Uma sibuyela kungqikithi, impumelelo ye-AIGC ebunyeni iyinhlanganisela yezinto ezintathu:

1. I-GPT iwumfanekiso wama-neurons womuntu

I-GPT AI emelwe yi-NLP iyi-algorithm ye-neural network algorithm, okuyisisekelo sayo kuwukulingisa amanethiwekhi e-neural ku-human cerebral cortex.

Ukucubungula nokucabanga okukhaliphile kolimi, umculo, izithombe, ngisho nolwazi lokunambitha konke kuyimisebenzi eqoqwe ngumuntu.

ubuchopho "njengekhompyutha yamaprotheni" ngesikhathi sokuziphendukela kwemvelo kwesikhathi eside.

Ngakho-ke, i-GPT ngokwemvelo iwukulingisa ofaneleka kakhulu ekucubunguleni ulwazi olufanayo, okungukuthi, ulimi olungahlelekile, umculo, nezithombe.

Indlela yokucutshungulwa kwayo ayikona ukuqonda incazelo, kodwa kunalokho inqubo yokucwenga, ukuhlonza, nokuhlanganisa.Lokhu kubi kakhulu

into exakayo.

I-algorithms yokuqaphela i-semantic yenkulumo yangaphambi kwesikhathi isungule imodeli yohlelo kanye nesizindalwazi senkulumo, base bebeka inkulumo kumephu yamagama,

wabe esebeka isilulumagama kusizindalwazi sohlelo ukuze aqonde incazelo yamagama, futhi ekugcineni athole imiphumela yokuqashelwa.

Ukusebenza kahle kokuqashelwa kwalokhu "indlela enengqondo" yokuqashelwa kwe-syntax bekulokhu kuhamba cishe ku-70%, njengokuqashelwa kwe-ViaVoice.

I-algorithm eyethulwa yi-IBM ngawo-1990.

I-AIGC ayikona ukudlala kanje.Ingqikithi yayo ayikona ukukhathalela uhlelo lolimi, kodwa kunalokho ukusungula i-algorithm yenethiwekhi ye-neural evumela ukuthi i-

ikhompuyutha ukubala ukuxhumana okungenzeka phakathi kwamagama ahlukene, okuyizixhumanisi zemizwa, hhayi ukuxhumana kwe-semantic.

Njengokufunda ulimi lwethu lwebele lapho sisebancane, ngokwemvelo salufunda, kunokuba sifunde “isihloko, isilandiso, umenziwa, isenzo, umphelelisi,”

bese uqonda isigaba.

Lena imodeli yokucabanga ye-AI, okuwukuqashelwa, hhayi ukuqonda.

Lokhu futhi ukubaluleka kokuguqula i-AI kuwo wonke amamodeli emishini yakudala - amakhompyutha awadingi ukuqonda lolu daba ezingeni eliphusile,

kodwa kunalokho khomba futhi uqaphele ukuhlobana phakathi kolwazi lwangaphakathi, bese ukwazi ke.

Isibonelo, isimo sokugeleza kwamandla nokubikezela kwamagridi amandla kusekelwe ekulingiseni kwenethiwekhi yamandla yakudala, lapho imodeli yezibalo

Indlela yokusebenza iyasungulwa bese ihlanganiswa kusetshenziswa i-algorithm ye-matrix.Esikhathini esizayo, kungase kungadingeki.I-AI izokhomba ngokuqondile futhi ibikezele a

iphethini ethile ye-modal esekelwe esimweni senodi ngayinye.

Uma kukhona ama-node amaningi, i-algorithm ye-matrix yakudala idume kancane, ngoba ubunzima be-algorithm buyanda ngenani

ama-node kanye nokuqhubeka kwejometri kuyanda.Kodwa-ke, i-AI ikhetha ukuba ne-concurrency ye-node yesikali esikhulu kakhulu, ngoba i-AI inhle ekuhlonzeni futhi

ukubikezela amamodi enethiwekhi okungenzeka kakhulu.

Kungakhathaliseki ukuthi ukubikezela okulandelayo kwe-Go (i-AlphaGO ingabikezela inqwaba yezinyathelo ezilandelayo, namathuba amaningi esinyathelo ngasinye) noma isibikezelo se-modal

wezinhlelo zezulu eziyinkimbinkimbi, ukunemba kwe-AI kuphakeme kakhulu kunaleyo yamamodeli emishini.

