Ingabe i-AI izothatha indawo yonjiniyela bedatha? [Ividiyo kanye nombuzo]

Impendulo emfushane: I-AI ngeke ithathe indawo yonjiniyela bedatha ngokuphelele; izokwenza umsebenzi ophindaphindwayo ube ngokuzenzakalelayo njengokubhala i-SQL, ukusika amapayipi, izivivinyo, kanye namadokhumenti. Uma indima yakho ikakhulukazi ingumnikazi ophansi, umsebenzi oqhutshwa amathikithi, ivele kakhulu; uma unokwethenjelwa, izincazelo, ukuphatha, kanye nokuphendula kwezigameko, i-AI ikwenza usheshe kakhulu.

Izinto ezibalulekile okufanele uzicabangele:

Ubunikazi: Beka phambili ukuzibophezela ngemiphumela, hhayi nje ukukhiqiza ikhodi ngokushesha.

Ikhwalithi: Ukuhlolwa kokwakha, ukubonwa, kanye nezinkontileka ukuze amapayipi ahlale ethembekile.

Ukubusa: Gcina ubumfihlo, ukulawula ukufinyelela, ukugcina, kanye nezindlela zokuhlola eziphethwe ngabantu.

Ukumelana nokusetshenziswa kabi: Phatha imiphumela ye-AI njenge-drafts; yibuyekeze ukuze ugweme ukungalungi okuqinisekile.

Ukushintsha indima: Chitha isikhathi esincane uthayipha i-boilerplate nesikhathi esiningi ukuklama izinhlelo eziqinile.

Uma uchithe imizuzu engaphezu kwemihlanu uzungeze amaqembu edatha, uke wezwa inkulumo ethi “remark” - ngezinye izikhathi ihletshwa, ngezinye izikhathi iqaliswa emhlanganweni ofana noshintsho lwendaba: Ingabe i-AI izothatha indawo yonjiniyela bedatha?

Futhi… Ngiyakuqonda. I-AI ingakhiqiza i-SQL, yakhe amapayipi, ichaze umkhondo wezitaki, idwebe amamodeli e-dbt, iphakamise ngisho nezinhlelo zokugcina impahla ngokuzethemba okukhathazayo. I-GitHub Copilot ye-SQL Mayelana namamodeli e-dbt I-GitHub Copilot
Kuzwakala sengathi ukubuka i-forklift ifunda ukuxuba. Kuyamangalisa, kuyathusa kancane, futhi awuqiniseki ngokugcwele ukuthi kusho ukuthini ngomsebenzi wakho 😅

Kodwa iqiniso alihlelekile njengesihloko. I-AI ishintsha ngokuphelele ubunjiniyela bedatha. Izenzakalela izingcezu ezingathandeki neziphindaphindekayo. Isheshisa izikhathi zokuthi “Ngiyazi engikufunayo kodwa angikhumbuli i-syntax”. Iphinde idale izinhlobo ezintsha zesiphithiphithi.

Ngakho-ke masiyibeke kahle, ngaphandle kokulindela noma ukwesaba okusongela ikusasa.

Izihloko ongase uthande ukuzifunda ngemva kwalesi:

🔗 Ingabe i-AI izothatha indawo yodokotela bemisebe?
Indlela ukuthwebula izithombe ze-AI okushintsha ngayo ukuhamba komsebenzi, ukunemba, kanye nezindima zesikhathi esizayo.

🔗 Ingabe i-AI izothatha indawo yababhali bezimali?
Bona ukuthi yimiphi imisebenzi yokubala eyenziwa yi-AI ngokuzenzakalelayo nokuthi yini ehlala ingumuntu.

🔗 Ingabe i-AI izothatha indawo yamabhange okutshalwa kwezimali?
Qonda umthelela we-AI kumadili, ucwaningo, kanye nobudlelwano namakhasimende.

🔗 Ingabe i-AI izothatha indawo yama-ejenti omshwalense?
Funda ukuthi i-AI iguqula kanjani ukubhala phansi, ukuthengisa, kanye nokusekelwa kwamakhasimende.

Kungani umbuzo othi “i-AI ithatha indawo yonjiniyela bedatha” uqhubeka uvela kabusha 😬

Ukwesaba kuvela endaweni ethile: ubunjiniyela bedatha bunomsebenzi omningi ophindaphindwayo.

• Ukubhala nokulungisa i-SQL

• Izikripthi zokungenisa izakhiwo

• Ukumaka amasimu kusuka kwesinye isimiso kuya kwesinye

• Ukudala izivivinyo kanye nemibhalo eyisisekelo

• Ukulungisa amaphutha ezinkinga zamapayipi... uhlobo lokubikezela

I-AI inhle kakhulu kumaphethini aphindaphindwayo. Futhi ingxenye enkulu yobunjiniyela bedatha injalo kanye - amaphethini ahlanganiswe kumaphethini. Iziphakamiso zekhodi ye-GitHub Copilot

Futhi, uhlelo lwamathuluzi seluvele "lufihla" ubunzima:

• Izixhumi ze-ELT eziphethwe Amadokhumenti e-Fivetran

• Ikhompyutha ye- AWS Lambda (ikhompyutha engenaseva)

• Ukuhlinzekwa kwempahla yokugcina impahla ngokuchofoza okukodwa

• Ukuhlelwa kokukhulisa ngokuzenzakalelayo amadokhumenti e-Apache Airflow

• Uhlaka lokuguqulwa oluchazayo Kuyini i-dbt?

