamandla, abasebenzi, ukulungiswa, kanye nokukhula.

Kusu le -AI elisezindaweni , udinga ukutshala imali ku-hardware (amaseva, ama-GPU, isitoreji ( i-SSD vs. i-HDD ), ukupholisa, izikhungo zedatha), amalayisense uma kudingeka, futhi kaningi, abasebenzi abanamakhono okugcina yonke into. Njengokubuyisela, uma utshalomali selubuyisiwe, izindleko ngokuqagela ngakunye zingaba ziphansi kakhulu , ikakhulukazi uma unemithwalo yemisebenzi ezinzile nenkulu.

I-AI esekelwe efwini ilandela imodeli yokukhokha njengoba uqhubeka noma yokubhalisela . Awudingi ukutshala imali kusengaphambili; ungaqala ukusebenza ngamahora futhi ukhule kusuka kuzicelo eziyi-10 kuya kweziyi-10,000 cishe ngokushesha. Kodwa uma ivolumu ikhuphuka noma imodeli iba nezinsiza eziningi kakhulu, ibhili yanyanga zonke ingakhula ngokushesha okukhulu . Ngaphezu kwalokho, udinga ukubheka izindleko ezingacacile njengokudluliswa kwedatha, isitoreji esengeziwe, noma ukuqeqeshwa kwe-GPU okunamandla.

Yingakho izinkampani eziningi zigcina zinemodeli ehlanganisiwe nethuthukisiwe : zisebenzisa endaweni okujulile kodwa okubikezelwayo (isibonelo, imodeli yokuhlukanisa yangaphakathi noma i-chatbot yenkampani eqeqeshwe ngamadokhumenti ayo), futhi zisebenzisa ifu ukuthola iziqongo, amacala okuhlola noma amamodeli asemngceleni angenakubanjwa ngaphakathi.

Ukufinyeleleka, ukubambisana, kanye nezimo ezivamile zokusetshenziswa

Indlela amaqembu akho asebenzisana ngayo ne-AI nayo iyashintsha kakhulu kuye ngokuthi isetshenziswa kuphi. Ukusebenzisana ngemodeli yendawo akulula njengokubambisana ngesevisi yamafu eyabiwe.

Nge -AI yendawo , imodeli ivame ukufinyeleleka kuphela kusuka kudivayisi noma kunethiwekhi lapho ibanjwe khona . Lokhu kulungele izindawo ezihlukanisiwe, izinqubo zangaphakathi, noma amaphrojekthi adinga ukuhlala evaliwe ngokuphelele (amalabhorethri, izinhlelo ezibalulekile, izitshalo zezimboni). Kodwa-ke, kwenza kube nzima ukubambisana okusatshalaliswe phakathi kwamaqembu namahhovisi ngaphandle kokuthi usethe ungqimba lwenkonzo yangaphakathi oluvezwe kahle.

Izinsizakalo ze -AI ezisekelwe efwini ziphumelela kakhulu ekufinyelelekeni komhlaba wonke kanye nokusebenza ngokubambisana . Noma yiliphi ilungu elinezimvume kanye nokuxhumeka kwe-inthanethi lingafinyelela i-API kusuka noma kuphi. Lokhu kwenza kube ukukhetha okunengqondo kwabasizi ababonakalayo bomphakathi, i-SaaS yabasebenzisi abaningi, amathuluzi okubambisana, noma imikhiqizo yabasebenzisi bokugcina.

Ezinye izibonelo ezicacile ze-AI esetshenziswe kabanzi efwini yilezi: ukuhlanganiswa kwe-ChatGPT noma i-Azure OpenAI kuzinhlelo zokusebenza ze-Windows , ukwenziwa kwesithombe nge -DALL-E , abasizi bezincomo abakhelwe ku-.NET MAUI kanye nama-LLM amafu , noma izinsizakalo zokusesha umbono, izwi, ukuhumusha, kanye ne-semantic ezitholakala ngama-REST API.

Kulula ukusebenzisa, ukunakekela, kanye nesimiso sezinto eziphilayo

Isici esisodwa esivame ukunganakwa: ubani ophethe ukubuyekeza nokugcina imodeli , kanye namathuluzi azungezile.

Ekusetshenzisweni kwe -AI yendawo , unesibopho cishe sayo yonke into: ukufaka nokubuyekeza amamodeli, ukuphatha i-ONNX Runtime noma ezinye izikhathi zokusebenza, ukugcina abashayeli be-GPU, ukuqapha ukusebenza, ukulungisa ubuthakathaka, nokuqinisekisa ukuhambisana nezinhlelo zakho zokusebenza. Izixazululo ezifana ne -Windows ML, i-Foundry Local, noma i-Microsoft Foundry ku-Windows zenza impilo ibe lula ngokuhlanganisa ngqo amamodeli kuzinhlelo zokusebenza zedeskithophu noma ezisemaphethelweni, kodwa umthwalo wokusebenza usaphezu kwakho.

Njengoba i-AI isefwini , umhlinzeki uphatha izibuyekezo, ama-patches, izici ezintsha, kanye nokulinganisa kwangaphakathi . Amapulatifomu afana ne -Microsoft Foundry, i-Azure AI Services, kanye ne-Azure OpenAI Service anikeza ama-API nama-SDK athuthukiswe kahle, ahlanganiswe ne -Azure DevOps, i-GitHub Copilot, i-Semantic Kernel , kanye nezinye izinsizakalo ze-DevOps. Eqenjini lokuthuthukisa, lokhu kuhumusha ekugxileni kakhulu ekuqondeni kwebhizinisi futhi kungagcini ekunakekeleni ingqalasizinda.

I-ecosystem nayo ibalulekile: abasizi abaningi bokuthuthukiswa okusekelwe efwini njenge -GitHub Copilot, i-Cursor, i-Claude Code, noma i-Google Gemini Code Assist bathembele kulezi zinsizakalo zamafu, banikeza ama-plugin e-IDE, ama-CLI, namathuluzi okuhlaziya ikhodi acwebezelisiwe kakhulu, ngezinhlelo zamanani ezisukela emazingeni amahhala kuya kokubhaliselwe kwe-Pro noma ukukhokha ngokusetshenziswa ngakunye nge-API.

I-AI Yendawo Yabathuthukisi Namabhizinisi: Izinzuzo Nokungekuhle

Kwabathuthukisi nezinhlangano eziningi, ukusetha indawo ye-AI endaweni sekuyindlela ekhangayo kakhulu, kokubili ngezizathu zobumfihlo kanye nokonga izindleko zesikhathi eside. Inhlanganisela evamile ivame ukuba yi -VS Code + izandiso ezifana ne-Continue noma i-Cline + i-LLM yendawo nge-Ollama noma iseva ezinikele.

Le ndlela ivumela yonke ikhodi nemibuzo ukuthi ihlale emshinini wakho , okubaluleke kakhulu uma usebenza ngempahla yengqondo ebucayi kakhulu noma ikhodi engakwazi ukuvezwa kubantu besithathu. Amamodeli omthombo ovulekile njenge- Qwen, i-Code Llama, i-Llama 3, i-Mistral, noma izinhlobo eziqondiswe kuhlelo lokusebenza zingasebenza kahle ngokumangazayo, ikakhulukazi ngobukhulu obuphakathi (14B, 32B) kuma-GPU amahle.

Intengo okufanele uyikhokhe ukuthi udinga ihadiwe enamandla (isibonelo, i-GPU ene-VRAM okungenani engu-8 GB ukuze usebenzise kahle amamodeli angu-7-14B) futhi kufanele ubhekane nokucushwa ngesandla, ukulinganisa, ukulungiswa, kanye nezibuyekezo. Kuyindlela ephelele yezinkampani ezinedatha ebucayi kakhulu , amalebhu obuchwepheshe, amaqembu ajabulela ukulungisa, kanye nalabo abafuna ukugwema ukubhalisa kwanyanga zonke.

Emkhakheni webhizinisi, izinkampani eziningi zokubonisana kanye nabathuthukisi besofthiwe abangokwezifiso bangochwepheshe ekuklameni izakhiwo ezisezindaweni noma ezihlanganisiwe ezihlanganisa ukuzenzekela kwenqubo, ama-ejenti ahlakaniphile, ukuphepha kwe-inthanethi, kanye nobuhlakani bebhizinisi. Izinkampani ezisebenza ebhange, kwezempilo, noma emikhakheni elawulwayo zivame ukukhetha i-AI esezindaweni ukuze zicubungule idatha ebucayi kanye nefu lokuhlaziya okuhlanganisiwe noma izinsizakalo ezibhekene namakhasimende.

I-AI efwini: amathuba, izimo zokusetshenziswa kanye nezinselele

Ngakolunye uhlangothi, i-AI esekelwe efwini isibe yindinganiso yangempela yezinhlangano eziningi ezifuna isivinini, ukukhulisa, kanye nokufinyelela kumamodeli asezingeni eliphezulu ngaphandle kwesidingo sokwakha isikhungo sazo sedatha. Yilapho bobabili abahlinzeki be-hyperscalers abakhulu (i-Azure, i-AWS, i-Google Cloud) kanye nabahlinzeki be-inference abakhethekile (i-runpod, i-vast.ai, i-together, i-deepinfra, njll.) bephumelela khona.

Ezindaweni zebhizinisi, i-cloud AI ivumela ukuzenzekela kwemisebenzi ephindaphindwayo, ukuhlaziya okubikezelayo, ukutholwa kokukhwabanisa, ukulungiswa kwempahla, kanye nokusekelwa kwamakhasimende amahora angama-24 ngosuku, izinsuku eziyi-7 ngesonto ngama-chatbot athuthukile. Imikhakha efana nokuthengisa nokukhiqiza ithembele kakhulu efwini ukuhlaziya ukuthengisa, ukulungiswa kokubikezela, kanye nombono womshini emigqeni yokukhiqiza.

Ngokusho kwedatha yemboni, iLatin America ibhekene nokukhula okunamandla kokutshalwa kwezimali kwengqalasizinda yamafu , okuqhutshwa ngqo amaphrojekthi e-AI. Ukusetshenziswa kwamamodeli akhiqizayo ukudala okuqukethwe, ukuncoma imikhiqizo, ukuhlaziya amadokhumenti, nokusiza ekwenzeni izinqumo kusiza ekunciphiseni ukuhlukana kwedijithali nezifunda ezivuthiwe njengeNyakatho Melika neYurophu.

Izihloko zemakethe ziqokomisa ukuthi iphesenti eliphezulu kakhulu lezinkampani esezisebenzisa ifu nazo zisebenzisa izinsizakalo ze-AI ezinikezwa yilabo banikezeli abafanayo , ngokuvamile kuma-multi-cloud noma ama-hybrid architectures. Amamodeli Olimi Olukhulu (ama-LLM) awasetshenziswa kuphela kuma-chatbot: aphinde anikeze amandla amathuluzi angaphakathi, abasizi bokubhala, izinjini zokusesha ezihlakaniphile, kanye nezixazululo zokuhlaziya amadokhumenti ezinkulu.

Ama-ejenti azimele kanye ne-Agency Era: kungani kubaluleke kakhulu

Ngokufika kwama -ejenti e-AI azimele (i-Agent Era) , angaphenduli nje kuphela kodwa futhi enza (asebenzise imiyalo, ashintshe izinhlelo, aphathe izinqubo), ukukhetha phakathi kwendawo kanye nefu kuba bucayi kakhulu.

Ekusethweni kwe -ejenti yendawo , yonke indlela yokusebenzisana, izinqumo, kanye nomjikelezo wesenzo isebenza ngaphakathi kwendawo yakho: amalogi, idatha yamakhasimende, imiyalo yezinhlelo zangaphakathi, njalo njalo. Lokhu kunikeza ukulawula okuphezulu kanye nokulandeleka , okubalulekile emikhakheni elawulwayo. Ungabuyekeza indlela abenza ngayo izinqumo, uhlole ukuziphatha kwabo, futhi unciphise ububanzi babo ngokunemba okukhulu.

Efwini , ukusebenzisa ama-ejenti kulula kakhulu: abahlinzeki banikeza izinhlaka, ama-orchestrator, kanye nezixhumi ezilungele ukuhlanganiswa ne-SaaS, ama-CRM, ama-ERP, nazo zonke izinhlobo zama-API angaphandle. Kodwa futhi zikhulisa izingozi zokuphepha nobumfihlo : ukulungiswa okungalungile noma isibuyekezo somthengisi esingahlolwanga kahle singagcina sithinte ulwazi lwamakhasimende akho noma sidalule idatha ngokungahlosile.

Sekuvele kube nezimo zangempela zezinkampani lapho ama-ejenti amafu aqale ukukhipha izimpendulo ezingaguquki noma eziyingozi ngemuva kokushintsha kwamamodeli angaphansi kwamandla awo. Lokhu kubonisa ngokusobala ukuthi ukulahlekelwa ukulawula okuphelele kweketanga kungaba yinkinga kangakanani ezindaweni zama-ejenti.

I-AI emaphethelweni: lapho udinga ukwenza izinqumo okwamanje futhi ngaphandle kwe-inthanethi

Esinye isiqephu okufanele sifakwe kulesi sibalo yi -edge AI , uhlobo olusondelene kakhulu lwe-AI yendawo egxile kumadivayisi axhunyiwe.

