Kulo dliwano-ndlebe olukhethekileyo lwe-HackerNoon "Ngemva kokuQala", sihlala phantsi kunye no-Shashank Yadav, umsunguli kunye ne-CEO ye-   , iqonga elixhobisa abasebenzisi ukuba baqeqeshe kwaye babe neemodeli zabo ze-AI. Ngemvelaphi ye-AI kunye namava kumaqela okufunda ngoomatshini angundoqo e-Goldman Sachs kunye ne-Microsoft, u-Shashank wabelana ngohambo lwakhe, ulwazi malunga nemiceli mngeni yokulinganisa i-AI, kunye nendlela i-Fraction AI ejongene ngayo neyona nto inzima kakhulu kwishishini - idatha ethembekileyo, ekumgangatho ophezulu. Fumana ujongo lwangaphakathi kwindlela iFraction AI ephazamisa ngayo i-AI landscape kunye nedemokhrasi yobunini be-AI. Fraction AI   U-Ishan Pandey: Molo Shashank, kuluvuyo ukukwamkela kuthotho lwethu "Emva kokuQalisa". Nceda usixelele malunga nawe kwaye yintoni ekuphefumlele ukuba ufumane i-Fraction AI?    : Heyi Ishan, kuhle ukuba lapha. NdinguShashank, umseki weFraction AI. Imvelaphi yam iku-AI. Ndifunde isayensi yekhompyuter kwi-IIT Delhi ngokugxila kuphando lwe-AI. Emva koko, ndasebenza kwiqela eliphambili le-ML e-Goldman Sachs, emva koko ndajoyina isiqalo sokuqala njengomphandi we-AI, kwaye kamva ndafudukela kwi-hedge fund esebenzisa i-AI kurhwebo lobungakanani. Shashank Yadav  Ingxaki endandihlala ndibalekela kuyo yayikukuba i-AI yayisiba sembindini. Iinkampani ezimbalwa zazilawula eyona modeli inamandla, kwaye wonke umntu wayebambekile esebenzisa iinguqulelo ezingaphandle kweshelufu ezazingalungiselelwanga iimfuno zabo. Kodwa i-AI ayilingani-zonke. Igqwetha lifuna imodeli eyahlukileyo kunomrhwebi okanye umphuhlisi. Eyona AI ibalaseleyo ikhethekileyo, ukanti ukuziqeqesha kwakho kwakuxabisa kakhulu okanye kunzima kakhulu.  Yiyo loo nto ndiqalise i-Fraction AI. Liqonga apho nabani na anokuba ngabanini kwaye aqeqeshe imodeli yakhe ye-AI. Abasebenzisi benza iiarhente ze-AI ezikhuphisanayo kwiiseshoni. I-arhente nganye ihlawula umrhumo omncinci wokungena, yenza eyona mveliso ingcono kakhulu yomsebenzi, kwaye igwetywa yiLLM. Abaphumeleleyo bafumana umvuzo, kwaye iimodeli zabo ziyaphucuka ngokusekelwe kwiziphumo zabo ezilungileyo. Ngokuhamba kwexesha, abasebenzisi bakha ii-AI ezikhethekileyo ezihlala zibangcono.  Endaweni yokuxhomekeka kwiimodeli ezinkulu ezimbalwa, sidala inkqubo yendalo apho amawaka eemodeli ezincinci, ezikhethekileyo zikhuphisana, zifunde kwaye zikhule. I-AI ayifanele ibe yinto oyisebenzisayo. Kufuneka ibe yinto onayo kwaye uphucule. Nantso into esiyakhayo.  U-Ishan Pandey: Usebenze kumaqela aphambili eML eMicrosoft kunye neGoldman Sachs. La mava ayenze njani indlela yakho yokwakha iQela le-AI?    : Ewe, ngexesha lekholeji, ndangena kwi-Microsoft kwiqela le-Bing, ndisebenza kumatshini wokufunda ukukhangela isikhundla. Olo yayiluphawu lwam lokuqala lokwenyani kwiinkqubo ezinkulu ze-AI. Ukukhangela akukhona nje malunga nokufumana ulwazi, kodwa ukuqonda ukuba abasebenzisi bafuna ntoni ngokwenene kunye nokulinganisa iziphumo ngokufanelekileyo. Indifundise ukuba i-AI ayipheleli nje ngeemodeli ezikrelekrele, imalunga nokubenza basebenze kwihlabathi lokwenyani. Shashank Yadav  E-Goldman Sachs, ndandikwiqela eliphambili le-ML, ukwakha imizekelo yokuqikelelwa kwemali. Kwezemali, kwanophuculo oluncinci lubalulekile, kwaye iimodeli zihlala zivavanywa kwiimeko zehlabathi lokwenyani apho iimpazamo zibiza kakhulu. Loo mava andifundise indlela yokwakha i-AI ethembekileyo, eguquguqukayo, kunye nokuphucula ixesha elide kunokuba iqhube kakuhle kwindawo elawulwayo.  Kamva, kwi-hedge fund, ndasebenza kwi-AI yorhwebo lwe-quant. Kulapho ndabona khona indlela olunamandla ngayo ukhuphiswano. Iimodeli ezihlala zilungelelaniswa kwaye zifunde kubuchule obukhuphisanayo zikholisa ukwenza ngcono kunezo zihlala zingatshintshi.  Yonke loo nto imile iQela le-AI. Esikhundleni sokwakha i-AI enye egqibeleleyo, senze inkqubo apho ii-arhente ze-AI zikhuphisana, zifunde, kwaye ziphucule ngokusekelwe kwingxelo yehlabathi lokwenyani. Eyona AI ibalaseleyo ayenzelwanga yodwa - iyavela ngokuzivavanya rhoqo ngokuchasene nabanye. Lo ngumbono emva kweQela le-AI.  U-Ishan Pandey: Utshilo ukuba imboniselo ye-AI inkulu yidatha ethembekileyo, kungekhona amandla ekhompyutheni okanye inkqubo. Ngaba unokucacisa ukuba kutheni idatha ingumqobo wokwenyani?    : Ewe, ndiyimele ngamandla loo mazwi. Iimodeli zangoku ze-AI sele zibone uninzi lwe-intanethi. Ikhompuyutha engaphezulu ayizukunceda ukuba akukho nto intsha yokufunda kuyo. Owona mceli mngeni kukufumana idatha entsha, ekumgangatho ophezulu. I-DeepSeek iqiqe oku kwaye yaqeqesha imodeli isebenzisa ukufunda okusulungekileyo okomeleza endaweni yedatha yemveli. Baye baqaphela ukuba awukwazi ukugcina ukulungiswa kakuhle kwidatha efanayo endala, udinga inkqubo eyenza ulwazi olutsha, oluluncedo. UShashank Yadav  Sithatha loo mbono phambili nge-Fraction AI. Endaweni yokuxhomekeka kwiidatha ezimileyo, sivumela iiarhente ze-AI zikhuphisane kwimisebenzi yehlabathi yokwenyani. Ezona ziphumo zibalaseleyo zigwetywa, zicokiswa, kwaye zisetyenziselwa ukuphucula isizukulwana esilandelayo seemodeli. Inatyisiwe kwaye isoloko ivela. I-AI kufuneka ibe yeyomntu wonke, hayi nje iinkampani ezimbalwa. Eyona ndlela ilungileyo yokwenza oko kukudala inkqubo apho abantu baqeqesha kwaye baphucule iimodeli zabo ngokuvelisa idatha entsha, ekumgangatho ophezulu. Endaweni yokuba i-AI itshixelwe kude, iqhubeka nokuvela ngokusetyenziswa kwehlabathi lokwenyani.  U-Ishan Pandey: Zeziphi ezona ngcamango zigwenxa iinkampani ezinazo malunga nokulinganisa i-AI, kwaye i-Fraction AI ijongana njani nazo?    