Ngakho-ke uma i-AI ifika, ingazwakala sengathi iyingcezu yokugcina. Uma inqwaba isivele isusiwe, futhi i-AI ingabhala ikhodi yeglue… yini esele? 🤷

Kodwa nansi into abantu abayigwemayo: ubunjiniyela bedatha abukona ukuthayipha kuphela. Ukuthayipha yingxenye elula. Ingxenye enzima ukwenza iqiniso lebhizinisi elifiphele, lezepolitiki, nelishintshashintshayo liziphathe njengohlelo oluthembekile.

Futhi i-AI isabhekene nalolo daba. Abantu nabo bayahlupheka - bamane bazenzela izinto ezintsha kangcono.

Lokho onjiniyela bedatha abakwenzayo usuku lonke (iqiniso elibi) 🧱

Masikhulume iqiniso - isihloko somsebenzi esithi “Data Engineer” sizwakala sengathi wakha izinjini zamarokhethi ngezibalo ezimsulwa. Empeleni, wakha ukwethembana.

Usuku olujwayelekile aludingi “ukusungula ama-algorithms amasha” kodwa ludinga okunye:

• Ukuxoxisana namaqembu asezingeni eliphezulu mayelana nezincazelo zedatha (kubuhlungu kodwa kuyadingeka)

• Ukuphenya ukuthi kungani i-metric ishintshile (nokuthi ingabe ingokoqobo)

• Ukusingatha ukuzulazula kweskimu kanye nokumangala kokuthi “othile wengeze ikholomu phakathi kwamabili”

• Ukuqinisekisa ukuthi amapayipi awakwazi ukubuthwa, ayabuyiselwa, futhi ayaqapheleka

• Ukwakha izivikelo ukuze abahlaziyi abangaphansi komfula bangakhi ngengozi amadeshibhodi angenangqondo

• Ukuphatha izindleko ukuze i-warehouse yakho ingaphenduki ibe umlilo omkhulu 🔥

• Ukuqinisekisa ukufinyelela, ukuhlola, ukuthobela imithetho, izinqubomgomo zokugcina izimpahla Izimiso ze-GDPR (iKhomishini yaseYurophu) Umkhawulo wesitoreji (ICO)

• Ukwakha imikhiqizo yedatha abantu abangayisebenzisa ngempela ngaphandle kokukuthumelela i-DM imibuzo engu-20

Ingxenye enkulu yomsebenzi imayelana nomphakathi kanye nokusebenza:

• "Ubani ongumnikazi waleli tafula?"

• "Ingabe le ncazelo isasebenza?"

• "Kungani i-CRM ithumela kwamanye amazwe amakhophi aphindaphindiwe?"

• "Singakwazi yini ukuthumela lesi silinganiso kubaphathi ngaphandle kokuhlazeka?" 😭

I-AI ingasiza ngezingxenye zalokhu, yebo. Kodwa ukuyishintsha ngokuphelele... kuyinkimbinkimbi.

Yini eyenza inguqulo eqinile yendima yobunjiniyela bedatha ibe yinguqulo eqinile? ✅

Lesi sigaba sibalulekile ngoba inkulumo yokufaka esikhundleni ivame ukuthi onjiniyela bedatha ikakhulukazi “bangabakhi bamapayipi.” Lokho kufana nokucabanga ukuthi abapheki “basika imifino” kakhulu. Kuyingxenye yomsebenzi, kodwa akuwona umsebenzi.

Inguqulo eqinile yonjiniyela wedatha ivame ukusho ukuthi bangenza okuningi kwalokhu:

• Umklamo wokushintsha
  Izinguquko zedatha. Amaqembu ayashintsha. Amathuluzi ayashintsha. Unjiniyela omuhle wakha izinhlelo ezingawi njalo lapho iqiniso lithimula 🤧

• Chaza izinkontileka kanye nokulindelekile
  Kusho ukuthini ukuthi “ikhasimende”? Kusho ukuthini ukuthi “okusebenzayo”? Kwenzekani uma ingxabano ifika sekwephuzile? Izinkontileka zivimbela isiphithiphithi ngaphezu kwekhodi yefancy. I-Open Data Contract Standard (ODCS) I-ODCS (GitHub)

• Yakha ukubonwa kukho konke
  Hhayi nje ukuthi “kwasebenza” kodwa “kwasebenza kahle.” Ukusha, ukungalingani kwevolumu, ukuqhuma okungenamsebenzi, ukushintsha kokusatshalaliswa. Ukubonwa kwedatha (i-Dynatrace) Kuyini ukubonwa kwedatha?

• Yenza ukuguquguquka okufana nokushesha komuntu omdala
  uma kuqhathaniswa nokunemba, izindleko uma kuqhathaniswa nokubambezeleka, ukuguquguquka uma kuqhathaniswa nokuba lula. Ayikho ipayipi ephelele, amapayipi kuphela ongaphila ngawo.

• Guqula izidingo zebhizinisi zibe izinhlelo eziqinile
  Abantu bacela izilinganiso, kodwa abakudingayo umkhiqizo wedatha. I-AI ingabhala ikhodi, kodwa ayikwazi ukwazi ngokuyisimangaliso izimayini zomhlaba zebhizinisi.

• Gcina idatha ithule
  Into encomekayo kakhulu ngeplatifomu yedatha ukuthi akekho okhuluma ngayo. Idatha engazelelwe iyidatha enhle. Njengokufaka amapayipi. Uyibona kuphela uma yehluleka 🚽

Uma wenza lezi zinto, umbuzo othi “Ingabe i-AI izothatha indawo yonjiniyela bedatha?” uqala ukuzwakala sengathi awulungile kancane. I-AI ingathatha indawo yemisebenzi, hhayi ubunikazi.

