Kwa-HackerNoon, siyazazisa izindaba ezimayelana nokwakha ubuchwepheshe obudala. Le mibuzo ayigcini nje ngemishini ye-AI—imayelana nokuqamba okusha, izinselele, kanye nobuchule bokwenza amathuluzi aphambili ekuphileni, afanele umphakathi wethu wabaholi bezobuchwepheshe, abakhi, nabafundi abanomqondo wekusasa.
Igama lami ngingu- Aniruth . Ngisebenza eqenjini lesitoreji e -Databricks , lapho sisebenzela khona ukunika amandla amakhasimende ukuthi alondoloze inani elikhulu ledatha ngefomethi evulekile, engakala nge-Data Intelligence Platform. Ngokukhethekile, ngisebenza emizamweni yethu yokusebenzisana neDelta Lake kanye ne-Apache Iceberg.
I-Aniruth : I-Databricks ihlanganisa idatha ne-AI ukuze inikeze amakhasimende ubuhlakani obusebenzisekayo—lokho esikubiza ngokuthi ubuhlakani bedatha. Lokhu kufaka phakathi ukufaka idatha enkulu, i-ETL, isitoreji esikhulu, imibuzo yobuhlakani bebhizinisi, kanye nemithwalo yemisebenzi ye-AI. Amasu asetshenziswa emshinini wokufunda eminyakeni eyishumi edlule abekhona kusukela ngawo-1980; ukwanda kwedatha enkulu kwenze kwaba nokwenzeka ukusebenzisa ama-algorithm esikalini.
Amasu afana nokwaziswa kokudutshulwa okuncane noma i-RAG ancike kudatha yekhwalithi ephezulu. Amamodeli anedatha engcono awina uma kuqhathaniswa nalawo anezakhiwo ezingcono. I-Databricks ifake utshalomali olukhulu emizamweni ehamba phambili esikhaleni sedatha, iphayona ekwakhiweni kwe-lakehouse ngamafomethi edatha avulekile kanye nokuphatha okuvulekile, lapho amakhasimende ekwazi ukuthola imininingwane engcono kakhulu ngokusebenza okungcono kakhulu okuvela kumachibi edatha.
Sisebenzisa amamodeli e-AI ngezindlela eziningi emkhiqizweni - njenge-Llama 3 yomsizi we-AI. Sikholelwa kudatha evulekile kanye ne-AI ecosystem futhi sikhuthaza amakhasimende ethu ukuthi asebenzise noma iyiphi imodeli azikhethele yona. Sisiza ukwenza isiqiniseko sokuthi amakhasimende anokuphatha kokuphela kuze kube sekugcineni kuwo wonke umjikelezo wempilo we-AI kungakhathaliseki ukuthi amamodeli awasebenzisayo anjani, ukuze akwazi ukugxila ekwenzeni amamodeli awo akhelwe inhloso izimo zawo zokusebenzisa.
Sisebenzise umzamo omkhulu kanye nokutshalwa kwezimali ekubekeni phambili ukunemba nezimpendulo ezingachemile zokusetshenziswa kwe-AI phakathi kwemikhiqizo yethu, futhi siyaqhubeka nokuhlola njalo.
Idatha nesikhala se-AI sishintsha ngokushesha, ngakho-ke kubaluleke kakhulu ukugcina usesikhathini samanje. Usuku lwami lungahlanganisa ukukhuluma namakhasimende, ukuhlaziya imakethe, ukuhlanganisa idokhumenti yezidingo zomkhiqizo, noma ukulungiselela izinto zokuthengisa. Ingxenye engiyithandayo ukwenza imidwebo ekhombisa ukuthi izinto zizosebenza kanjani, njengoba kumnandi kakhulu ukuguqula umbono ube obonakalayo.
Kuningi ukuphumelela okukhulu okuzayo maduze. Okunye engikuthakaselayo ikakhulukazi ukwenziwa komuntu siqu kokuqukethwe. Kule minyaka eyishumi edlule, izikhangiso zicushwe kahle kumbheki othile. Ezinye izici zokuqukethwe zishuniwe, njengalokho isithonjana i-Netflix esibonisa umsebenzisi, kodwa okuqukethwe kwangempela (ividiyo ngokwayo) ayikalungiswa. Kuzokujabulisa ukubona ukuthi abaqondisi/abakhiqizi babhalansisa kanjani indaba abayifunayo kanye nezintshisekelo zabasebenzisi.
Ngisebenza ekugcinweni kwedatha enkulu, okungadida kakhulu ukuyiqonda. Sinokulungiselelwa okuhlukahlukene kwe-AI kudatha, kodwa kuvame ukuba nemibuzo mayelana nokuthi lokhu kulungiselelwa kwenziwa nini, ukuthi kusebenza kanjani, lokho okungakubandakanyi, njll. Ngalezi zinhlobo zemibuzo, kubalulekile ukuqinisekisa ukuthi sinemiyalezo ecacile nengaguquki mayelana sakha ini nokuthi kungani. Ngithole ukuthi ukuchaza imbangela yokulinganiselwa kuthinta kahle kakhulu amakhasimende.
Amamodeli e-Multimodal azoba ngcono kakhulu eminyakeni ezayo, okuzoshintsha indlela yethu eyinhloko yokusebenzisana ne-AI. Ukuthola imizwa yomuntu kulula kakhulu kusuka olwazini olubonakalayo noma olulalelwayo uma kuqhathaniswa nombhalo. Ngicabanga ukuthi kunethuba lokudala ukusebenzisana kwemvelo okwengeziwe kuchungechunge olubanzi lwezimo.
Ngokuvamile sifuna ukubona impendulo nokusetshenziswa okuhle. Ngikhuluma namakhasimende kaningi ukuze ngithole umuzwa wokuthi kungani futhi ecabanga kanjani ngemikhiqizo yethu, okuwukhiye wokuchaza ukuthi kungani sibona amathrendi athile kumamethrikhi.
Imikhiqizo yedatha emikhulu idume ngenselele enkulu ukuyisebenzisa. Izibonelo ezilula kulula ukuzisetha, kodwa umsebenzi omningi wokukhiqiza ngokuvamile ubandakanya ukucushwa okudidayo nekhodi. Kube yinto ehamba phambili kimina ukwakha ukusebenza okudingwa ngamakhasimende, kuyilapho ngenza umkhiqizo ube lula kakhulu ukuwusebenzisa.
Ikusasa yilapho noma yiliphi ibhizinisi lithola imininingwane kudatha yalo kalula. Emhlabeni wamanje, imininingwane yebhizinisi eqhutshwa idatha ivamise ukukhawulelwa ezinkampanini ezinkulu - kodwa nazo zingancamela ukuzizwisa okulula.
Kumuntu ngamunye, ngijabule kakhulu ngokuhlanganiswa kwe-AI ku-hardware. Kuze kube manje, sesiyibonile kakhulu i-AI ezinhlelweni zesoftware ezifana namawebhusayithi. Kunezinhlelo zokusebenza eziningi ezinkulu zemishini yokwakha esebenzisa i-AI, futhi sesivele siqala ukubona eminye yaleyo mithelela ezimotweni nasemafonini.
I-Databricks isendleleni ebheke ekubeni ibe lula futhi ibe namandla kakhulu ngesikhathi esifanayo. Miningi imizamo esisebenza kuyo yonke indawo, kusukela ekwenzeni idatha yezinga elikhulu nokubala kalula ukusebenza ngayo kuya ekuthuthukiseni ukusebenza kwemibuzo nokugeleza komsebenzi. Ngokwami, ngicabanga ukuthi sinezici ezijabulisayo ezizayo maduze kuwo wonke umkhiqizo ezenza ukuhamba komsebenzi kube lula nge-AI. Izibonelo zifaka amazwana akhiqizwe yi-AI kudatha, iziphakamiso zekhodi ye-AI kubahleli bamanothibhuku, kanye nezokuxhumana ze-AI ukuze uxoxe nedatha (isibonelo, i-Databricks AI/BI Genie).
Kunokukhathazeka ngokuthi i-AI izokwehlisa yini inani lemisebenzi. Imikhiqizo yethu yakhelwe ukwandisa ulwazi olubalulekile, oluvame ukuza ngokuhlangana nabasebenzisi. Isibonelo, nge-AI/BI Genie, abasebenzisi bangakha izixhumanisi kudatha yabo. Lona okuhlangenwe nakho komlingo, lapho abasebenzisi bengabuza khona imibuzo futhi bathole izimpendulo eziqondile kubo. Eqinisweni, abasebenzisi bangabheka i-SQL esetshenziswayo ukuze baqinisekise ukuthi yilokho abakufunayo. Lokhu kubanjiswana nabahlaziyi, kunciphisa isikhathi esibathathayo ukusuka emcabangweni baye ekuqondeni.
Okunye okwangimangaza kakhulu wubunzima obukhona kwezinye izinkampani ezinkulu. Lokhu kwethula izidingo emkhiqizweni ebengingeke ngizicabangele ngokwami. Isibonelo esivamile ukucabanga ngamasu okufuduka lapho wethula umkhiqizo omusha. Ngokuvamile, izinkampani ezinkulu zizobe zihlanganise ubuchwepheshe obukhona (imvamisa isofthiwe yomthombo ovulekile) noma zakhe isofthiwe yangokwezifiso ukuxazulula inkinga ebhekana nomkhiqizo wethu omusha. Ngokuvamile kuthatha isikhathi ukuqonda ukuthi kungani futhi kanjani lokhu kuhlanganiswa ukuze kuqinisekiswe ukuthi sinesixazululo esihlanganisa wonke amathuba.
Ungathanda ukugxila ekuphenduleni eminye yale mibuzo?