Isizathu sokuthi kungani igridi yamandla okwamanje ingadingi i-AI ukuthi inani lamanodi ku-220 kV nangaphezulu kwamanethiwekhi kagesi aphethwe yisifundazwe.

ukuthunyelwa akukukhulu, futhi izimo eziningi zisethwe ukuze zihlukanise futhi zinciphise i-matrix, kunciphisa kakhulu inkimbinkimbi yokubala

imodeli yemishini.

Kodwa-ke, esigabeni sokugeleza kwamandla enethiwekhi yokusabalalisa, ebhekene namashumi ezinkulungwane noma amakhulu ezinkulungwane zamanodi amandla, izindawo zokulayisha, kanye nendabuko.

I-matrix algorithms kunethiwekhi enkulu yokusabalalisa ayinawo amandla.

Ngikholwa ukuthi ukubonwa kwephethini ye-AI ezingeni lenethiwekhi yokusabalalisa kuzokwenzeka esikhathini esizayo.

2. Ukuqoqwa, ukuqeqeshwa, kanye nokukhiqizwa kolwazi olungahlelekile

Isizathu sesibili sokuthi kungani i-AIGC yenze impumelelo ukuqoqwa kolwazi.Kusukela ku-A/D ukuguqulwa kwenkulumo (microphone+PCM

sampling) ekuguquleni kwe-A/D kwezithombe (CMOS+color space mapping), abantu baqongelele idatha ye-holographic kokubonakalayo nokuzwayo.

amasimu ngezindlela ezingabizi kakhulu emashumini ambalwa eminyaka adlule.

Ikakhulukazi, ukuthandwa okukhulu kwamakhamera nama-smartphones, ukuqoqwa kwedatha engahlelekile emkhakheni we-audiovisual wabantu.

cishe ngezindleko eziphansi, futhi ukuqoqwa kolwazi lombhalo okuqhumayo ku-inthanethi kuyisihluthulelo sokuqeqeshwa kwe-AIGC - amasethi edatha yokuqeqesha awabizi.

Isibalo esingenhla sibonisa ukuthambekela kokukhula kwedatha yomhlaba wonke, eveza ngokucacile ithrendi ekhulayo.

Lokhu kukhula okungekona komugqa kokuqoqwa kwedatha kuyisisekelo sokukhula okungewona umugqa wamakhono e-AIGC.

KODWA, iningi lale datha liyidatha yomsindo nokubukwayo engahlelekile, eqoqwa ngezindleko ezinguziro.

Emkhakheni wamandla kagesi, lokhu akukwazi ukufezwa.Okokuqala, iningi lemboni yamandla kagesi linemininingwane ehlelekile futhi ehlelekile, njenge

i-voltage neyamanje, okungamasethi edatha yephoyinti ochungechunge lwesikhathi kanye ne-semi structured.

Amasethi edatha yesakhiwo adinga ukuqondwa ngamakhompyutha futhi adinga “ukuqondanisa”, njengokuqondanisa kwedivayisi - i-voltage, idatha yamanje, namandla

I-switch kumele ihambisane nale nodi.

Okukhathaza kakhulu ukuqondanisa isikhathi, okudinga ukuvumelanisa i-voltage, amandla amanje, asebenzayo nasebenzayo ngokususelwa esikalini sesikhathi, ukuze

ukuhlonza okulandelayo kungenziwa.Kukhona nezikhombisi-ndlela eziya phambili nezihlehlayo, eziwukuqondanisa kwendawo ngamaquadranti amane.

Ngokungafani nedatha yombhalo, engadingi ukuqondanisa, isigaba simane siphonswe kukhompuyutha, ekhomba ukuhlotshaniswa kolwazi okungenzeka

ngokwalo.

Ukuze uqondanise lolu daba, njengokuqondanisa kwemishini yedatha yokusatshalaliswa kwebhizinisi, ukuqondanisa kuyadingeka njalo, ngoba okuphakathi nendawo

inethiwekhi yokusabalalisa amandla kagesi aphansi iyengeza, isusa, futhi ilungisa amathuluzi nemigqa nsuku zonke, futhi izinkampani zegridi zisebenzisa izindleko ezinkulu zokusebenza.

Njengokuthi "isichasiselo sedatha," amakhompyutha awakwazi ukwenza lokhu.

Okwesibili, izindleko zokuthola idatha emkhakheni wamandla ziphezulu, futhi izinzwa ziyadingeka esikhundleni sokuba nomakhalekhukhwini ukuze ukhulume futhi uthathe izithombe.”