Lapho i-AI isivele isiza onjiniyela bedatha (futhi kuhle kakhulu) 🤖✨

I-AI ayiyona nje ukumaketha. Uma isetshenziswa kahle, iyisiphindaphindi samandla esisemthethweni.

1) Umsebenzi we-SQL osheshayo kanye nokuguqulwa

• Ukudweba ama-complex joins

• Imisebenzi yefasitela yokubhala ongathanda ukuyicabanga

• Ukuguqula i-logic yolimi olulula ibe yi-skeletons yemibuzo

• Ukulungisa imibuzo emibi ibe yi-CTEs efundekayo ye-GitHub Copilot ye-SQL

Lokhu kukhulu ngoba kunciphisa umphumela "wekhasi elingenalutho". Kusadingeka uqinisekise, kodwa uqala ku-70% esikhundleni sika-0%.

2) Ukulungisa amaphutha kanye nokususa izinhlayiya zesinkwa ezibangela izimpande

I-AI ilungile ku:

• Ukuchaza imiyalezo yamaphutha

• Iphakamisa ukuthi kufanele ubheke kuphi

• Ukuncoma izinyathelo zohlobo "lokuhlola ukungafani kweskimu" GitHub Copilot
  Kufana nokuba nonjiniyela omncane ongakhathali ongalali futhi ngezinye izikhathi oqamba amanga ngokuzethemba 😅

3) Ukubhala imibhalo kanye nokucebisa ikhathalogi yedatha

Kukhiqizwe ngokuzenzakalela:

• Izincazelo zekholomu

• Izifinyezo zemodeli

• Izincazelo zozalo

• “Leli tafula lisetshenziselwani?” imibhalo ye-dbt

Akuphelele, kodwa kuphula isiqalekiso semipayipi engabhalisiwe.

4) Hlola i-scaffolding bese uhlola

I-AI ingaphakamisa:

• Ukuhlolwa okuyisisekelo okungenalutho

• Ukuhlolwa kobungqayizivele

• Imibono yobuqotho bokubhekisela

• "Lesi silinganiso akufanele sinciphe" izitatimende zesitayela se-dbt izivivinyo zedatha Ukulindela Okukhulu: Ukulindela

Futhi - usanquma ukuthi yini ebalulekile, kodwa kuyasheshisa izingxenye ezijwayelekile.

5) Ikhodi "yeglu" yepayipi

Amathempulethi okulungiselela, izisekelo ze-YAML, ukuhlelwa kwe-DAG drafts. Lezo zinto ziyaphindaphindwa futhi i-AI idla ngokuphindaphindwayo kwasekuseni 🥣 Ama-Apache Airflow DAG

Lapho i-AI isabhekene nobunzima (futhi lokhu kuyimbangela yayo) 🧠🧩

Lena yingxenye ebaluleke kakhulu, ngoba iphendula umbuzo wokufaka esikhundleni ngokuthungwa kwangempela.

1) Izincazelo ezingacacile nezishintshashintshayo

Ukuqonda kwebhizinisi akuvamile ukuba kube lula. Abantu bashintsha izingqondo zabo phakathi nomusho. “Umsebenzisi osebenzayo” uba “umsebenzisi okhokhayo osebenzayo” uba “umsebenzisi okhokhayo osebenzayo ngaphandle kokubuyiselwa kwemali ngaphandle kwezinye izikhathi”… uyazi ukuthi kunjani.

I-AI ayikwazi ukuba nelungelo lalokho kungaqondakali. Ingaqagela kuphela.

2) Ukuziphendulela kanye nengozi

Uma ipayipi liphuka futhi ideshibhodi ye-exec ikhombisa ukungabi nangqondo, othile kufanele:

• ukuhlelwa kweqembu

• umthelela wokuxhumana

• lungisa

• vimbela ukuphinda kwenzeke

• bhala ukuhlolwa kwesidumbu

• nquma ukuthi ibhizinisi lisengazethemba yini izinombolo zesonto eledlule

I-AI ingasiza, kodwa ayikwazi ukuphendula ngendlela enenjongo. Izinhlangano azisebenzisi imizwa - zisebenza ngomthwalo wemfanelo.

3) Ukucabanga kwezinhlelo

Amapulatifomu edatha ayizinto eziphilayo: ukungenisa, ukugcina, ukuguqulwa, ukuhlela, ukuphatha, ukulawula izindleko, ama-SLA. Ushintsho ku-ripples yesendlalelo esisodwa. Imiqondo ye-Apache Airflow

I-AI ingaphakamisa ukulungiswa kwendawo okudala ubuhlungu emhlabeni jikelele. Kufana nokulungisa umnyango ocijile ngokususa umnyango 😬

4) Ukuphepha, ubumfihlo, ukuthobela imithetho

Yilapho amaphupho okufaka esikhundleni efa khona.

• Izilawuli zokufinyelela

• Ukuphepha kwezinga lomugqa Izinqubomgomo zokufinyelela umugqa weqhwa Ukuphepha kwezinga lomugqa we-BigQuery

• Ukuphathwa kwe-PII Uhlaka Lobumfihlo lwe-NIST

• Imithetho yokugcina Umkhawulo wesitoreji (ICO) Isiqondiso se-EU mayelana nokugcina

• Imizila yokuhlola i- NIST SP 800-92 (ukuphathwa kwamalogi) Ukulawulwa kwe-CIS 8 (Ukuphathwa kwamalogi okuhlola)

• Imikhawulo yokuhlala kwedatha

I-AI ingabhala izinqubomgomo, kodwa ukuzisebenzisa ngokuphephile kuwubunjiniyela bangempela.

