Ukusebenzisa i-GPT-n Ukuze Uthuthukise Inqubo Ye-Qlik Ethuthukisiwe

Njengoba kungenzeka uyazi, ithimba lami kanye nami silethe emphakathini wakwa-Qlik isandiso sesiphequluli esihlanganisa i-Qlik ne-Git ukuze ulondoloze izinguqulo zedeshibhodi kalula, senza izithonjana zamadeshibhodi ngaphandle kokushintshela kwamanye amawindi. Ngokwenza kanjalo, silondoloza onjiniyela be-Qlik isikhathi esibalulekile futhi sinciphisa ingcindezi nsuku zonke.

Ngihlala ngibheka izindlela zokuthuthukisa inqubo ye-Qlik futhi ngithuthukise izindlela zansuku zonke. Kungakho kunzima kakhulu ukugwema isihloko esishube kakhulu, i-ChatGPT, ne-GPT-n, nge-OpenAI noma i-Large Language Model efanayo.

Ake seqe ingxenye yokuthi Amamodeli Olimi Amakhulu, i-GPT-n, asebenza kanjani. Kunalokho, ungabuza i-ChatGPT noma ufunde incazelo engcono kakhulu yomuntu kaSteven Wolfram.

Ngizoqala ngethisisi engathandwayo, “I-GPT-n Generated Insights from the data is a Curiosity-Quenching Toy,” bese ngabelana ngezibonelo zangempela lapho umsizi we-AI esisebenza kuye ekwazi ukuzenzela imisebenzi evamile, isikhathi samahhala sokuthola izinto eziyinkimbinkimbi. ukuhlaziya nokwenza izinqumo kwabathuthukisi/abahlaziyi be-BI.

Umsizi we-AI kusukela ebuntwaneni bami

Ungavumeli i-GPT-n Ikukhohlise

… isho nje izinto “ezizwakala kahle” ezisuselwe ekutheni izinto “zazizwakala kanjani” ezintweni zayo zokuqeqesha. © USteven Wolfram

Ngakho, uxoxa ne-ChatGPT usuku lonke. Futhi kungazelelwe, kufike umqondo okhaliphile engqondweni: "Ngizokwazisa i-ChatGPT ukuthi ikhiqize imininingwane ebambekayo kusuka kudatha!"

Ukondla amamodeli e-GPT-n kusetshenziswa i-OpenAI API ngayo yonke idatha yebhizinisi namamodeli edatha yisilingo esikhulu sokuthola ukuqonda okubambekayo, kodwa nansi into ebalulekile — umsebenzi oyinhloko Wemodeli Yolimi Olukhulu njenge-GPT-3 noma ngaphezulu ukuthola ukuthi kanjani ukuqhubeka nocezu lombhalo onikiwe. Ngamanye amazwi, “Ilandela iphethini” yalokho okukhona kuwebhu nasezincwadini nezinye izinto ezisetshenziswa kuyo.

Ngokusekelwe kuleli qiniso, kunezimpikiswano eziyisithupha eziphusile zokuthi kungani imininingwane ekhiqizwe i-GPT-n iyithoyizi nje yokuqeda ilukuluku lakho kanye nomphakeli kaphethiloli wejeneretha yombono ebizwa ngokuthi ubuchopho bomuntu:

1. I-GPT-n, i-ChatGPT ingase ikhiqize imininingwane engafanele noma engasho lutho ngenxa yokuthi ayinawo umongo odingekayo ukuze iqonde idatha nama-nuances ayo—ukuntuleka komongo.
2. I-GPT-n, i-ChatGPT ingase ikhiqize imininingwane enganembile ngenxa yamaphutha ekucutshungulweni kwedatha noma ama-algorithms anephutha — ukuntuleka kokunemba.
3. Ukuthembela kuphela ku-GPT-n, i-ChatGPT ukuze uthole imininingwane ingaholela ekuntulekeni kokucabanga okujulile nokuhlaziya okuvela kochwepheshe abangabantu, okungase kuholele eziphethweni ezingalungile noma ezingaphelele — ukuthembela ngokweqile kokuzenzakalela.
4. I-GPT-n, i-ChatGPT ingase ikhiqize imininingwane echemile ngenxa yedatha eqeqeshwe ngayo, okungase kuholele emiphumeleni eyingozi noma ebandlululayo — ingozi yokuchema.
5. I-GPT-n, i-ChatGPT ingase intule ukuqonda okujulile kwemigomo yebhizinisi nezinjongo eziqhuba ukuhlaziywa kwe-BI, okuholela ezincomweni ezingahambelani nesu lonke - ukuqonda okulinganiselwe kwemigomo yebhizinisi.
6. Ukwethemba idatha ebalulekile yebhizinisi nokwabelana ngayo "ngebhokisi elimnyama" elingakwazi ukuzifundela kuzoveza umbono ekuphathweni kwe-TOP amakhanda akhanyayo ofundisa izimbangi zakho ukuthi zingawina kanjani - ukuntula ukwethemba. Besesivele sikubonile lokhu lapho isizindalwazi sokuqala samafu njenge-Amazon DynamoDB siqala ukuvela.