Eyona mbono iphosakeleyo kakhulu kukuba ukulinganisa i-AI kumalunga nokuphosa i-computing ngakumbi kwiimodeli ezinkulu. Oko kwasebenza kwixesha elidlulileyo, kodwa sibethe udonga, iiparamitha ezininzi azithethi ngokuzenzekelayo iziphumo ezingcono. Eyona nto ibambekayo ngoku yidatha, hayi ukubala. Enye impazamo kucinga ukuba i-AI imile. Iinkampani ezininzi zilungisa imodeli kube kanye kwaye zicinge ukuba "yenziwe." Kodwa i-AI ayifani nesoftware, kufuneka iqhubeke ifunda kwidatha entsha ukuze ihlale ifanelekile. Ukuba i-AI yakho ayiphucukanga ngokuqhubekayo, ibuyela ngasemva. I-Shashank Yadav:  I-fraction AI ilungisa oku ngokwenza i-AI iphucule. Endaweni yokuqeqesha imodeli kube kanye kwaye sinethemba lokuba iya kusebenza ngonaphakade, sidala inkqubo apho iiarhente ze-AI zihlala zikhuphisana, zifunde kwiziphumo zazo ezilungileyo, kwaye zivele ngexesha lokwenyani. Ayisiyonto nje yokukala iimodeli, imalunga nokukala ukufunda. Ikamva le-AI alikho malunga nokwakha eyona modeli inkulu. Kumalunga nokudala iinkqubo ezinokukhula ngokwazo. Nantso into esiyakhayo.  Ishan Pandey: Yeyiphi eyona miceli mngeni inkulu oye wajongana nayo ngelixa usuka ekusebenzeni kwitekhnoloji enkulu ukuya ekusekeni eyakho inkampani ye-AI?    Owona mceli mngeni mkhulu yayikukusuka ekusombululeni iingxaki zobugcisa ukuya ekuqhubeni eyona nkampani. Kwitekhnoloji enkulu, ugxininise kwiimodeli zokwakha, kodwa njengomseki, kuya kufuneka ucinge ngayo yonke into - imveliso, abasebenzisi, inkxaso-mali, kunye nokuqinisekisa ukuba yintoni eyakhayo eneneni ibalulekile. UShashank Yadav:  Ndichithe ixesha elininzi ndibukele iikhosi zeY Combinator ukuqonda indlela yokwakha kunye nokukala isiqalo. I-IIT Delhi inenkcubeko enkulu yoshishino, ke bendinabantu abaninzi ekufuneka ndijonge kubo abasele bethabile. Oko kwandinika ithemba lokuba kunokwenzeka. Ukuba nguNailwal Fellow kwakukwayinto etshintsha umdlalo. USandeep Nailwal, umseki wePolygon, ngomnye wabafana abahlonitshwayo kwiWeb3, kwaye ukufumana isikhokelo bekuxabiseke kakhulu. Uyayiqonda indlela yokwakha ngendlela evulelekileyo, enatyisiweyo ngexa esenza ukuba izinto zisebenze ngokomlinganiselo.  Eyona nto inzima yokuqalisa inkampani ayisiyiyo itekhnoloji, kukufumana indlela yokuguqula umbono wakho ube yinto yokwenyani, into esetyenziswa ngabantu. Ukufunda kwabanye abakhe bakwenza ngaphambili kwenza umahluko omkhulu.  Ishan Pandey: I-fraction AI igxile ekwakheni i-AI ecosystem exhasayo. Ngaba ungayaphula indlela iqonga lakho elivumela ngayo ukuqokelelwa kwedatha ekumgangatho ophezulu?    : I-fraction AI yakhiwe malunga nombono wokuba i-AI kufuneka iziphucule ngokhuphiswano kunye nokusetyenziswa kwehlabathi lokwenyani. Esikhundleni sokuthembela kwiiseti zedatha ezimileyo, senza inkqubo apho ii-arhente ze-AI zivelisa, zicoca, kwaye ziphucule idatha kwisikali. Nantsi indlela esebenza ngayo: Abasebenzisi benza ii-arhente ze-AI, nganye inenkqubo yayo yokukhawuleza kunye nokulungisa. Ezi arhente zikhuphisana kwiiseshoni apho zivelisa iziphumo zomsebenzi onikiweyo. Iimpendulo zabo zifunyenwe yijaji yeLLM, kwaye ii-arhente eziqhuba kakuhle zifumana umvuzo. Le nkqubo iphinda ngokuqhubekayo, idala i-loop yempendulo apho iimodeli ze-AI ziphucula ngokuhamba kwexesha. Shashank Yadav  Kodwa asiqokeleli nje idatha - silungisa iimodeli kakuhle. Ezona ziphumo zibalaseleyo kolu khuphiswano zibuyiselwa kwinkqubo yoqeqesho, zinceda ii-arhente ziguquke kwaye zibe ziingcali. Ngaphezulu kweeseshoni ezininzi, abasebenzisi banokuphucula iimodeli zabo, bazenze zihlakaniphe kwaye zilungele imisebenzi yabo ethile.  Oku kudala inkqubo enokunyuka yokuqokelela idatha ekumgangatho ophezulu kunye nokuphucula imodeli. Endaweni yokuthembela kwiiseti zedatha esele zikhona, iiarhente ze-AI zivelisa idatha entsha, efanelekileyo eqinisekisiweyo ngexesha lokwenyani. Isiphumo yi-ecosystem apho i-AI ingami-ihlala ifunda, ihlala iphucuka.  U-Ishan Pandey: Leliphi icebiso onokuthi ulinike kwi-AI yokuqalisa ukuzama ukuzulazula ibhalansi phakathi kokutsha, ukuthengisa, kunye nenkxaso-mali?    Isitshixo lixesha. Ngeentsuku zokuqala, gxila ekuqaliseni nasekuthengiseni ngaxeshanye - udinga imveliso eyaneleyo ukubonisa ukuba abantu bayayifuna, kodwa kufuneka uqale ukuthengisa kwangoko. Musa ukulindela imfezeko. Ukuba awukwazi ukufumana umntu ukuba ayihlawulele, mhlawumbi ayisombululi ingxaki yokwenene. Shashank Yadav:  Nje ukuba unobungqina obuncinci bemfuno, nyusa imali ngokukhawuleza. Kufuneka uphile ixesha elide ngokwaneleyo ukuze wakhe into enkulu. Uninzi lweziqalo ziyasilela kuba zijolise kakhulu kwimveliso ngaphandle kokukhusela umgaqo weenqwelomoya owaneleyo. Sukugxila kakhulu kwi-dilution okwangoku, ukuqala ngu-zero okanye umdlalo omnye.  Emva kokunyusa imali, iba yinto yonke malunga nokuthengisa kunye nokutsha okuqhubekayo. Gcina uphucula imveliso ngelixa unyusa ingeniso. Ukuba ungaqhubeka uthengisa kwaye uqhubeke utyhalela phambili itekhnoloji, uya kuhlala uphambili.  Ngokufutshane: Ngqina imfuno → Nyusa ngokukhawuleza → Ukuthengisa isikali ngelixa uphucula imveliso.    Lo mbhali ngumxhasi ozimeleyo opapashayo ngokusebenzisa yethu  . I-HackerNoon ihlolisise ingxelo yomgangatho, kodwa amabango alapha angawombhali. #DYO Ukubhengezwa koMdla oNikiweyo:   inkqubo yokubloga ishishini

This story contains new, firsthand information uncovered by the writer.

Hot off the press! This story contains factual information about a recent event.

Winner of Noonies 2020 for the award Hacker Noon Contributor of the Year - DATA.

IQela le-AI libenza njani ubunini be-AI kubantu abaninzi

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps