Idatha yokwenziwa ikhulisa kanjani i-AI esikalini

I-Artificial intelligence (AI) ithembele kakhulu kumadathasethi amakhulu, ahlukahlukene futhi anamalebula ngokucophelela ukuze aqeqeshe ukufunda ngomshini (ML) ama-algorithms. Esikhathini samanje, idatha isiphenduke impilo ye-AI, futhi ukuthola idatha efanele kuthathwa njengento ebaluleke kakhulu neyinselele yokuthuthukisa amasistimu e-AI aqinile.

Nokho, ukuqoqa nokulebula amasethi edatha amakhulu anezigidi zezinto ezithathwe emhlabeni wangempela kudla isikhathi futhi kuyabiza. Ngenxa yalokho, lawo mamodeli e-ML okuqeqesha aseqalile ukuthembela kakhulu kuwo idatha yokwenziwanoma idatha ekhiqizwa ngokungeyikho esikhundleni sokukhiqizwa izehlakalo zomhlaba wangempela.

Idatha yokwenziwa inyuke kakhulu ekudumeni eminyakeni yamuva nje, yethula isisombululo esisebenzayo enkingeni yekhwalithi yedatha futhi inikeza amandla okubumba kabusha ukuthunyelwa kwe-ML enkulu. Ngokusho kuka-a Ukufunda kukaGartneridatha yokwenziwa kulindeleke ukuthi ihlanganise u-60% wayo yonke idatha esetshenziswa ekuthuthukisweni kwe-AI ngo-2024.

I-Turbocharging AI/ML enedatha yokwenziwa

Umqondo ulula ngokumangalisayo. Ivumela ongoti ukuthi bakhiqize idatha abayidingayo ngedijithali, ngokufunwa, nanganoma iyiphi ivolumu abayifunayo, eklanyelwe ukucaciswa kwabo okunembayo. Abacwaningi sebengakwazi manje ukuphendukela kumadathasethi okwenziwa adalwe kusetshenziswa amamodeli e-3D ezigcawu, izinto kanye nabantu ukuze bakhiqize iziqeshana zesenzo ngokushesha – ngaphandle kokuhlangabezana nezinkinga ze-copyright noma ukukhathazeka kwezimiso zokuziphatha ezihlobene nedatha yangempela.

“Ukusebenzisa idatha yokwenziwa yokuqeqeshwa komshini wokufunda kuvumela izinkampani ukuthi zakhe amamodeli ezimo ebezingafinyeleleki ngaphambilini ngenxa yokuthi idatha edingekayo iyimfihlo, ikhwalithi ephansi kakhulu noma ingekho nhlobo,” umhlaziyi weForrester uRowan Curran etshela i-VentureBeat. “Ukwenza ama-synthetic datasets kusebenzisa amasu afana nala amanethiwekhi ezitha ezikhiqizayo (ama-GAN) ukuthatha idathasethi yezinkulungwane ezimbalwa zabantu futhi bayiguqule ibe yidathasethi eyenza okufanayo lapho iqeqesha imodeli ye-ML – kodwa engenalo noma yiluphi ulwazi olungamhlonza umuntu (PII) lwedathasethi yoqobo.”

Abasekeli bakhomba izinzuzo ezihlukahlukene ekukhetheni amasethi edatha okwenziwa. Okokuqala, ukusebenzisa idatha yokwenziwa kunganciphisa kakhulu izindleko zokukhiqiza idatha yokuqeqeshwa. Ingase futhi ibhekane nezinkinga zobumfihlo ezihlobene nedatha engase ibe bucayi etholwe emhlabeni wangempela.

Idatha yokwenziwa ingasiza ekunciphiseni ukuchema, uma kuqhathaniswa nedatha yangempela, okungenzeka ingameli ngokunembile uhla olugcwele lolwazi olumayelana nomhlaba wangempela. Ukuhlukahluka okukhulu kungase kubalwe kumasethi edatha okwenziwa ngokuhlanganisa izimo ezingandile ezimelela amathuba angokoqobo kodwa okunzima ukuzithola kudatha yangempela.

U-Curran uchaze ukuthi amasethi edatha okwenziwa asetshenziselwa ukudala idatha yamamodeli ezimeni lapho idatha edingekayo ingekho khona ngenxa yokuthi isimo sokuqoqwa kwedatha senzeka njalo.

“Umhlinzeki wezokunakekelwa kwempilo wayefuna ukwenza umsebenzi ongcono wokubamba umdlavuza wamaphaphu usaqala, kodwa imininingwane encane yezithombe ibitholakala. Ngakho-ke ukuze bakhe imodeli yabo, badale idathasethi yokwenziwa eyayisebenzisa izithombe zamaphaphu ezinempilo ezihlanganiswe nezimila ezisanda kuqala ukuze kwakhiwe idathasethi entsha yokuqeqeshwa ezosebenza njengokungathi idatha efanayo eqoqwe emhlabeni wangempela,” kusho uCurran.

Uthe idatha yokwenziwa iphinde ithole amandla kwezinye izimboni ezivikelekile, njengezinsizakalo zezezimali. Lezi zinkampani zinemikhawulo ebalulekile yokuthi zingasebenzisa futhi zihambise kanjani idatha yazo, ikakhulukazi emafini.

Idatha yokwenziwa inamandla okuthuthukisa ukuthuthukiswa kwesofthiwe, ukusheshisa ucwaningo nokuthuthukiswa, ukusiza ukuqeqeshwa kwamamodeli e-ML, inike amandla izinhlangano ukuthi zithole ukuqonda okujulile kwedatha yazo yangaphakathi nemikhiqizo, futhi zithuthukise izinqubo zebhizinisi. Lezi zinzuzo, nazo, zingakhuthaza ukukhula kwe-AI ngezinga elikhulu.

Isebenza kanjani emhlabeni wangempela we-AI?

Kodwa umbuzo usasele: Ingabe idatha ekhiqizwe ngokuzenzisa ingasebenza njengedatha yangempela? Isebenza kahle kangakanani imodeli eqeqeshwe ngedatha yokwenziwa lapho ihlukanisa izenzo zangempela?

U-Yashar Behzadi, i-CEO kanye nomsunguli wesiteji sedatha yokwenziwa Ukuhlanganiswa kwe-AIithi izinkampani zivame ukusebenzisa idatha yokwenziwa neyomhlaba wangempela ngokuhlanganyela, ukuze ziqeqeshe amamodeli azo futhi ziqinisekise ukuthi zilungiselelwe ukusebenza kahle kakhulu.

“Idatha yokwenziwa ivamise ukusetshenziselwa ukwandisa nokwelula idatha yomhlaba wangempela, kuqinisekiswe amamodeli aqinile futhi enza kahle,” etshela i-VentureBeat. Isibonelo, uthe i-Synthesis AI isebenza nedlanzana labakhiqizi bezimoto besigaba 1 nezinkampani zesoftware.

“Silokhu sizwa ukuthi idatha yokuqeqeshwa etholakalayo iphansi kakhulu noma ayanele – futhi abanayo imvume yamakhasimende abo yokuqeqesha amamodeli abona ngekhompyutha ngayo noma kanjani,” esho. “Idatha yokwenziwa ixazulula zontathu izinselele – ikhwalithi, ubuningi nobumfihlo.”

Izinkampani ziphinde ziphendukele kudatha yokwenziwa lapho zingakwazi ukuthola izichasiselo ezithile kumalebula abantu, njengamamephu ajulile, okujwayelekile kwendawo, izimpawu zendawo ze-3D, amamephu okuhlukaniswa anemininingwane kanye nezakhiwo ezibonakalayo, echaza.

“I-Bias kumamodeli we-AI ibhalwe kahle, futhi ihlobene nedatha yokuqeqeshwa engaphelele engenakho ukuhlukahluka okudingekayo okuhlobene nobuhlanga, ithoni yesikhumba noma ezinye izibalo zabantu,” esho. “Ngenxa yalokhu, ukuchema kwe-AI kuba nomthelela ongemuhle kwezibalo zabantu futhi kuholela ekufakweni okuncane kwezinhlelo zokusebenza nemikhiqizo.” Ngokusebenzisa idatha yokwenziwa, eqhubeka, izinkampani zingachaza ngokucacile isethi yedatha yokuqeqeshwa ukuze kuncishiswe ukuchema futhi kuqinisekiswe amamodeli ahlanganisayo, agxile kubantu ngaphandle kokwephula ubumfihlo bomthengi.

Ukushintsha ngisho nengxenye encane yedatha yokuqeqeshwa yomhlaba wangempela ngedatha yokwenziwa kwenza kube nokwenzeka ukusheshisa nokwenza lula ukuqeqeshwa nokusetshenziswa kwamamodeli e-AI azo zonke izikali.

Kwa-IBM, ngokwesibonelo, abacwaningi basebenzise i- I-ThreeDWorld simulator nokuhambisana kwayo I-Task2Sim inkundla yokukhiqiza izithombe ezilingisayo zezigcawu nezinto ezingokoqobo, ezingasetshenziswa ukuqeqesha abahlukanisi bezithombe kusengaphambili. Lezi zithombe zokwenziwa zinciphisa inani ledatha yokuqeqeshwa kwangempela edingekayo, futhi zitholwe zisebenza ngokulinganayo kumamodeli wokuqeqesha kusengaphambili imisebenzi efana nokuthola umdlavuza kumaskeni wezokwelapha.

Ngaphezu kwalokho, ukugcwalisa idatha eyiqiniso ngedatha ekhiqizwe ngokuzenzisa kunganciphisa ubungozi bemodeli eqeqeshwe kusengaphambili ngedatha eluhlaza ekhishwe ku-inthanethi ebonisa ukucwasa noma ukucwasa ngokobulili. Idatha yokwenziwa ngokwezifiso ihlolwa kusengaphambili ukuze kuncishiswe ukuba khona kokuchema, kuncishiswe ubungozi bokuziphatha okungafuneki okunjalo kumamodeli.

UDavid Cox, umqondisi we-MIT-IBM Watson AI Lab, uDavid Cox, uthe: ucwaningo lwe-AI lokuhlola.

Idatha yokwenziwa kanye nekhwalithi yemodeli

U-Alp Kucukelbir, umsunguli kanye nososayensi omkhulu weplathifomu yokwenza imboni I-Fero Labs kanye noprofesa we-adjunct at Columbia Universityyathi nakuba idatha yokwenziwa ingahambisana nedatha yomhlaba wangempela yokuqeqeshwa kwamamodeli e-AI, iza ne-caveat enkulu: Udinga ukwazi ukuthi iliphi igebe olixhumayo kudathasethi yakho yomhlaba wangempela.

“Ake uthi usebenzisa i-AI ukuze ukhiphe ikhabhoni umshini wokugaya insimbi. Ufuna ukusebenzisa i-AI ukuze uvule futhi udalule ukusebenza okuqondile kwalelo gall (isb, ngokuqondile ukuthi imishini efekthri ethile isebenza kanjani ndawonye) futhi ungatholi kabusha i-metallurgy eyisisekelo ongayithola encwadini yokufunda. Kulokhu, ukusebenzisa idatha yokwenziwa, kuzodingeka ulingise ukusebenza okunembile kwesigayo sensimbi ngaphezu kolwazi lwethu lwe-metalurgy yezincwadi zokufunda,” kuchaza uKucukelbir. “Ukube ubune simulator enjalo, ubungeke udinge i-AI ukuqala ngayo.”

Ukufunda ngomshini kuhle ekuhlanganiseni, kodwa kungase kube ngcono ekuhlukaniseni amadathasethi okuqeqeshwa. Kodwa-ke, idatha ekhiqizwe ngokungemthetho ivumela abacwaningi kanye nodokotela ukuthi banikeze idatha “yecala-ekhoneni” ku-algorithm, futhi ekugcineni ingasheshisa imizamo ye-R&D, kwengeza uJulian Sanchez, umqondisi wobuchwepheshe obusafufusa kwa- John Deere.

“Sizamile idatha yokwenziwa ngendlela yokuhlola e-John Deere, futhi kukhombisa isithembiso. Isethi evamile yezibonelo ibandakanya ezolimo, lapho kungenzeka ube nezinga eliphansi kakhulu lokwenzeka kwamacala athile akhona,” kusho uSanchez etshela i-VentureBeat. “Idatha yokwenziwa ihlinzeka ngama-algorithms e-AI/ML ngamaphoyinti ereferensi adingekayo ngedatha futhi inika abacwaningi ithuba lokuqonda ukuthi abaqeqeshiwe [model] ikwazi ukuphatha amacala ahlukene okusetshenziswa. Kuzoba yingxenye ebalulekile yokuthi i-AI/ML ikala kanjani. ”

Ngokufanayo, u-Sebastian Thrun, owayeyi-VP ye-Google kanye nosihlalo wamanje kanye nomsunguli wenkundla yokufunda ye-inthanethi Ukungaziphathi kahle, ithi lolu hlobo lwedatha ngokuvamile alunangqondo ngokuhambisana nobukhulu obuthile. Ukulingisa ngedatha yokwenziwa kuyindlela esheshayo nephephile yokusheshisa ukufunda, kodwa ngokuvamile kunamaphutha aziwayo.

“Lokhu kuyiqiniso ikakhulukazi ngedatha ekubonweni kwayo (izithombe zekhamera, inkulumo, njll.). Kodwa isu elilungile livamise ukuhlanganisa idatha yomhlaba wangempela nedatha yokwenziwa,” kusho uThrun etshela iVentureBeat. “Ngesikhathi ngisephrojekthi ye-Google ezishayelayo Waymo, sisebenzise inhlanganisela yakho kokubili. Idatha yokwenziwa izodlala indima enkulu ezimeni esingafuni ukuzithola emhlabeni wangempela. “

Izinselelo zokusebenzisa idatha yokwenziwa ye-AI

UMichael Rinehart, i-VP ye-AI e- amafu amaningi inkundla yokuphepha kwedatha Securiti AIithi kukhona ukuhwebelana phakathi kokusebenziseka kwedatha yokwenziwa kanye nobumfihlo obunikezayo.