5) “Izinto ezingaziwa”

Izehlakalo zedatha zivame ukungalindeleki:

• I-API yomthengisi ishintsha izincazelo buthule

• Ukuqagela kwendawo yesikhathi kuyashintsha

• I-backfill iphinda kabili i-partition

• Indlela yokuzama kabusha ibangela ukubhala okuphindwe kabili

• Isici esisha somkhiqizo sethula amaphethini amasha emicimbi

I-AI iba buthaka uma isimo singeyona iphethini eyaziwayo.

Ithebula Lokuqhathanisa: yini enciphisa lokho, empeleni 🧾🤔

Ngezansi umbono osebenzayo. Hhayi “amathuluzi athatha indawo yabantu,” kodwa amathuluzi nezindlela ezinciphisa imisebenzi ethile.

Qaphela ukuthi yini engekho: umugqa othi “cindezela inkinobho ukuze ususe onjiniyela bedatha.” Yebo… lowo mugqa awukho 🙃

Ngakho-ke... ingabe i-AI izothatha indawo yoNjiniyela Bedatha, noma izomane ishintshe indima? 🛠️

Nansi impendulo engeyona ephawulekayo: I-AI izothatha indawo yezingxenye zomsebenzi, hhayi umsebenzi.

Kodwa kuzoyilungisa kabusha indima. Futhi uma ungayinaki leyo, uzozizwa ucindezelekile.

Yini eshintshayo:

• Isikhathi esincane sokubhala i-boilerplate

• Isikhathi esincane sokusesha amadokhumenti

• Isikhathi esiningi sokubuyekeza, ukuqinisekisa, ukuklama

• Isikhathi esiningi sokuchaza izinkontileka kanye nokulindelwe kwekhwalithi I-Open Data Contract Standard (ODCS)

• Isikhathi esiningi sibambisana nomkhiqizo, ezokuphepha, ezezimali

Lolu ushintsho oluncane: ubunjiniyela bedatha abubalulekanga kangako “ngokwakha amapayipi” kodwa bubaluleka kakhulu “ngokwakha uhlelo lomkhiqizo wedatha oluthembekile.”

Futhi ngokushintshana okuthule, lokho kuyigugu kakhulu, hhayi kancane.

Futhi - futhi ngizokusho lokhu noma kuzwakala kumangalisa - i-AI yandisa inani labantu abangakhiqiza izinto zobuciko zedatha, okwandisa isidingo sokuba othile agcine yonke into iphilile. Ukuphuma okwengeziwe kusho ukudideka okwengeziwe. I-GitHub Copilot

Kufana nokunikeza wonke umuntu ithuba lokubhoboza ngamandla. Kuhle! Manje othile udinga ukuphoqelela umthetho othi “ngicela ungabhobozi epayipini lamanzi” 🪠

Iqoqo elisha lamakhono elihlala liwusizo (ngisho ne-AI yonke indawo) 🧠⚙️

Uma ufuna uhlu lokuhlola olusebenzayo "oluvikela ikusasa", lubukeka kanje:

Indlela yokucabanga ngokuklama uhlelo

• Ukumodela kwedatha okusinda ekushintsheni

• Ukuhwebelana kweqembu vs ukusakaza

• Ukubambezeleka, izindleko, ukucabanga kokuthembeka

Ubunjiniyela bekhwalithi yedatha

• Izinkontileka, ukuqinisekiswa, ukutholwa kwe-anomaly I-Open Data Contract Standard (ODCS) Ukubonwa kwedatha (i-Dynatrace)

• Ama-SLA, ama-SLO, imikhuba yokusabela ezigamekweni

• Ukuhlaziywa kwembangela eyinhloko ngesiyalo (hhayi ama-vibes)

Ukwakhiwa kokubusa kanye nokwethembana

• Amaphethini okufinyelela

• Ukuhlolwa kwe- NIST SP 800-92 (ukuphathwa kwamalogi)

• Ubumfihlo ngomklamo Uhlaka Lobumfihlo lwe-NIST

• Ukuphathwa komjikelezo wokuphila kwedatha Isiqondiso se-EU mayelana nokugcinwa kwedatha

Ukucabanga kwepulatifomu

• Amathempulethi angasetshenziswa kabusha, izindlela zegolide

• Amaphethini ajwayelekile okungenisa, ukuguqula, ukuhlola -Fivetran dbt

• Amathuluzi okuzisebenzela wona angancibiliki

Ukuxhumana (yebo, ngempela)

• Ukubhala amadokhumenti acacile

• Izincazelo zokuqondanisa

• Ukuthi “cha” ngenhlonipho kodwa ngokuqinile

• Ukuchaza ukungezwani ngaphandle kokuzwakala njengerobhothi 🤖

Uma ukwazi ukwenza lokhu, umbuzo othi “Ingabe i-AI izothatha indawo yonjiniyela bedatha?” awube usabeka kakhulu. I-AI iba yi-exoskeleton yakho, hhayi indawo yakho.

