Kudzokera kune essence, kubudirira kweAIGC muhumwe musanganiswa wezvinhu zvitatu:

1. GPT mufananidzo wetsinga dzevanhu

GPT AI inomiririrwa neNLP ndeyekombuta neural network algorithm, ine musimboti kutevedzera neural network mumunhu cerebral cortex.

Iko kugadzirisa uye kungwara fungidziro yemutauro, mimhanzi, mifananidzo, uye kunyange kuravira ruzivo ese mabasa akaunganidzwa nemunhu.

uropi se "protein komputa" panguva yekushanduka-shanduka kwenguva refu.

Naizvozvo, GPT ndiyo yakajairika yekutevedzera yakanyatsokodzera kugadzirisa ruzivo rwakafanana, kureva, mutauro usina kurongeka, mimhanzi, uye mifananidzo.

Iyo nzira yekugadziriswa kwayo hakusi kunzwisiswa kwezvinoreva, asi kuti maitiro ekunatsa, kuona, uye kubatanidza.Izvi zvakanyanya

paradoxical chinhu.

Matauriro ekutanga semantic recognition algorithms akasimbisa modhi yegirama uye dhatabhesi rekutaura, ndokuronga kutaura kune mazwi,

ndokuisa mazwi mudura regirama kuti vanzwisise zvinoreva mazwi, uye pakupedzisira vakawana mhedzisiro yekuzivikanwa.

Iko kugona kwekuziva kweiyi "inonzwisisika meshini" yakavakirwa syntax kuzivikanwa yanga ichitenderera ichitenderedza 70%, senge ViaVoice kuzivikanwa.

algorithm yakaunzwa neIBM muma1990s.

AIGC haisi yekutamba seizvi.Chinhu chayo hachisi chekuitira hanya nezve girama, asi panzvimbo pekumisa neural network algorithm inobvumira iyo

Komputa kuverenga zvinogoneka zvakabatanidzwa pakati pemazwi akasiyana, ari neural connections, kwete semantic connections.

Kungofanana nokudzidza rurimi rwaamai patakanga tichiri vaduku, takarudzidza nenzira yomuzvarirwo, panzvimbo pokudzidza “musoro, chivakashure, chinhu, chiito, chikwaniriso,”

uyezve kunzwisisa ndima.

Iyi ndiyo nzira yekufunga yeAI, inova kuzivikanwa, kwete kunzwisisa.

Uku ndiko zvakare kukanganisa kukosha kweAI kune ese emhando dzemhando dzemhando - makomputa haafanire kunzwisisa nyaya iyi pamwero unonzwisisika,

asi ziva uye ziva kuwirirana pakati peruzivo rwemukati, wobva wazviziva.

Semuyenzaniso, iyo simba rekuyerera mamiriro uye kufanotaura kwemagetsi magidhi akavakirwa pane classical simba network simulation, uko masvomhu modhi ye

meshini inomiswa uye yobva yashandurwa uchishandisa matrix algorithm.Mune ramangwana, zvingasava zvakakosha.AI ichazivisa zvakananga uye kufanotaura a

imwe modal pateni yakavakirwa pachimiro cheimwe node.

Iyo yakawanda node iripo, iyo inonyanya kufarirwa iyo classical matrix algorithm ndeye, nekuti kuoma kweiyo algorithm kunowedzera nehuwandu hwe.

nodes uye kufambira mberi kwejometri kunowedzera.Nekudaro, AI inofarira kuve neyakakura chiyero node concurrency, nekuti AI yakanaka pakuziva uye

kufanotaura zvingangoitika network modes.

Kungave kuri kufanotaura kunotevera kweGo (AlphaGO inogona kufanotaura makumi emakumi enhanho anotevera, ine isingaverengeke mikana yedanho rega rega) kana modal kufanotaura.

yemamiriro ekunze akaomarara, kurongeka kweAI kwakakwira zvakanyanya kupfuura iyo yemakina mamodheru.

