Ubuhanga bwo gukora imibonano mpuzabitsina (AI) muri Cloud Computing ni iki?

Ubuhanga bwo gukora imibonano mpuzabitsina (AI) muri Cloud Computing ni iki?

Igisubizo kigufi: Ubuhanga bwo gukora imibonano mpuzabitsina mu buryo bwa “cloud computing” bujyanye no gukoresha porogaramu za “cloud platforms” mu kubika amakuru, gukodesha “compute”, guhugura “models”, kuyakoresha nk'amaserivisi, no kuyakurikirana mu musaruro. Ni ingenzi kuko amakosa menshi aturuka ku makuru, kuyakoresha, n'imikorere, ntabwo ari imibare. Niba ukeneye “surveillance” yihuse cyangwa “releases” zishobora gusubirwamo, “cloud + MLOps” niyo nzira ifatika.

Ibintu by'ingenzi byakunzwe:

Ubuzima : Gushyira amakuru ku murongo, kubaka ibintu, guhugura, gushyiraho, hanyuma ukurikirane uburyo ibintu bigenda bihinduka, gutinda, n'ikiguzi.

Ubuyobozi : Kubaka uburyo bwo kugenzura uburyo bwo kwinjira, inyandiko z'igenzura, no gutandukanya ibidukikije kuva mu ntangiriro.

Kongera gukosorwa : Kwandika verisiyo z'amakuru, kode, ibipimo, n'ibidukikije kugira ngo ibikorwa bikomeze gusubirwamo.

Kugenzura ikiguzi : Koresha uburyo bwo gukusanya amakuru, kuyabika mu buryo bwihuse, kuyapima mu buryo bwikora, n'uburyo bwo kuyakoresha mu kwirinda gutakaza amafaranga.

Imiterere yo gushyiraho urubuga : Hitamo urubuga rucungwa, imikorere ya lakehouse, Kubernetes, cyangwa RAG hashingiwe ku kuri kw'ikipe.

Ubuhanga bwo gukora imibonano mpuzabitsina (AI) muri Cloud Computing ni iki? Infographic

Ingingo ushobora gukunda gusoma nyuma y'iyi:

🔗 Ibikoresho by'ingenzi byo gucunga ubucuruzi bwa AI mu bicu
Gereranya urubuga runini rwo mu bicu runoza imikorere, imari, n'amatsinda.

🔗 Ikoranabuhanga rikenewe kugira ngo habeho ikoranabuhanga rinini rya AI
Ibikorwa remezo by'ingenzi, amakuru, n'ubuyobozi birakenewe kugira ngo GenAI ishyirwe mu bikorwa.

🔗 Ibikoresho bya AI ku buntu byo gusesengura amakuru
Uburyo bwiza bwo gukora isuzumabumenyi ku buntu bwo gusukura, kwerekana no kwerekana amakuru.

🔗 Serivisi yo gukora imibonano mpuzabitsina n'ubwenge (AI) ni iki?
Isobanura AIAaS, inyungu, ibiciro, n'uburyo busanzwe bwo gukoresha mu bucuruzi.

Ubuhanga bwo gukora imibonano mpuzabitsina mu bicu: Ibisobanuro byoroshye 🧠☁️

Mu by'ingenzi, ubuhanga bwo gukora imibonano mpuzabitsina mu buryo bwa “cloud computing” bivuze gukoresha porogaramu zikoresha ikoranabuhanga rya “cloud platforms” kugira ngo ugere kuri:

Aho kugura ibikoresho byawe bihenze, ukodesha ibyo ukeneye, iyo ubikeneye NIST SP 800-145 . Nk'uko wakoresha siporo yo gukora imyitozo ngororamubiri imwe ikomeye aho kubaka siporo muri gareji yawe hanyuma ntuzongere gukoresha treadmill ukundi. Bibaho ku beza muri twe 😬

Mu buryo bweruye: ni AI ikora ku rugero, yohereza, ivugurura, kandi igakora binyuze mu bikorwa remezo bya cloud NIST SP 800-145 .