Izimo ezingokoqobo lapho ezinye izindima zobunjiniyela bedatha zincipha khona 📉

Kulungile, hlola iqiniso ngokushesha, ngoba akuyona yonke into ehlekisayo ne-emoji confetti 🎉

Ezinye izindima zibonakala kakhulu:

• Izindima zokungenisa kuphela lapho konke kuyizixhumi ezijwayelekile Izixhumi ze-Fivetran

• Amaqembu enza imibiko ephindaphindwayo kakhulu ngokungabi nalutho kwesizinda

• Ama-Org lapho ubunjiniyela bedatha buphathwa “njengezinkawu ze-SQL” (kunonya, kodwa kuyiqiniso)

• Imisebenzi yobunikazi obuphansi lapho umsebenzi umane nje ungowokukopisha nokunamathisela

Ubuchwepheshe be-AI kanye namathuluzi aphethwe kahle kunganciphisa lezo zidingo.

Kodwa ngisho nalapho, ukushintshwa kuvame ukubukeka kanje:

• Bambalwa abantu abenza umsebenzi ofanayo ophindaphindayo

• Ukugcizelela kakhulu ubunikazi beplatifomu nokuthembeka kwayo

• Ukushintshela kokuthi “umuntu oyedwa angasekela amapayipi amaningi”

Ngakho yebo - amaphethini esibalo sabantu angashintsha. Izindima ziyashintsha. Izihloko ziyashintsha. Leyo ngxenye ingokoqobo.

Noma kunjalo, inguqulo yendima enobunikazi obuphezulu nethembekile kakhulu isalokhu ikhona.

Isifinyezo sokuvala 🧾✅

Ingabe i-AI izothatha indawo yoNjiniyela Bedatha? Hhayi ngendlela ehlanzekile nephelele abantu abacabanga ngayo.

I-AI izokwenza lokhu:

• imisebenzi ephindaphindwayo ezenzakalelayo

• ukusheshisa ukubhala amakhodi, ukulungisa amaphutha, kanye nemibhalo I-GitHub Copilot yemibhalo ye-SQL dbt

• ukunciphisa izindleko zokukhiqiza amapayipi

Kodwa ubunjiniyela bedatha bumayelana kakhulu nalokhu:

• ukuziphendulela

• umklamo wesistimu

• ukwethembana, ikhwalithi, kanye nokuphatha i-Open Data Contract Standard (ODCS) NIST Privacy Framework

• ukuhumusha iqiniso lebhizinisi elifiphele libe imikhiqizo yedatha ethembekile

I-AI ingasiza ngalokho… kodwa “ayiyona eyakho”.

Uma ungunjiniyela wedatha, isinyathelo silula (akulula, kodwa silula):
gxila ekubeni ngumnikazi, ikhwalithi, ukucabanga kwepulatifomu, kanye nokuxhumana. Vumela i-AI iphathe i-boilerplate ngenkathi uphatha izingxenye ezibalulekile.

Futhi yebo - ngezinye izikhathi lokho kusho ukuthi ungumuntu omdala ekamelweni. Akuyona into ekhangayo. Kodwa-ke unamandla kancane 😄

Ingabe i-AI izothatha indawo yoNjiniyela Bedatha?
Izothatha indawo yeminye imisebenzi, ishintshe ileli, futhi yenze onjiniyela bedatha abangcono kakhulu babe yigugu nakakhulu. Yileyo indaba yangempela.

Isibonelo sangempela: Ukwakha umsebenzi wokubuyekezwa kwepayipi ledatha elisizwa yi-AI 🛠️

Isimo

Cabanga ngenkampani encane ye-ecommerce enonjiniyela wedatha oyedwa, abahlaziyi ababili, kanye nenkinga ejwayelekile kakhulu: ideshibhodi yezezimali iyaqhubeka nokungasebenzi noma nini lapho umhlinzeki wezinkokhelo eshintsha igama lensimu.

Ithimba alifuni ukuthi i-AI "ibe ngumnikazi" wombhobho. Lokho kungaba yingozi. Esikhundleni salokho, basebenzisa i-AI njengomsizi wohlaka lokuqala lomsebenzi ojwayelekile kodwa obalulekile: ukubhala ama-skeleton emodeli ye-dbt, ukusikisela izivivinyo, ukubhala amadokhumenti, nokudala uhlu lokuhlola lokubuyekezwa kwekhodi.

Unjiniyela wedatha yomuntu usengumnikazi womklamo wokugcina, izincazelo zedatha, imithetho yokufinyelela, kanye nokusetshenziswa komkhiqizo. I-AI imane isheshise ibanga eliphakathi eliyinkimbinkimbi.

Okudingwa umsebenzi wokusebenza

Ngaphambi kokusebenzisa i-AI, ithimba linikeza umongo owanele ukuze ube wusizo:

• Uhlelo lwethebula lokukhokha olukhona

• Izincazelo zesilinganiso sezimali eziqondiwe, njengokuthi “imali engenayo eseleyo”, “inani lembuyiselo”, kanye “nenkokhelo ekhokhelwe”

• Imithetho yokuqamba amagama yamamodeli e-dbt

• Izibonelo zezivivinyo ezivunyiwe

• Inkontileka yedatha emfushane yokuphakelwa kwezinkokhelo

• Imithetho yokusingatha i-PII, izinkokhelo ezihlulekile, amakhophi aphindwe kabili, kanye namarekhodi afika sekwephuzile

• Isibonelo sezigameko zesikhathi esidlule, okuhlanganisa ukuthi yini ehambe kabi nokuthi yalungiswa kanjani

Isihluthulelo akukhona ukuthi “cela i-AI yakhe ipayipi”. Lokho akucaci kahle.

Indlela enamandla yile: “Nansi imithetho yethu, nansi uhlelo, nansi indlela yokuziphatha elindelekile. Bhala okuthile esingakubuyekeza.”