Ukufakazela okungenani impikiswano eyodwa, ake sihlole ukuthi i-ChatGPT ingazwakala kanjani ikholisa. Kodwa kwezinye izimo, akulungile.

Ngizocela i-ChatGPT ukuthi ixazulule isibalo esilula 965 * 590 bese ngiyicela ukuthi ichaze imiphumela isinyathelo nesinyathelo.

568 350! OOPS... kukhona okungahambanga kahle.

Endabeni yami, i-hallucination yagqashuka empendulweni ye-ChatGPT ngoba impendulo engu-568,350 ayilungile.

Masenze ishothi yesibili futhi sicele i-ChatGPT ukuthi ichaze imiphumela isinyathelo nesinyathelo.

Isithombe esihle! Kodwa kusengalungile...

I-ChatGPT izama ukukholisa ngencazelo yesinyathelo nesinyathelo, kodwa ayikalungi.

Umongo ubalulekile. Ake sizame futhi kodwa sondle inkinga efanayo ngomyalo othi “yenza njenge …”.

I-BINGO! 569 350 impendulo efanele

Kodwa lokhu kuyisimo lapho uhlobo lokujwayezwa kwenethi ye-neural lungakwenza kalula - yini 965*590 - lungeke lwanele; i-algorithm yangempela yokubala iyadingeka, hhayi nje indlela esekelwe kwizibalo.

Kwazi bani… mhlawumbe i-AI isanda kuvumelana nothisha bezibalo esikhathini esidlule futhi ayisebenzisi isibali kuze kube amabanga aphezulu.

Njengoba ukwaziswa kwami ​​esibonelweni sangaphambilini kuqondile, ungakwazi ukuhlonza ngokushesha amanga empendulo evela ku-ChatGPT bese uzama ukukulungisa. Kodwa kuthiwani uma i-hallucination idlula ekuphenduleni imibuzo efana nalena:

1. Imuphi umthengisi osebenza kahle kakhulu?
2. Ngibonise Imali Engenayo yekota yokugcina.

Kungasiholela ekwenziweni kwezinqumo eziqhutshwa yi-HALLUCINATION, ngaphandle kwamakhowe.

Yebo, nginesiqiniseko sokuthi izimpikiswano zami eziningi ezingenhla zizobe zingasebenzi ezinyangeni ezimbalwa noma iminyaka ngenxa yokuthuthukiswa kwezixazululo ezigxile kancane emkhakheni we-Generative AI.

Nakuba ukulinganiselwa kwe-GPT-n kungafanele kushaywe indiva, amabhizinisi asengakwazi ukudala inqubo yokuhlaziya enamandla kakhulu futhi ephumelelayo ngokusebenzisa amandla abahlaziyi abangabantu (kuyahlekisa ukuthi kufanele ngigqamise u-HUMAN) nabasizi be-AI. Ngokwesibonelo, cabangela isimo lapho abahlaziyi abangabantu bezama ukuhlonza izici ezibangela ukuhilizisana kwamakhasimende. Esebenzisa abasizi be-AI abaxhaswe yi-GPT-3 noma ngaphezulu, umhlaziyi angakhiqiza ngokushesha uhlu lwezinto ezingaba khona, njengamanani entengo, isevisi yamakhasimende, kanye nekhwalithi yomkhiqizo, bese ehlola lezi ziphakamiso, aphenye idatha ngokuqhubekayo, futhi ekugcineni ahlonze izici ezifanele kakhulu. lokho kuxosha ikhasimende.

NGIKHOMBISE IMIBHALO EFANA NOMUNTU

UMHLAZIYI WABANTU wenza ukwaziswa ku-ChatGPT

Umsizi we-AI angasetshenziselwa ukwenza imisebenzi oyisebenzisa amahora amaningi uyenza manje. Kusobala, kodwa ake sibhekisise indawo lapho abasizi be-AI abanikezwa amandla Amamodeli Olimi Amakhulu njenge-GPT-3 nangaphezulu bavivinywa kahle khona — bakhiqiza imibhalo efana nomuntu.

Kunenqwaba yazo emisebenzini yansuku zonke yonjiniyela be-BI:

1. Amashadi okubhala, izihloko zeshidi, nezincazelo. I-GPT-3 nangaphezulu ingasisiza ukuthi sikhiqize ngokushesha izihloko ezifundisayo nezimfushane, siqinisekise ukuthi ukubonwa kwedatha yethu kulula ukukuqonda nokuzulazula kubenzi bezinqumo futhi sisebenzisa “ukwenza njengo..” ukwaziswa.
2. Amadokhumenti ekhodi. Nge-GPT-3 nangaphezulu, singakha ngokushesha amazwibela ekhodi abhalwe kahle, senze kube lula kumalungu eqembu lethu ukuthi aqonde futhi agcine i-codebase.
3. Ukudala izinto eziyinhloko (isichazamazwi sebhizinisi). Umsizi we-AI angasiza ekwakheni isichazamazwi sebhizinisi esiphelele ngokunikeza izincazelo ezinembayo nezimfishane zamaphoyinti edatha ahlukahlukene, ehlise ukungaqondakali, kanye nokukhuthaza ukuxhumana kweqembu okungcono.
4. Ukudala isithonjana esikhangayo (amakhava) samashidi/amadeshibhodi kuhlelo lokusebenza. I-GPT-n ingakhiqiza izithonjana ezihehayo nezibukekayo, ithuthukise ulwazi lomsebenzisi futhi ikhuthaze abasebenzisi ukuthi bahlole idatha etholakalayo.
5. Ukubhala amafomula okubala ngezinkulumo zokuhlaziya ezihleliwe kumibuzo ye-Qlik Sense / DAX ku-Power BI. I-GPT-n ingasisiza ukuthi sibhale lezi zinkulumo kanye nemibuzo ngendlela ephumelela kakhudlwana, sinciphise isikhathi esichithwa ekubhaleni amafomula futhi sisivumele ukuthi sigxile ekuhlaziyeni idatha.
6. Ukubhala imibhalo yokulayisha idatha (ETL). I-GPT-n ingasiza ekudaleni imibhalo ye-ETL, iguqule idatha ngokuzenzakalelayo, kanye nokuqinisekisa ukuvumelana kwedatha kuwo wonke amasistimu.
7. Ukuxazulula izinkinga zedatha nezinkinga zohlelo lokusebenza. I-GPT-n inganikeza iziphakamiso nemininingwane ukusiza ukuhlonza izinkinga ezingaba khona futhi inikeze izixazululo zedatha evamile nezinkinga zohlelo lokusebenza.
8. Ukuqamba kabusha izinkambu ukusuka kwezobuchwepheshe kuye kwezebhizinisi Kumodeli Yedatha. I-GPT-n ingasisiza sihumushele amagama obuchwepheshe olimini lwebhizinisi olufinyeleleka kalula, senze imodeli yedatha ibe lula ukuyiqonda kubabambiqhaza abangebona abezobuchwepheshe ngokuchofoza okumbalwa.

Abasizi be-AI abanikwa amandla amamodeli e-GPT-n bangasisiza ukuthi sisebenze kahle kakhulu futhi siphumelele emsebenzini wethu ngokwenza imisebenzi evamile kanye nesikhathi esikhululekile sokuhlaziya okuyinkimbinkimbi nokwenza izinqumo.

Futhi lena indawo lapho isandiso sesiphequluli sethu se-Qlik Sense singaletha khona inani. Silungiselele ukukhishwa okuzayo - komsizi we-AI, okuzoletha izihloko nokukhiqizwa kwencazelo kubathuthukisi be-Qlik ohlelweni lokusebenza kuyilapho kwakhiwa izinhlelo zokusebenza zezibalo.

Besebenzisa i-GPT-n ecushwe kahle nge-OpenAI API kule misebenzi evamile, abathuthukisi be-Qlik nabahlaziyi bangathuthukisa kakhulu ukusebenza kahle kwabo futhi banikeze isikhathi esengeziwe sokuhlaziya okuyinkimbinkimbi nokwenza izinqumo. Le ndlela iphinde iqinisekise ukuthi sisebenzisa amandla e-GPT-n kuyilapho sinciphisa ubungozi bokuthembela kuyo ukuze kuhlaziywe idatha ebalulekile nokukhiqizwa kwemininingwane.

Isiphetho

Sengiphetha, ngicela ungivumele ngivumele i-ChatGPT:

Ukubona kokubili imikhawulo kanye nokusetshenziswa okungenzeka kwe-GPT-n ngaphakathi komongo we-Qlik Sense namanye amathuluzi obuhlakani bebhizinisi kusiza izinhlangano zenze ngokugcwele lobu buchwepheshe be-AI obunamandla kuyilapho zinciphisa izingozi ezingaba khona. Ngokugqugquzela ukubambisana phakathi kwemininingwane ekhiqizwe yi-GPT-n-ubuchwepheshe bomuntu, izinhlangano zingakha inqubo yokuhlaziya eqinile esebenzisa amandla akho kokubili i-AI nabahlaziyi abangabantu.

Ukuze ube phakathi kwabokuqala ukuthola izinzuzo zokukhishwa komkhiqizo wethu ozayo, sithanda ukukumema ukuthi ugcwalise ifomu lohlelo lwethu lokufinyelela ngaphambi kwesikhathi. Ngokujoyina uhlelo, uzothola ukufinyelela okukhethekile ezicini zakamuva nezithuthukisi ezizokusiza ukuthi usebenzise amandla omsizi we-AI ekugelezeni komsebenzi wakho wokuthuthukisa i-Qlik. Ungaphuthelwa leli thuba lokuhlala uphambi kwejika futhi uvule amandla aphelele emininingwane eshayelwa yi-AI yenhlangano yakho.

Joyina Uhlelo Lwethu Lokufinyelela Ngaphambili