Impamvu AI + Cloud ari ikibazo gikomeye cyane 🚀

Reka tuvugishe ukuri - imishinga myinshi ya AI ntitsindwa kubera ko imibare igoye. Iratsindwa kubera ko "ibintu bikikije icyitegererezo" bivangavanze:

  • amakuru aratatanye

  • ibidukikije ntibihuye

  • moderi ikora kuri mudasobwa igendanwa y'umuntu ariko nta handi hantu

  • Gushyira mu bikorwa bifatwa nk'igitekerezo cya nyuma

  • umutekano n'amategeko biraza byibuze nk'umubyara utatumiwe 😵

Porogaramu zo mu bicu zirafasha kuko zitanga:

1) Igipimo cya elastic 📈

Toza icyitegererezo ku itsinda rinini ry’ibikoresho mu gihe gito, hanyuma ugifunge NIST SP 800-145 .

2) Igerageza ryihuse ⚡

Kora vuba porogaramu za mudasobwa zicungwa, imiyoboro yubatswe mbere, na porogaramu za GPU. Google Cloud: GPUs za AI .

3) Gushyira mu bikorwa byoroshye 🌍

Shyiraho moderi nka API, imirimo ya batch, cyangwa serivisi zishyirwa kuri interineti Red Hat: API ya REST ni iki? SageMaker Batch Transform .

4) Urusobe rw'amakuru rwahujwe 🧺

Imiyoboro yawe y'amakuru, ububiko, n'isesengura akenshi biba biri mu bicu. AWS: Ububiko bw'amakuru vs ikiyaga cy'amakuru .

5) Ubufatanye n'imiyoborere 🧩

Uruhushya, inyandiko z’igenzura, verisiyo, n’ibikoresho bisangiwe bikorerwa mu bitabo bya Azure ML registries (MLOps) .

Uburyo AI muri Cloud Computing ikora mu buryo bufatika (Umuvuduko Nyawo) 🔁

Dore uko ubuzima busanzwe bumeze. Ntabwo ari verisiyo "y'igishushanyo mbonera cyiza" ... ni iyabayeho.

Intambwe ya 1: Amakuru agwa mu bubiko bw'ibicu 🪣

Ingero: amasanduku yo kubika ibintu, ibiyaga by'amakuru, ububiko bw'amakuru bw'ibicu Amazon S3 (ububiko bw'ibintu) AWS: Ikiyaga cy'amakuru ni iki? Incamake ya Google Cloud Storage .

Intambwe ya 2: Gutunganya amakuru + kubaka imikorere 🍳

Urayisukura, ukayihindura, ukarema ibintu biyiranga, wenda ukayishyira kuri interineti.

Intambwe ya 3: Imyitozo yo kwitoza 🏋️

Ukoresha cloud compute (akenshi GPU) kugira ngo uhugure Google Cloud: GPUs kuri AI :

Intambwe ya 4: Gushyira mu bikorwa 🚢

Moderi zipakiye kandi zigatangwa binyuze muri:

Intambwe ya 5: Gukurikirana + amakuru mashya 👀

Indirimbo:

  • gutinda

  • Uburyo bwo kugenzura SageMaker Model Monitor

  • ihinduka ry'amakuru Vertex AI Model

  • ikiguzi kuri buri giteganyijwe

  • ibintu bikubangamira cyane bigatuma uvuga uti “ibi ntibikwiye kubaho…” 😭

Iyo ni moteri. Iyo ni AI muri Cloud Computing iri kugenda, ntabwo ari nk'igisobanuro gusa.

Ni iki gituma verisiyo nziza ya AI muri Cloud Computing iba nziza? ✅☁️🤖

Niba ushaka ishyirwa mu bikorwa "ryiza" (atari icyitegererezo gitangaje gusa), ibande kuri ibi:

A) Gutandukanya ibibazo neza 🧱

  • urwego rw'amakuru (ububiko, imiyoborere)

  • urwego rw'amahugurwa (igerageza, imiyoboro)

  • urwego rwo gutanga (API, gupima)

  • urwego rw'igenzura (ibipimo, inyandiko, amatangazo) SageMaker Model Monitor

Iyo byose bihujwe, gukosora amakosa biba kwangiza amarangamutima.