Isibonelo semiyalelo

Usiza ekubhaleni imodeli ye-dbt yedatha yethu yokukhokha. Sebenzisa i-schema nemithetho engezansi ukudala imodeli yokupasa kokuqala, izivivinyo ze-dbt eziphakanyisiwe, kanye namanothi okubhala amadokhumenti.

Imodeli kumele ibale imali engenayo elungisiwe yansuku zonke nge-order_id kanye nomhlinzeki_wokukhokha. Khipha izinkokhelo ezihlulekile, khipha ukuthengiselana kokuhlola, bese ususa imbuyiselo kuphela uma i-refund_status = “iqinisekisiwe”.

Ungaqambi amakholomu. Uma ikholomu edingekayo ingekho, yibhale phansi ngaphansi kokuthi “Imibuzo yokubuyekezwa komuntu” esikhundleni sokuqagela.

Futhi phakamisa izivivinyo zokuhluka, amanani angenalutho, amanani amukelwe, kanye nokuqondakala kwemali engenayo. Maka noma yikuphi ukucabanga okungathinta ukubika kwezezimali.

Indlela yokuyihlola

Ukuhlolwa okunengqondo kuncane futhi kuyinto evamile ngamabomu:

• Nikeza i-AI uhlelo lokukhokha olwaziwayo bese uhlola ukuthi iyakugwema yini ukusungula amasimu.

• Nikeza i-schema eyodwa enekholomu ye-refund_status engekho bese ubona ukuthi ibuza umbuzo esikhundleni sokuqagela.

• Sebenzisa i-SQL ekhiqizwe ngokumelene nesethi yedatha yesiteji, hhayi ukukhiqiza.

• Qhathanisa umphumela namarekhodi okukhokha angu-20 ahlolwe ngesandla.

• Cela umhlaziyi kanye nonjiniyela wedatha ukuthi babuyekeze izincazelo ngaphambi kokuhlanganisa.

• Engeza izivivinyo ezamukelwe ku-CI ukuze ipayipi liqhubeke lizihlola ngemva kokufakwa.

Into ebalulekile ukuhlola i-AI ngezindlela zokwehluleka ozesaba kakhulu: amakholomu akhiwe, i-logic yemali engenayo engalungile, ukuphathwa kwemali engenayo okungekho, kanye nemigqa ephindaphindwayo ethule.

Umphumela

Umphumela obonisayo: ngokusekelwe ekubekeni isikhathi imisebenzi emithathu yokushintsha ipayipi ngaphambi nangemva kokusebenzisa lo msebenzi.

Ngaphambi kokusebenzisa i-AI, unjiniyela wachitha cishe amahora ama-5 nemizuzu engama-30 ngoshintsho ngalunye: cishe amahora ama-2 ebhala i-SQL, ihora eli-1 edala izivivinyo, imizuzu engama-45 ebhala amadokhumenti, kanti okunye ukuhlola amacala angaphandle ngezezimali.

Njengoba i-AI isetshenziswe kuphela kuma-draft okuqala, uhlobo olufanayo loshintsho luthathe cishe amahora ama-2 nemizuzu eyi-10. Ukonga okukhulu kakhulu kuvele ekuhlolweni kwe-scaffolding kanye nama-draft amadokhumenti, okwehle kusukela ehoreni eli-1 nemizuzu engama-45 kuya cishe kumaminithi angama-25.

Isinyathelo sokubuyekezwa komuntu sisathathe cishe imizuzu engama-45, futhi akufanele sisuswe.

Esivivinyweni semisebenzi emithathu, i-AI yaphakamisa ukuhlolwa okungu-18. Unjiniyela wamukele okungu-11, wahlela okungu-5, futhi wenqaba okungu-2 ngoba babecabanga ukuthi imithetho yebhizinisi ayilona iqiniso. Ukuthi ukwenqatshwa kubalulekile: kufakazela ukuthi umsebenzi udinga ukubuyekezwa, hhayi ukwethenjwa okungaboni.

Yini engase ihambe kabi

I-AI ingenza ipayipi libukeke liphelele kakhulu kunalokho eliyikho.

Amaphuzu avamile okuhluleka afaka:

• Ukusungula amakholomu azwakala enengqondo

• Ukuphatha ukubuyiselwa kwemali, ukukhokhisa imali emuva, kanye nezinkokhelo ezihlulekile njengento efanayo

• Izinkinga zesikhathi ezingekho emalini engenayo yansuku zonke

• Ukusikisela ukuhlolwa okuvamile okungabambi amaphutha ezezimali

• Ukubhala imibhalo ezwakala iqinisekile kodwa ifihla ukungaqiniseki

• Ukukhohlwa imithetho yobumfihlo uma idatha yesampula iqukethe imininingwane yamakhasimende

Umthetho omuhle: I-AI ingaklama imodeli, kodwa umuntu kumele asayine izincazelo, ukuqonda imali, ukulawula ukufinyelela, kanye nokukhishwa komkhiqizo.

Ukudla okuwusizo

Inguqulo ewusizo ye-AI kubunjiniyela bedatha akuyona "ukufaka esikhundleni sonjiniyela wedatha". Iwukususa ikhasi elingenalutho, bese ubuyekeza kanzima".

Lokho kusho i-SQL esheshayo, ukuhlolwa okusheshayo, kanye nemibhalo engcono yokupasa kokuqala, kuyilapho unjiniyela esaphethe ingxenye ebaluleke kakhulu: ukuthi idatha ilungile, ithembekile, iphephile, futhi iyachazeka yini.