Ngaso sonke isikhathi i-voltage yehla ngezinga elilodwa (noma ubudlelwano bokusabalalisa amandla buncipha ngezinga elilodwa), ukutshalwa kwezimali kwenzwa edingekayo kuyanda

okungenani ngohlelo olulodwa lobukhulu.Ukufinyelela ohlangothini lomthwalo (ukuphela kwe-capillary end), kuwutshalomali olukhulu kakhulu lwedijithali. ”

Uma kudingekile ukukhomba imodi yesikhashana yegridi yamandla, ukunemba okuphezulu kwesampuli yamafrikhwensi kuyadingeka, futhi izindleko ziphezulu kakhulu.

Ngenxa yezindleko eziphakeme kakhulu ze-marginal zokutholwa kwedatha nokuqondanisa kwedatha, igridi yamandla okwamanje ayikwazi ukuqongelela okwanele okungewona umugqa.

ukukhula kolwazi lwedatha ukuqeqesha i-algorithm ukuze ifinyelele ububodwa be-AI.

Ingasaphathwa eyokuvuleka kwedatha, akwenzeki ukuthi i-AI yamandla iqale ithole le datha.

Ngakho-ke, ngaphambi kwe-AI, kuyadingeka ukuxazulula inkinga yamasethi wedatha, ngaphandle kwalokho ikhodi ye-AI evamile ayikwazi ukuqeqeshwa ukukhiqiza i-AI enhle.

3. Ukuphumelela emandleni wokubala

Ngokungeziwe kuma-algorithms nedatha, impumelelo yobubodwa ye-AIGC iphinde ibe impumelelo emandleni wokubala.Ama-CPU endabuko awekho

ilungele ikhompyutha enkulu ngesikhathi esisodwa ye-neuronal.Ukusetshenziswa kwama-GPUs emidlalweni ye-3D namamuvi okwenza ukufana okukhulu

indawo entantayo+ikhompuyutha yokusakaza ingenzeka.Umthetho ka-Moore uqhubeka unciphisa izindleko zokubala ngeyunithi ngayinye yamandla wokubala.

Igridi yamandla AI, inkambiso engenakugwemeka esikhathini esizayo

Ngokuhlanganiswa kwenani elikhulu lamasistimu okugcina amandla e-photovoltaic asabalalisiwe, kanye nezidingo zohlelo lokusebenza.

ukulayisha okuseceleni kwezitshalo zamandla abonakalayo, kuyadingeka ngokufanelekile ukwenza isibikezelo somthombo nokulayisha sezinhlelo zenethiwekhi yokusabalalisa umphakathi kanye nomsebenzisi

ukusatshalaliswa (ama-micro) amasistimu wegridi, kanye nokwenza kahle kokugeleza kwamandla ngesikhathi sangempela sokusabalalisa (ama-micro) amasistimu wegridi.

Ubunkimbinkimbi bekhompiyutha bohlangothi lwenethiwekhi yokusabalalisa empeleni bungaphezulu kwalokho okuhlelelwa kwenethiwekhi yokudlulisela.Ngisho nezentengiso

okuyinkimbinkimbi, kungase kube amashumi ezinkulungwane zamadivayisi okulayisha kanye namakhulu okushintsha, kanye nesidingo sokusebenza kwenethiwekhi encane/yokusabalalisa kwenethiwekhi ye-AI.

ukulawula kuzovela.

Ngezindleko eziphansi zezinzwa kanye nokusetshenziswa kabanzi kwamandla kagesi njengama-solid-state transformers, ama-solid-state switch, nama-inverters (ama-converter),

ukuhlanganiswa kwezinzwa, ikhompyutha, nokulawula emaphethelweni egridi yamandla nakho sekuphenduke inkambiso emisha.

Ngakho-ke, i-AIGC yegridi yamandla iyikusasa.Kodwa-ke, okudingekayo namuhla ukuthi ungakhiphi ngokushesha i-algorithm ye-AI ukwenza imali,

Kunalokho, qala ubhekane nezindaba zokwakha ingqalasizinda yedatha edingwa yi-AI

Ekukhuphukeni kwe-AIGC, kudingeka kube nokucabanga okwanele okuzolile mayelana nezinga lesicelo nekusasa lamandla e-AI.

Njengamanje, ukubaluleka kwamandla e-AI akubalulekanga: isibonelo, i-algorithm ye-photovoltaic enokunemba kokubikezela kwe-90% ibekwe emakethe yendawo.

ngomkhawulo wokuchezuka kokuhweba we-5%, futhi ukuchezuka kwe-algorithm kuzosula yonke inzuzo yokuhweba.

Idatha ingamanzi, futhi amandla okuhlanganisa we-algorithm yisiteshi.Njengoba kwenzeka, kuyoba njalo.