“Ukuthola i-tradeoff efanele kuyinselele ngoba kuncike enkampanini, njenganoma yikuphi ukuhlolwa komvuzo wobungozi,” kusho uRinehart. “Le nselele iphinde yengezwe yiqiniso lokuthi izilinganiso zobuningi bobumfihlo aziphelele, futhi ubumfihlo obuningi empeleni bungakwazi ukutholwa yidathasethi yokwenziwa kunalokho okushiwo yisilinganiso.”

Wachaza ukuthi ngenxa yalokho, ukulawula okuxekethile noma izinqubo kungase kusetshenziswe kulolu hlobo lwedatha. Isibonelo, izinkampani zingase zeqe amafayela edatha yokwenziwa aziwayo ngesikhathi sokuskenwa kwedatha ebucayi, alahlekelwe ukubonakala ekwandeni kwawo. Isayensi yedatha amaqembu angase aqeqeshe ngisho namamodeli amakhulu kuwo, akwazi ukubamba ngekhanda futhi akhiqize kabusha idatha yokwenziwa, bese eyisabalalisa.

“Uma idatha yokwenziwa noma okunye kokunye okuphuma kuyo kuhloswe ukuba kwabelwane ngakho noma kudalulwe, izinkampani kufanele ziqinisekise ukuthi ivikela ubumfihlo banoma yimaphi amakhasimende eziwamele, ngokwesibonelo, ngokusebenzisa ubumfihlo obuhlukile ngayo,” kweluleka u-Rinehart. “Idatha yokwenziwa eyimfihlo esezingeni eliphakeme iqinisekisa ukuthi amaqembu angakwazi ukwenza izivivinyo ngedatha engokoqobo engadaluli ulwazi olubucayi.”

UFernando Lucini, ohola umhlaba wonke kwisayensi yedatha kanye nobunjiniyela bokufunda komshini ku I-Accentureyengeza ngokuthi ukukhiqiza idatha yokwenziwa kuyinqubo eyinkimbinkimbi kakhulu, edinga abantu abanamakhono akhethekile nolwazi oluthuthuke ngempela lwe-AI.

“Inkampani idinga izinhlaka namamethrikhi aqondile futhi ayinkimbinkimbi ukuze iqinisekise ukuthi idale lokho ebikuhlosile,” echaza.

Yini elandelayo ngedatha yokwenziwa ku-AI?

U-Lucini ukholelwa ukuthi idatha yokwenziwa iyisibusiso kubacwaningi futhi maduzane izoba yithuluzi elivamile kuzo zonke izitaki zobuchwepheshe zenhlangano ukuze kukhuliswe amandla azo amamodeli e-AI/ML.

“Ukusebenzisa idatha yokwenziwa akunikezeli nje kuphela ithuba lokusebenza ezinkingeni ezithakazelisayo kakhulu kubacwaningi nokusheshisa izisombululo, kodwa futhi kunethuba lokuthuthukisa ama-algorithms aqanjwe kakhulu angase avule amacala amasha okusebenzisa ebesingakaze sicabange ukuthi angenzeka,” kwengeza uLucini. “Ngilindele ukuthi idatha yokwenziwa ibe yingxenye yakho konke ukufunda komshini, i-AI kanye nokuhamba komsebenzi kwesayensi yedatha futhi ngaleyo ndlela kube nesixazululo sedatha yanoma iyiphi inkampani.”

Ngokwengxenye yakhe, i-Synthesis AI’s Behzadi ibikezela ukuthi i-AI ekhiqizayo i-boom ibilokhu futhi izoqhubeka nokuba yi-catalyst enkulu yedatha yokwenziwa.

>>Landela okuqhubekayo kwe-VentureBeat i-AI ekhiqizayo ukufakwa <<

“Kube nokukhula okukhulu kulezi zinyanga ezimbalwa ezedlule, futhi ukubhanqa i-AI ekhiqizayo nedatha yokwenziwa kuzoqhubeka nokutholwa,” esho.

Ukuhlanganisa i-AI ekhiqizayo namapayipi wemiphumela yokubuka, ukuhlukahluka kanye nekhwalithi yedatha yokwenziwa kuzothuthuka kakhulu, esho. “Lokhu kuzothuthukisa ukwamukelwa ngokushesha kwedatha yokwenziwa kuzo zonke izimboni. Eminyakeni ezayo, wonke amaqembu ombono wekhompyutha azosebenzisa idatha yokwenziwa. ”