Chikonzero nei gidhi remagetsi parizvino risingade AI ndechekuti nhamba yemanodhi mu220 kV uye pamusoro pemagetsi network inotungamirwa nedunhu.

kutumira hakuna kukura, uye akawanda mamiriro akaiswa kuti aite mutsara uye apfuure matrix, zvichideredza zvakanyanya kuoma kwe computational

mechanic model.

Nekudaro, panguva yekugovera network simba rekuyerera nhanho, yakatarisana nemakumi ezviuru kana mazana ezviuru zvemagetsi node, mutoro node, uye zvechinyakare.

matrix algorithms mune yakakura yekugovera network haina simba.

Ini ndinotenda kuti kucherechedzwa kwemaitiro eAI padanho retiweki yekugovera zvinogoneka mune ramangwana.

2. Kuunganidzwa, kudzidziswa, uye kugadzirwa kwemashoko asina kurongeka

Chikonzero chechipiri nei AIGC yaita pundutso kuunganidzwa kweruzivo.Kubva kuA/D shanduko yekutaura (microphone+PCM

sampling) kune A/D shanduko yemifananidzo (CMOS + color space mepu), vanhu vakaunganidza holographic data mune yekuona uye yekunzwa.

minda nenzira dzinodhura zvakanyanya mumakumi mashoma emakore apfuura.

Kunyanya, kukuru-kukurumbira kwemakamera uye mafoni, kuunganidzwa kwedata risina kurongeka mundima yeaudiovisual yevanhu.

pamutengo ungangoita zero, uye kuputika kuunganidzwa kweruzivo rwemavara paInternet ndiyo kiyi yekudzidziswa kweAIGC - seti yedata yekudzidzisa haina kudhura.

Nhamba iri pamusoro inoratidza kukura kwe data yepasi rose, iyo inoratidza zvakajeka maitiro eexponential.

Uku kusiri-mutsara kukura kwekuunganidzwa kwedata ndiyo hwaro hwekusakura-mutsara kwehunyanzvi hweAIGC.

ASI, mazhinji eaya data haana kurongeka odhiyo-anoona data, iyo inounganidzwa pamutengo we zero.

Mumunda wemagetsi emagetsi, izvi hazvigone kuwanikwa.Chekutanga, mazhinji emagetsi emagetsi indasitiri yakarongeka uye semi yakarongeka data, senge

voltage uye zvazvino, ayo ari point data seti yenguva yakatevedzana uye semi yakarongeka.

Structural data seti inoda kunzwisiswa nemakomputa uye inoda "kurongeka", senge kurongeka kwemidziyo - voltage, yazvino, uye simba data.

yekuchinja inoda kuwiriraniswa kune iyi node.

Zvimwe zvinonetsa kurongeka kwenguva, izvo zvinoda kuenzanisa voltage, ikozvino, uye inoshanda uye inoshanda uye inoshanda simba zvichienderana nechiyero chenguva, kuti

chitupa chinotevera chinogona kuitwa.Kune zvakare mafambiro emberi neanodzokera kumashure, ayo ari kurongeka kwenzvimbo mune mana quadrants.

Kusiyana nedata data, iyo isingade kurongeka, ndima inongokandwa pakombuta, inotaridza hukama hweruzivo hunogona kuitika.

pachayo.

Kuti uenzanise nyaya iyi, senge kurongeka kwemidziyo yedata rekugovera bhizinesi, kurongeka kunogara kuchidiwa, nekuti yepakati uye.

low voltage distribution network iri kuwedzera, kudzima, nekugadzirisa midziyo nemitsetse zuva rega rega, uye makambani egridi anoshandisa mari yakawanda yebasa.

Kufanana ne "data annotation," makomputa haagone kuita izvi.

Chechipiri, mutengo wekutora data muchikamu chemagetsi wakakwira, uye masensa anodiwa pane kuve nenharembozha yekutaura nekutora mafoto.”

Nguva imwe neimwe iyo voltage inodzikira nedanho rimwe (kana hukama hwekugovera simba hunoderera nedanho rimwe), iyo inodiwa sensor investimendi inowedzera

nechero hurongwa humwe hwehukuru.Kuti uwane chikamu chekuremerwa (capillary end) kunzwa, inotonyanya kuwedzera mari yedhijitari."

Kana zvichidikanwa kuziva iyo yenguva pfupi yemagetsi grid, yakakwirira-chaiyo yakakwira-frequency sampling inodiwa, uye mutengo wakatokwira.

Nekuda kwemutengo wakanyanya wepamusoro wekutora data uye kurongeka kwedata, gidhi remagetsi parizvino harikwanisi kuunganidza zvakakwana zvisiri-mutsara.

kukura kwe data data kudzidzisa algorithm kusvika kune iyo AI humwe.

Tisingarevi kuvhurika kwedata, hazvigoneke kune simba AI kutanga kuwana aya data.

Naizvozvo, pamberi peAI, zvinodikanwa kugadzirisa dambudziko remaseti e data, zvikasadaro general AI kodhi haigone kudzidziswa kuburitsa yakanaka AI.

3. Kubudirira musimba rekuverenga

Pamusoro pealgorithms uye data, iyo singularity budiriro yeAIGC zvakare budiriro mune computational simba.Traditional CPUs haasi

yakakodzera kune yakakura-yakafanana neuronal computing.Ndizvo chaizvo mashandisirwo eGPUs mumitambo ye3D nemabhaisikopo anoita hukuru hwakafanana.

inoyangarara-nzvimbo + yekuyerera komputa inogoneka.Mutemo waMoore unowedzera kuderedza mutengo wekombuta pachikamu chesimba remakomputa.

Simba grid AI, maitiro asingadzivisike mune ramangwana

Nekubatanidzwa kwenhamba yakawanda yakagoverwa photovoltaic uye yakagoverwa ekuchengetedza simba masisitimu, pamwe chete nezvinodiwa zvekushandisa

load side virtual magetsi zvidyarwa, zvine hungwaru zvinodikanwa kuita sosi uye kurodha kufanotaura kweruzhinji kugovera network masisitimu uye mushandisi

kugovera (micro) grid masisitimu, pamwe neiyo chaiyo-nguva simba kuyerera optimization yekugovera (micro) grid masisitimu.

Iyo computational kuomarara kwedivi rekugovera network yakanyatso kukwira kupfuura iyo yekutumira network kuronga.Kunyangwe zvekutengesa

yakaoma, panogona kunge paine makumi ezviuru zvezvishandiso zvekutakura uye mazana ekuchinja, uye kudiwa kweAI yakavakirwa micro grid / kugovera network kushanda.

kutonga kuchamuka.

Nekudhura kwakaderera kwema sensors uye kushandiswa kwakapararira kwemagetsi emagetsi emagetsi akadai seakasimba-state transformers, solid-state switch, uye inverters (converters),

kubatanidzwa kwekunzwa, komputa, uye kutonga pamucheto wegidhi remagetsi rakavawo maitiro matsva.

Naizvozvo, iyo AIGC yemagetsi grid ndiyo ramangwana.Nekudaro, chinodiwa nhasi hachisi chekungotora AI algorithm kuita mari,

Pane kudaro, tanga wagadzirisa nyaya dzekuvaka dhata dzinodiwa neAI

Mukukwidziridzwa kweAIGC, panofanirwa kuve nekudzikama kwakaringana kufunga nezvedanho rekushandisa uye ramangwana resimba AI.

Parizvino, kukosha kwesimba AI hakuna kukosha: semuenzaniso, photovoltaic algorithm ine fungidziro yechokwadi ye90% inoiswa mumusika wenzvimbo.

nekutengesa kutsauka pachikumbaridzo che5%, uye kutsauka kwegorgorithm kuchabvisa purofiti yese yekutengesa.

Iyo data imvura, uye iyo computational simba regorgorithm inzira.Sezvazvinoitika, zvichava.