B) Uburyo bwo kongera kubyaza umusaruro ubusanzwe 🧪

Sisitemu nziza igufasha kuvuga, udakoresheje ukuboko:

  • amakuru yatoje ubu buryo

  • verisiyo ya kode

  • ibipimo bya hyperparameter

  • ibidukikije

Niba igisubizo ari “uhh, ndatekereza ko byari ukwiruka kuwa kabiri…” usanzwe uri mu kaga 😅

C) Igishushanyo mbonera gishingiye ku giciro 💸

Cloud AI ifite imbaraga, ariko ni nabwo buryo bworoshye bwo gukora umushinga w'itegeko utera kwibaza ku mahitamo yawe y'ubuzima.

Imiterere myiza irimo:

D) Umutekano no kubahiriza amategeko bikubiye muri 🔐

Ntabwo ihambiriweho umugozi nyuma nk'aho umuyoboro w'amazi uva.

E) Inzira nyayo kuva ku igenamigambi kugeza ku ikorwa 🛣️

Iyi ni yo nkuru. "Verisiyo" nziza ya AI mu bicu irimo MLOps, imiterere yo kuyikoresha, no kuyikurikirana kuva ku ntangiriro Google Cloud: MLOps ni iki? . Bitabaye ibyo ni umushinga w'iserukiramuco rya siyansi ufite inyemezabwishyu nziza.

Imbonerahamwe yo Kugereranya: Amahitamo AI Azwi cyane mu bicu (n'abo agenewe) 🧰📊

Hasi hari imbonerahamwe yihuse kandi ifite ibitekerezo birambuye. Ibiciro ni byinshi ku bushake kuko ibiciro byo mu bicu ari nko gutumiza ikawa - igiciro fatizo ntabwo ari cyo giciro 😵💫

Igikoresho / Urubuga Abareba Igiciro gikwiye Impamvu ikora (harimo n'inyandiko zidasanzwe)
AWS SageMaker Amakipe ya ML, ibigo Wishyure uko ugenda Urubuga rwa ML rwuzuye - amahugurwa, imipaka, imiyoboro. Ifite imbaraga, ariko hari menus hose.
Google Vertex AI (AI) Amatsinda ya ML, amashami y'ubumenyi bw'amakuru Wishyure uko ugenda Amahugurwa akomeye acungwa neza + iyandikwa ry'icyitegererezo + iyuzuzanya. Birumva byoroshye iyo bikanda.
Kwiga imashini za Azure Ibigo, imiryango ishingiye kuri MS Wishyure uko ugenda Ikina neza na Azure ecosystem. Amahitamo meza y'imiyoborere, hari uduce twinshi.
Databricks (ML + Lakehouse) Amatsinda akomeye y'ubuhanga mu by'ikoranabuhanga Kwiyandikisha + ikoreshwa Ni byiza cyane mu kuvanga imiyoboro y'amakuru + ML ahantu hamwe. Akenshi bikundwa n'amatsinda akorera mu buryo bufatika.
Ibiranga ubuhanga bwo gukora imibonano mpuzabitsina ku rubura (Snowflake AI) Isosiyete za mbere mu gusesengura Hashingiwe ku mikoreshereze Biba byiza iyo isi yawe isanzwe iri mu bubiko. “Laboratwari ya ML” nkeya, “AI” nyinshi muri SQL
IBM watsonx Inganda zigenzurwa Ibiciro by'ibigo Imiyoborere n'igenzura ry'ibigo ni ibintu by'ingenzi cyane. Akenshi bitoranywa mu gushyiraho politiki nyinshi.
Kubernetes icungwa (DIY ML) Injeniyeri za platform Ihinduka Yoroshye kandi isanzwe. Nanone… iyo ububabare bucitse uba ufite uburenganzira bwo kubugira 🙃
Igitekerezo kidakoresha seriveri (imikorere + iherezo) Amakipe y'ibicuruzwa Hashingiwe ku mikoreshereze Ni byiza cyane ku modoka zirimo urujya n'uruza rw'imodoka. Reba uko imodoka zitangira kandi zitinda nk'iy'umukara.

Ibi ntabwo ari uguhitamo "icyaba cyiza" - ahubwo ni uguhuza n'ukuri kw'ikipe yawe. Iryo ni ibanga rituje.

Uburyo Busanzwe bwo Gukoresha AI muri Cloud Computing (Hamwe n'Ingero) 🧩✨

Aha niho uburyo bwo gushyiraho porogaramu za AI-in-cloud bukora neza cyane:

1) Uburyo bwo gufasha abakiriya 💬

  • abafasha mu biganiro

  • uburyo bwo gutumiza amatike

  • incamake

  • yo kumenya amarangamutima n'umugambi mu bicu

2) Sisitemu zo gutanga inama 🛒

  • ibitekerezo ku bicuruzwa

  • ibikubiye mu nkuru

  • “Abantu nabo baguze”
    Ibi akenshi biba bikeneye ibisobanuro birambuye no kuvugururwa mu gihe nyacyo.

3) Gutahura uburiganya no gutanga amanota ku byago 🕵️

Cloud yorohereza guhangana n'ibitero bitunguranye, gukurikirana ibikorwa, no kuyobora amatsinda.

4) Ubutasi bw'inyandiko 📄

  • Imiyoboro ya OCR

  • gukuramo imiterere

  • isesengura ry'amasezerano

  • Gusesengura inyemezabuguzi Imikorere ya Snowflake Cortex AI
    Mu miryango myinshi, aha niho igihe gisubizwa bucece.

5) Guteganya no kunoza ubumenyi 📦

Guteganya ibikenewe, gutegura ububiko bw'ibikoresho, kunoza inzira. Cloud irafasha kuko amakuru ari menshi kandi kongera guhugura biba kenshi.

6) Porogaramu zikora za AI 🪄

  • gutegura ibikubiye mu nyandiko

  • ubufasha ku makode

  • roboti z'ubumenyi bw'imbere (RAG)

  • yo gukusanya amakuru hakoreshejwe uburyo bwa “synthetic retrieval-Augmented Generation (RAG)”
    Iyi ni yo mpamvu amasosiyete akunze kuvuga ati: “Tugomba kumenya aho amategeko yacu yo kubona amakuru aherereye.” 😬

Imiterere y'Ubwubatsi Uzabona Ahantu Hose 🏗️

Icyitegererezo cya 1: Urubuga rwa ML rucungwa (inzira ya "turashaka ko ibibazo by'umutwe bike") 😌

Ikora neza iyo umuvuduko ari ngombwa kandi ntushaka kubaka ibikoresho by'imbere kuva ku busa.

Icyitegererezo cya 2: Lakehouse + ML (inzira ya "gutanga amakuru mbere na mbere") 🏞️

  • guhuza ikoranabuhanga ry'amakuru + imikorere ya ML

  • koresha amakaye, imiyoboro, ikoranabuhanga ry'ibintu hafi y'amakuru

  • ikomeye ku miryango isanzwe ikorera muri sisitemu nini zo gusesengura Databricks Lakehouse

Icyitegererezo cya 3: ML iri mu gikoresho kuri Kubernetes (inzira ya "turashaka kugenzura") 🎛️

  • ingero z'amapaki mu bikoresho

  • urwego hamwe na politiki yo gupima ubwikorezi Kubernetes: Gupima Pod itambitse

  • guhuza serivisi, uburyo bwo kugaragara, amabanga mgmt

Bizwi kandi nka: “Dufite icyizere, kandi dukunda gukemura ibibazo mu masaha adasanzwe.”

Icyitegererezo cya 4: RAG (Gukuramo ibintu byiyongera) (inzira yo "gukoresha ubumenyi bwawe") 📚🤝

Iki ni igice cy'ingenzi cy'ibiganiro bigezweho bya AI-in-cloud kuko ari cyo gikoresho kinini nyacyo gikoresha AI ibyara umusaruro mu buryo bwizewe.

MLOps: Igice buri wese asuzugura 🧯

Niba ushaka ko ubuhanga bwa AI mu bicu bwitwara neza mu musaruro, ukeneye MLOps. Si ukubera ko ari ikintu kigezweho - kuko moderi zigenda zihinduka, amakuru ahinduka, kandi abakoresha bakaba bahanga mu buryo bubi cyane . Google Cloud: MLOps ni iki?

Ibice by'ingenzi:

Iyo wirengagije ibi, uzasanga ufite "icyitegererezo cya pariki y'inyamaswa" 🦓 aho byose biba ari bizima, nta kintu na kimwe cyanditseho, kandi utinya gufungura irembo.