Umqondo ulula: susa ukubala eduze ngangokunokwenzeka lapho idatha ikhiqizwa khona . Esikhundleni sokuthumela zonke izithombe kusuka kukhamera yezimboni ziye efwini ukuze zihlaziywe, zicutshungulwa ngqo kudivayisi enqenqemeni enemodeli esivele ilayishiwe. Lokhu kusebenza futhi kuma-smartwatches, ama-smartphone, izinzwa ku-supply chain, kanye nezikrini zezokwelapha ezigqokwayo.

Lokhu kwenza kube nokwenzeka ukwenza izinqumo ngesikhathi sangempela , noma ngabe uxhumano lwe-inthanethi luhlala lukhona noma alukho, futhi kunciphisa ithrafikhi yenethiwekhi, izindleko ze-bandwidth, kanye nokuthembela ezikhungweni zedatha ezikude. I-AI enqenqemeni iyathakazelisa kakhulu ekukhiqizeni , ekuthuthweni kwempahla, ekuphathweni kwamandla, kanye nasekunakekelweni kwezempilo , lapho ukubambezeleka nokuqina kubalulekile khona.

Ngesikhathi esifanayo, lawa madivayisi asemaphethelweni angahlanganiswa nezinsizakalo zamafu zokuqeqesha, ukuhlanganiswa kwedatha, kanye nokuhlanganiswa komhlaba wonke , ngaleyo ndlela kufezwe ukwakheka kwe-AI okunamandla kakhulu okusatshalaliswa.

Amasu ahlanganisiwe kanye nezindlela ezisebenzayo zokukhetha

Konke okungenhla kusiholela ephuzwini elisobala: akuvamile ukuba kube nomnqobi ocacile phakathi kwe-AI esezindaweni kanye ne-AI esekelwe efwini . Izinhlangano eziningi ezivuthiwe zigcina zinendlela ehlanganisiwe, ngokuklama ngamabomu ukuthi yini ehamba kuphi.

Isu elifanele lamaphrojekthi amaningi lihilela ukusebenzisa i-AI yendawo: ulwazi olubucayi kakhulu (idatha yezokwelapha, yezezimali, kanye ne-R&D), ama-chatbot angaphakathi ezinkampani, ukuzenzekela kwenqubo yasehhovisi, kanye nezimo ezingenakho ukuxhumana okuqinisekisiwe. Lapha, okubaluleke kakhulu ubumfihlo, ukulawula, kanye nezindleko ezingabikezelwa.

Ngesikhathi esifanayo, kunengqondo ukusebenzisa i-AI efwini: abasizi bomphakathi, izinsizakalo ezinabasebenzisi abayizinkulungwane ngasikhathi sinye, ukuhlolwa okusheshayo kwamacala amasha okusetshenziswa, ukuhlaziywa kwedatha okukhulu, kanye nokufinyelela kumamodeli asemngceleni ongeke ukwazi ukuwasingatha ngaphakathi. Lapha, ukusabalala, ukulula, kanye nesikhathi sokumaketha kubalulekile.

Ezimweni lapho kokubili kudingeka khona, ungaklama izakhiwo ze-hybrid kanye ne-multicloud : ingxenye yomsebenzi iphathwa endaweni (ukuhlanzwa kwedatha, ukungachazwa, izinqumo eziphuthumayo) kanti enye ingxenye idluliselwa efwini (ukuqeqeshwa, ukuhlaziywa okuhlanganisiwe, ukuphetha okunzima lapho kukhona ukuxhumana).

Isihluthulelo ukuhlola ngokucophelela, ngaphambi kokunquma, izici ezifana nengozi yokulawula, ukuzwela kwedatha, iphrofayili yethrafikhi, umkhathi wokukhula, isabelomali, ithalente lobuchwepheshe elitholakalayo, kanye nokuthembela okwamukelekayo kubantu besithathu . Njengoba lokho kusobala, kulula kakhulu ukunquma ukuthi kungcono yini ukutshala imali eningi ezixazululweni zendawo, efwini, noma endaweni ephakathi ecatshangelwe kahle.

Esimweni lapho cishe zonke izinkampani sezivele zisebenza efwini ngandlela thile futhi ngesikhathi esifanayo zifuna ukukhulisa imbuyiselo ekutshalweni kwazo kwe-AI ngaphandle kwezindleko noma izingozi ezikhuphukayo, ukuhlanganisa ngobuhlakani i-AI yendawo, i-AI enqenqemeni kanye ne-AI yamafu kuba yindlela ephumelelayo : usebenzisa okungcono kakhulu kwendawo ngayinye, uvumelanisa ukuphepha, ukusebenza kanye nesabelomali, futhi ugcina ukuguquguquka okwanele ukuze uzivumelanise nobuchwepheshe obushintsha ngokushesha.