Imibuzo Evame Ukubuzwa

Ingabe i-AI izothatha indawo yonjiniyela bedatha ngokuphelele?

Ezinhlanganweni eziningi, i-AI inamathuba amaningi okuthatha imisebenzi ethile kunokusula indima ngokuphelele. Ingasheshisa ukubhalwa kwe-SQL, ukufakwa kwe-scaffolding yamapayipi, ukudlula kokuqala kwamadokhumenti, kanye nokudalwa kokuhlolwa okuyisisekelo. Kodwa ubunjiniyela bedatha buphinde bube nobunikazi kanye nokuziphendulela, kanye nomsebenzi ongemuhle wokwenza iqiniso lebhizinisi elingcolile liziphathe njengohlelo oluthembekile. Lezo zingxenye zisadinga abantu ukuthi banqume ukuthi "okulungile" kubukeka kanjani futhi bathathe umthwalo wemfanelo lapho izinto ziphuka.

Yiziphi izingxenye zobunjiniyela bedatha i-AI esezivele zizenzakalela?

I-AI isebenza kahle kakhulu emsebenzini ophindaphindwayo: ukubhala nokulungisa i-SQL, ukukhiqiza ama-skeleton emodeli ye-dbt, ukuchaza amaphutha avamile, nokukhiqiza uhlaka lwamadokhumenti. Ingaphinde ihlole izikali njengokuhlola okungenalutho noma okuhlukile futhi ikhiqize ikhodi yethempulethi ethi "glue" yamathuluzi okuhlanganisa. Ukunqoba umfutho - uqala eduze kwesisombululo esisebenzayo - kodwa kusadingeka uqinisekise ukunemba futhi uqinisekise ukuthi kufanelana nendawo okuyo.

Uma i-AI ikwazi ukubhala i-SQL kanye namapayipi, yini esele konjiniyela bedatha?

Okuningi: ukuchaza izinkontileka zedatha, ukuphatha ukugeleza kweskimu, nokuqinisekisa ukuthi amapayipi awabonakali, ayaqapheleka, futhi ayabuyiselwa. Onjiniyela bedatha bachitha isikhathi bephenya izinguquko ze-metric, bakha izivikelo zabasebenzisi abasezansi, kanye nokuphatha ukuhwebelana ngezindleko nokuthembeka. Umsebenzi uvame ukuhambisana nokwakha ukwethembana nokugcina ipulatifomu yedatha “ithule,” okusho ukuthi izinzile ngokwanele ukuthi akekho okufanele acabange ngakho nsuku zonke.

I-AI iwushintsha kanjani umsebenzi wansuku zonke wonjiniyela wedatha?

Ngokuvamile kunciphisa i-boilerplate kanye "nesikhathi sokubheka," ngakho-ke uchitha isikhathi esincane uthayipha futhi uchitha isikhathi esiningi ubuyekeza, uqinisekisa, futhi uklama. Lolo shintsho lushukumisa indima ekuchazeni okulindelwe, amazinga ekhwalithi, kanye namaphethini angasetshenziswa kabusha kunokubhala yonke into ngesandla. Empeleni, cishe uzokwenza umsebenzi wokubambisana owengeziwe nomkhiqizo, ukuphepha, kanye nezezimali - ngoba umkhiqizo wobuchwepheshe uba lula ukuwudala, kodwa kunzima ukuwulawula.

Kungani i-AI inenkinga ngezincazelo zebhizinisi ezingacacile njengokuthi “umsebenzisi osebenzayo”?

Ngenxa yokuthi i-logic yebhizinisi ayimi ndawonye noma ayiqondile - ishintsha phakathi kwephrojekthi futhi iyahlukahluka ngabathintekayo. I-AI ingabhala incazelo, kodwa ayikwazi ukuba ngumnikazi wesinqumo lapho izincazelo zishintsha noma zingqubuzana. Ubunjiniyela bedatha buvame ukudinga ukuxoxisana, ukubhala phansi izinkolelo-mbono, nokuguqula izidingo ezingacacile zibe yizinkontileka ezihlala isikhathi eside. Lowo msebenzi "wokuqondanisa abantu" uyisizathu esiyinhloko sokuthi indima inganyamalali ngisho noma amathuluzi ethuthuka.

Ingabe i-AI ingaphatha ukuphathwa kwedatha, ubumfihlo, kanye nokuthobela imithetho kusebenza ngokuphephile?

I-AI ingasiza ekubhaleni izinqubomgomo noma iphakamise izindlela, kodwa ukusetshenziswa okuphephile kusadinga ubunjiniyela bangempela kanye nokuqapha ngokucophelela. Ukuphatha kuhilela ukulawulwa kokufinyelela, ukuphathwa kwe-PII, imithetho yokugcina, izindlela zokuhlola, kanye nemikhawulo yokuhlala. Lezi yizindawo ezinobungozi obukhulu lapho "cishe kulungile" kungamukeleki khona. Abantu kumele baklame imithetho, baqinisekise ukuphoqelelwa, futhi bahlale benomthwalo wemfanelo ngemiphumela yokuthobela imithetho.

Yimaphi amakhono ahlala ewusizo konjiniyela bedatha njengoba i-AI ithuthuka?