Umutekano, Ubuzima bwite, no Kuyubahiriza (Si igice cyo kwishimisha, ariko… Yego) 🔐😅

Ubuhanga bwo gukora imibonano mpuzabitsina mu buryo bwa “cloud computing” butera ibibazo bike bikomeye:

Kugenzura uburyo bwo kubona amakuru 🧾

Ni nde ushobora kubona amakuru y'amahugurwa? Inyandiko z'ibitekerezo? Ibitekerezo? Ibisubizo?

Guhisha no kubika amakuru ahishe 🗝️

Imfunguzo, tokens, n'ibyangombwa bikeneye gukoreshwa neza. "Muri dosiye yo gushyiraho" ntabwo ari ugukoresha.

Kwigunga no gukodesha 🧱

Hari imiryango ikenera ahantu hatandukanye ho gukora, gukora, no gukora. Cloud irafasha - ariko gusa iyo uyishyizeho neza.

Igenzura ry'ibaruramari 📋

Imiryango igenzurwa ikunze kugomba kugaragaza:

  • ni ayahe makuru yakoreshejwe

  • uburyo ibyemezo byafashwe

  • ni nde washyizeho icyo

  • igihe byahinduraga IBM watsonx.governance

Uburyo bwo gucunga ingaruka z'icyitegererezo ⚠️

Ibi birimo:

  • kugenzura uruhande rubogamye

  • isuzuma ry'inzitizi

  • ubwirinzi bwihuse bwo gutera inshinge (ku bijyanye n'ubukorano bw'ubukorano)

  • kuyungurura umusaruro mu buryo bwizewe

Ibi byose bigaruka ku ngingo: si "ubukorano bwa gihanga bukorerwa kuri interineti gusa." Ni ubukorano bwa gihanga bukorerwa munsi y'inzitizi nyazo.

Inama ku biciro n'imikorere (kugira ngo utazarira nyuma) 💸😵💫

Inama nke zageragejwe n'intambara:

  • Koresha moderi ntoya cyane ihuye n'ibyo ukeneye
    . Inini si ko buri gihe iba nziza. Hari igihe iba nini gusa.

  • Gusobanura mu byiciro byose igihe bishoboka
    ihendutse kandi ikora neza Batch Transform .

  • Bika
    cyane cyane ku bibazo bisubirwamo no gusimbuza.

  • Kwiyongerera ubushobozi bwo gukora igenzura, ariko gabanya
    ubushobozi bwo gukora igenzura ritagira umupaka bishobora gusobanura amafaranga atagira umupaka Kubernetes: Horizontal Pod Kwiyongerera ubushobozi bwo gukora igenzura . Mbaza uko mbizi… mu by'ukuri, ntubikore 😬

  • Kurikirana ikiguzi kuri buri musozo na buri gikorwa.
    Bitabaye ibyo uzanoza ikintu kitari cyo.

  • Koresha ikoranabuhanga ryo kubara rishobora gukoreshwa mu mahugurwa.
    Uzigama amafaranga menshi niba akazi kawe ko guhugura gashobora guhangana n'ibibazo. Amazon EC2 Spot Instances Google Cloud Preemptible VMs .

Amakosa Abantu Bakora (Ndetse n'amakipe y'ubwenge) 🤦♂️

  • Gufata ikoranabuhanga rya "cloud anit" nk'aho ari "gushyiramo moderi gusa"

  • Kwirengagiza ireme ry'amakuru kugeza ku munota wa nyuma

  • Kohereza icyitegererezo nta kugenzura SageMaker Model Monitor

  • Ntabwo nteganya kongera gutoza cadence Google Cloud: MLOps ni iki?

  • Twibagiwe ko amakipe y'umutekano ahari kugeza mu cyumweru cyo gutangiza 😬

  • Gukoresha ubuhanga burenze urugero kuva ku munsi wa mbere (rimwe na rimwe ibintu byoroshye bitsinda)

Nanone, ikintu kibi cyane: amakipe agabanya uburyo abakoresha banga gutinda. Icyitegererezo kidahuje neza gato ariko cyihuta akenshi kiratsinda. Abantu ntibagira kwihangana mu bitangaza bito.

Ibintu by'ingenzi 🧾✅

Ubuhanga bwo gukora imibonano mpuzabitsina mu buryo bwa Cloud Computing ni uburyo bwuzuye bwo kubaka no gukoresha ubuhanga bwo gukora imibonano mpuzabitsina hakoreshejwe ibikorwa remezo by'ibicu - kongera amahugurwa, koroshya ishyirwa mu bikorwa, guhuza imiyoboro y'amakuru, no gushyira mu bikorwa moderi hamwe na MLOps, umutekano, n'ubuyobozi Google Cloud: MLOps ni iki? NIST SP 800-145 .

Incamake y'incamake:

  • Cloud iha AI ibikorwa remezo byo kwagura no kohereza 🚀 NIST SP 800-145

  • AI itanga "ubwonko" bw'imirimo yo mu bicu butuma ibyemezo byikora 🤖

  • Ubumaji si imyitozo gusa, ahubwo ni ugushyira mu bikorwa, gukurikirana, no kuyobora 🧠🔐 SageMaker Model Monitor

  • Hitamo urubuga rushingiye ku byo ikipe ikeneye, ntabwo ari ugukoresha ikoranabuhanga ribi 📌

  • Reba ibiciro n'ibikorwa nk'igikona cyambaye indorerwamo 🦅👓 (ikigereranyo kibi, ariko urabyumva)

Niba waraje hano utekereza uti "AI muri cloud computing ni API y'icyitegererezo gusa," oya - ni ibidukikije byose. Rimwe na rimwe ni byiza, rimwe na rimwe bihindagurika, rimwe na rimwe byombi ku gicamunsi kimwe 😅☁️

Ibibazo Bikunze Kubazwa

Icyo "AI" mu bijyanye na mudasobwa mu bicu (cloud computing) bivuze mu magambo asanzwe

Ubuhanga bwo gukora imibonano mpuzabitsina mu bicu bivuze ko ukoresha urubuga rwo kubika amakuru, gukora imashini zikora imashini (CPUs/GPUs/TPUs), gutoza moderi, kuzikoresha, no kuzikurikirana - nta bikoresho byawe bwite. Mu by'ukuri, igicu kiba ahantu ubuzima bwawe bwose bwa AI bukorera. Ukodesha ibyo ukeneye igihe ubikeneye, hanyuma ukagabanya igihe urangije.

Impamvu imishinga ya AI inanirwa nta bikorwa remezo byo mu buryo bwa cloud na MLOps

Ibyinshi mu bibazo bibaho hafi ya moderi, ntabwo bibaho imbere muri yo: amakuru adahuye, ibidukikije bidahuye, uburyo bworoshye bwo gukoresha porogaramu, kandi nta gukurikirana. Uburyo bwo gukoresha porogaramu mu bicu bufasha mu guhuza uburyo bwo kubika, kubara, no gukoresha porogaramu kugira ngo moderi zidafata umwanya kuri "byakoraga kuri mudasobwa yanjye igendanwa." MLOps yongeraho kole ibura: gukurikirana, kwandika, imiyoboro, no gusimbuza porogaramu kugira ngo sisitemu ikomeze kuba nziza kandi ikomeze kubungabungwa.

Imikorere isanzwe ya AI mu bijyanye na mudasobwa ikoresha ikoranabuhanga ryo mu bicu, kuva ku makuru kugeza ku musaruro

Uburyo busanzwe bwo gukoresha amakuru ni ubu: amakuru agwa mu bubiko bw'ibicu, agatunganywa mu buryo bufatika, hanyuma models zigahuzwa ku buryo bwo kubara bushobora kwaguka. Hanyuma, ukoresha ukoresheje API endpoint, batch job, serverless setup, cyangwa serivisi ya Kubernetes. Amaherezo, ukurikirana latency, drift, n'ikiguzi, hanyuma ukongera ukoresheje retraining no deployments zitekanye. Imiyoboro myinshi nyayo ihora ihuza aho kohereza rimwe.

Guhitamo hagati ya SageMaker, Vertex AI, Azure ML, Databricks, na Kubernetes

Hitamo ukurikije ukuri kw'ikipe yawe, ntabwo ari urusaku rwo kwamamaza "rwa platform nziza". Amashyirahamwe ya ML acungwa (SageMaker / Vertex AI / Azure ML) agabanya ibibazo by'imikorere binyuze mu mahugurwa, amashami, ububiko, no gukurikirana. Databricks akenshi zihuza amakipe menshi ashinzwe igenamigambi ry'amakuru ashaka ko ML ijya hafi y'imiyoboro n'isesengura. Kubernetes itanga ubuyobozi n'uburyo bwo guhindura amakuru, ariko kandi ufite icyizere, politiki zo kwagura, no gukemura ibibazo iyo ibintu byangiritse.