Amakhono enza izinhlelo ziqine: ukucabanga kokuklama uhlelo, ubunjiniyela bekhwalithi yedatha, kanye nokulinganisa okugxile kwipulatifomu. Izinkontileka, ukubonwa, imikhuba yokuphendula izehlakalo, kanye nokuhlaziywa kwezimbangela eziyinhloko eziqondisiwe kuba kubaluleke kakhulu lapho abantu abaningi bengakhiqiza izinto zobuciko zedatha ngokushesha. Ukuxhumana nakho kuba yinto ehlukanisayo - ukuvumelanisa izincazelo, ukubhala amadokhumenti acacile, kanye nokuchaza ukuhwebelana ngaphandle kwedrama kuyingxenye enkulu yokugcina idatha ithembekile.

Yiziphi izindima zobunjiniyela bedatha ezisengozini enkulu ngenxa ye-AI kanye namathuluzi aphethwe?

Izindima ezigxile kakhulu ekungenisweni okuphindaphindiwe noma emipayipini yokubika ejwayelekile zibonakala kakhulu, ikakhulukazi lapho izixhumi ze-ELT eziphethwe zimboza imithombo eminingi. Umsebenzi wobunikazi obuphansi, oqhutshwa yithikithi ungancipha ngoba i-AI kanye nokuqalwa kunciphisa umzamo ngepayipi ngalinye. Kodwa lokhu kuvame ukubonakala sengathi bambalwa abantu abenza imisebenzi ephindaphindayo, hhayi "abangenabo onjiniyela bedatha." Izindima zobunikazi obuphezulu ezigxile ekuthembekeni, ikhwalithi, kanye nokwethenjwa zihlala ziqinile.

Ngingawasebenzisa kanjani amathuluzi afana ne-GitHub Copilot noma i-dbt ene-AI ngaphandle kokudala isiphithiphithi?

Phatha umkhiqizo we-AI njengohlaka, hhayi isinqumo. Sebenzisa ukukhiqiza ama-skeleton emibuzo, ukuthuthukisa ukufundeka, noma ukuhlolwa kwe-dbt kanye namadokhumenti, bese uqinisekisa ngokumelene nedatha yangempela kanye namacala onqenqema. Hlanganisa nemigomo eqinile: izinkontileka, amazinga okuqamba amagama, ukuhlolwa kokubuka, kanye nemikhuba yokubuyekeza. Umgomo ukulethwa okusheshayo ngaphandle kokulahla ukuthembeka, ukulawula izindleko, noma ukuphatha.

Izinkomba

1. IKhomishini YaseYurophu - Kuchazwe ukuvikelwa kwedatha: Izimiso ze-GDPR - commission.europa.eu

2. Ihhovisi likaKhomishana Wolwazi (i-ICO) - Umkhawulo Wesitoreji - ico.org.uk

3. IKhomishini YaseYurophu - Idatha ingagcinwa isikhathi esingakanani futhi ingabe kuyadingeka ukuyibuyekeza? - commission.europa.eu

4. Isikhungo Sikazwelonke Sezindinganiso Nobuchwepheshe (i-NIST) - Uhlaka Lobumfihlo - nist.gov

5. Isikhungo Sezinsiza Zokuphepha Kwamakhompyutha se-NIST (CSRC) - SP 800-92: Umhlahlandlela Wokuphathwa Kwelogi Yokuphepha Kwamakhompyutha - csrc.nist.gov

6. Isikhungo Sokuphepha Kwe-inthanethi (i-CIS) - Ukuphathwa Kwelogi Yokuhlola (Ukulawulwa Kwe-CIS) - cisecurity.org

7. Imibhalo ye-Snowflake - Izinqubomgomo zokufinyelela emigqeni - docs.snowflake.com

8. Idokhumenti ye-Google Cloud - Ukuphepha kwezinga lemigqa ye-BigQuery - docs.cloud.google.com

9. BITOL - I- Open Data Contract Standard (ODCS) v3.1.0 - bitol-io.github.io

10. I-BITOL (GitHub) - I-Open Data Contract Standard - github.com

11. Ukugeleza Komoya kwe-Apache - Imibhalo (ezinzile) - airflow.apache.org

12. Ukugeleza Komoya kwe-Apache - ama-DAG (imiqondo eyinhloko) - airflow.apache.org

13. Imibhalo YamaLabhu e-dbt - Kuyini i-dbt? - docs.getdbt.com

14. Imibhalo YamaLabhu e-dbt - Mayelana namamodeli e-dbt - docs.getdbt.com

15. Imibhalo YamaLabhu e-dbt - Imibhalo - docs.getdbt.com

16. Imibhalo YamaLabhu e-dbt - Ukuhlolwa kwedatha - docs.getdbt.com

17. Imibhalo Yamalebhu e-dbt - Isendlalelo Semantic se-dbt - docs.getdbt.com

18. Imibhalo ye-Fivetran - Ukuqalisa - fivetran.com

19. I-Fivetran - Izixhumi - fivetran.com

20. Imibhalo ye-AWS - Umhlahlandlela Wonjiniyela we-AWS Lambda - docs.aws.amazon.com

21. I-GitHub - I-GitHub Copilot - github.com

22. Amadokhumenti e-GitHub - Ukuthola iziphakamiso zekhodi ku-IDE yakho nge-GitHub Copilot - docs.github.com

23. I-Microsoft Learn - GitHub Copilot ye-SQL (isandiso sekhodi ye-VS) - learn.microsoft.com

24. Imibhalo ye-Dynatrace - Ukubonwa kwedatha - docs.dynatrace.com

25. I-DataGalaxy - Kuyini ukubonwa kwedatha? - datagalaxy.com

26. Imibhalo Yokulindela Okukhulu - Ukubuka konke Kokulindela - docs.greatexpectations.io

Thola i-AI Yakamuva Esitolo Esisemthethweni Somsizi we-AI

Mayelana NATHI