Imiterere y'inyubako igaragara cyane mu miterere ya AI cloud muri iki gihe

Uzabona imiterere ine buri gihe: platforms za ML zicungwa kugira ngo zikoreshwe ku muvuduko, lakehouse + ML ku ma orgs ya data-first, ML iri mu bubiko kuri Kubernetes kugira ngo igenzurwe, na RAG (gukoresha ubumenyi bwacu bw'imbere mu mutekano). RAG ikunze kuba irimo inyandiko mu bubiko bw'ibicu, embeddings + ububiko bwa vector, urwego rwo kugarura amakuru, n'uburyo bwo kugenzura ubwinjirire hamwe no kwandika amakuru. Imiterere uhitamo igomba guhuza imiyoborere yawe n'igihe ugeze ku gihe.

Uburyo amakipe akoresha moderi za cloud AI: REST APIs, batch jobs, serverless, cyangwa Kubernetes

REST APIs zikunze kugaragara mu guhanura mu gihe nyacyo iyo gutinda k'umusaruro ari ngombwa. Isuzuma ry'ibice byinshi ni ryiza mu gutanga amanota yagenwe no gukoresha neza ikiguzi, cyane cyane iyo ibisubizo bidakenewe ko biba ako kanya. Amaherezo adafite seriveri ashobora gukora neza ku bantu bafite uruziga rwinshi, ariko gutangira gukonje no gutinda bisaba kwitabwaho. Kubernetes ni nziza iyo ukeneye gupima no guhuza neza ibikoresho bya platform, ariko byongera imiterere y'imikorere.

Ibyo ugomba gukurikirana mu musaruro kugira ngo sisitemu za AI zikomeze kugira ubuzima bwiza

Nibura, kurikirana igihe cyo gutinda, igipimo cy'amakosa, n'ikiguzi kuri buri giteganyijwe kugira ngo icyizere n'ingengo y'imari bikomeze kugaragara. Ku ruhande rwa ML, kurikirana ihindagurika ry'amakuru n'ihindagurika ry'imikorere kugira ngo umenye igihe ukuri guhindutse muri moderi. Gukurikirana ibibazo by'inyuma n'umusaruro mubi nabyo ni ingenzi, cyane cyane ku bibazo by'ikoreshwa ry'amakuru aho abakoresha bashobora kuba abashyamiranye mu buryo bw'ubuhanga. Gukurikirana neza binashyigikira ibyemezo byo gusubira inyuma iyo moderi zisubiye inyuma.

Kugabanya ikiguzi cya AI yo mu bicu nta mikorere ya tanki

Uburyo busanzwe ni ugukoresha moderi ntoya yujuje ibisabwa, hanyuma ukanoza inference ukoresheje batching na caching. Autoscaling irafasha, ariko ikeneye imipaka kugira ngo "elastic" idahinduka "amafaranga atagira imipaka." Ku mahugurwa, compute ishobora kuzigama amafaranga menshi niba akazi kawe kihanganira guhagarara. Gukurikirana ikiguzi kuri buri endpoint na buri kintu bikurinda kunoza igice kitari cyo cya sisitemu.

Ingaruka zikomeye ku mutekano no kubahiriza amategeko hamwe n'ubuhanga bwo gukora imibonano mpuzabitsina mu buryo bw'ikoranabuhanga (AI) mu bicu

Ingaruka zikomeye ni ukubona amakuru mu buryo butagenzurwa, gucunga amabanga make, no kubura inzira zo kugenzura amakuru ku muntu watoje kandi agakoresha iki. Ikoranabuhanga rya AI ryongera ibibazo nko gutera amakuru vuba, gutanga amakuru mu buryo butekanye, no kugaragaza amakuru y’ingenzi mu nyandiko. Imiyoboro myinshi ikenera uburyo bwo kwirinda ibidukikije (dev/staging/prod) na politiki isobanutse neza yo gutanga amakuru, gutanga amakuru, no kwandika amakuru y’inyongera. Imiterere yizewe ifata ubuyobozi nk'ikintu cy'ingenzi gisabwa muri sisitemu, ntabwo ari igikorwa cyo gutangiza icyumweru.

Amareferensi

  1. Ikigo cy'Igihugu cy'Ubuziranenge n'Ikoranabuhanga (NIST) - SP 800-145 (Iheruka) - csrc.nist.gov

  2. Google Cloud - GPU za AI - cloud.google.com

  3. Google Cloud - Inyandiko za TPU zo mu gicu - docs.cloud.google.com

  4. Serivisi za interineti za Amazon (AWS) - Amazon S3 (ububiko bw'ibintu) - aws.amazon.com

  5. Serivisi za interineti za Amazon (AWS) - Ikiyaga cy'amakuru ni iki? - aws.amazon.com

  6. Serivisi za interineti za Amazon (AWS) - Ububiko bw'amakuru ni iki? - aws.amazon.com

  7. Serivisi za interineti za Amazon (AWS) - Serivisi za AWS AI - aws.amazon.com

  8. Google Cloud - Google Cloud AI APIs - cloud.google.com

  9. Google Cloud - MLOps ni iki? - cloud.google.com

  10. Google Cloud - Vertex AI Model Registry (Intangiriro) - docs.cloud.google.com

  11. Ingofero Itukura - API Y'ICYUBAHIRO ni iki? - redhat.com

  12. Inyandiko za Amazon Web Services (AWS) - SageMaker Batch Transform - docs.aws.amazon.com

  13. Serivisi za interineti za Amazon (AWS) - Ububiko bw'amakuru ugereranije n'uruziga rw'amakuru ugereranije n'amakuru y'amakuru - aws.amazon.com

  14. Microsoft Learn - Azure ML registries (MLOps) - learn.microsoft.com

  15. Google Cloud - Incamake y'ububiko bwa Google Cloud - docs.cloud.google.com

  16. arXiv - Inyandiko yo Kugarura Ibikomoka ku Buzima Budasanzwe (RAG) - arxiv.org

  17. Inyandiko za Amazon Web Services (AWS) - SageMaker Serverless Inference - docs.aws.amazon.com

  18. Kubernetes - Horizontal Pod Autoscaling - kubernetes.io

  19. Google Cloud - Vertex AI batch - docs.cloud.google.com

  20. Inyandiko za Amazon Web Services (AWS) - SageMaker Model Monitor - docs.aws.amazon.com

  21. Google Cloud - Vertex AI Model Monitoring (Gukoresha modelling mode) - docs.cloud.google.com

  22. Serivisi za interineti za Amazon (AWS) - Amazon EC2 Spot Instances - aws.amazon.com

  23. Google Cloud - Preemptible VMs - docs.cloud.google.com

  24. Inyandiko za Amazon Web Services (AWS) - AWS SageMaker: Uko ikora (Amahugurwa) - docs.aws.amazon.com

  25. Google Cloud - Google Vertex AI - cloud.google.com

  26. Microsoft Azure - Kwiga imashini za Azure - azure.microsoft.com

  27. Databricks - Databricks Lakehouse - databricks.com

  28. Inyandiko za Snowflake - Ibiranga Snowflake AI (Ubuyobozi bw'incamake) - docs.snowflake.com

  29. IBM - IBM watsonx - ibm.com

  30. Google Cloud - Inyandiko za API z'ururimi karemano rwa Cloud - docs.cloud.google.com

  31. Inyandiko za Snowflake - Imikorere ya Snowflake Cortex AI (AI SQL) - docs.snowflake.com

  32. MLflow - Gukurikirana MLflow - mlflow.org

  33. MLflow - Urwego rwo Kwandika Icyitegererezo cya MLflow - mlflow.org

  34. Google Cloud - MLOps: Uburyo buhoraho bwo gutanga no gukoresha ikoranabuhanga mu kwigira ku mashini - cloud.google.com

  35. Serivisi za interineti za Amazon (AWS) - Iduka ry'Ibiranga SageMaker - aws.amazon.com

  36. IBM - IBM watsonx.governance - ibm.com

Shaka ubuhanga bwa AI bugezweho mu iduka ryemewe rya AI Assistant Store

Ku bijyanye natwe

Garuka kuri blog