Amanethiwekhi e-neural ajulile: Ubuchopho obufihliwe obunika amandla i-AI

• About the Author:
• Igama lomlobi

  I-Quantumrun Foresight
• April 6, 2023

Ama-algorithms kanye nedatha enkulu sekuphenduke amagama ahamba phambili esikhaleni se-artificial intelligence (AI), kodwa amanethiwekhi emizwa yokwenziwa (ANN) yiwo awavumela ukuthi abe ngamathuluzi anamandla. Lawa ma-ANN asetshenziselwa ukubona amaphethini, ahlukanise idatha, futhi enze izinqumo ngokusekelwe kudatha yokufaka. 

Umongo wenethiwekhi ye-neural ejulile

Amanethiwekhi emizwa yokwenziwa azama ukulingisa inkimbinkimbi yobuhlakani bomuntu ngokwakha inethiwekhi yesofthiwe, amakhodi, ne-algorithms ukuze kucutshungulwe okokufaka (idatha/amaphethini) futhi awafanise nokuphumayo okusebenza kakhulu (umphumela/imiphumela). I-ANN isendlalelo esifihliwe esicubungula futhi sixhumanise ubudlelwano phakathi kwedatha nokwenza izinqumo. Uma i-ANN yakha phakathi kokufaka nokukhiphayo, umshini ufunda okwengeziwe ngenxa yokutholakala kwedatha eyinkimbinkimbi. Izendlalelo eziningi ze-ANN zaziwa ngokuthi amanethiwekhi e-neural ajulile ngenxa yokuthi angakwazi ukungena kumthamo omkhulu wedatha yokuqeqeshwa futhi athuthukise isisombululo noma amaphethini angcono kakhulu. 

Umshini "ufundisiwe" ngokuqhubekayo ngokusabalalisa emuva, inqubo yokulungisa amapharamitha akhona ukuze uqeqeshe ama-algorithms ukuze uqhamuke nomphumela/ukuhlaziywa okungcono kakhulu. Amanethiwekhi emizwa yokwenziwa angaqeqeshelwa ukwenza imisebenzi eyahlukene, efana nokubonwa kwesithombe nenkulumo, ukuhumusha ulimi, kanye nokudlala imidlalo. Benza lokhu ngokulungisa amandla okuxhumana phakathi kwama-neurons, aziwa njengezisindo, ngokusekelwe kudatha yokufaka abayitholayo phakathi nenqubo yokuqeqesha. Le ndlela ivumela inethiwekhi ukuthi ifunde futhi ihambisane nesikhathi, ithuthukise ukusebenza kwayo emsebenzini. Kunezinhlobo eziningi zama-ANN, okuhlanganisa amanethiwekhi e-feedforward, amanethiwekhi e-convolutional neural (CNNs), namanethiwekhi e-neural aphindaphindiwe (ama-RNN). Uhlobo ngalunye ludizayinelwe ukuthi lufaneleke ngokukhethekile umsebenzi othile noma isigaba sedatha.

Umthelela ophazamisayo

Ayikho imboni namuhla engasebenzisi amanethiwekhi ajulile we-neural kanye ne-AI ukwenza ngokuzenzakalelayo izinqubo zebhizinisi futhi iqoqe ubuhlakani bemakethe. Mhlawumbe ukusetshenziswa okusobala kakhulu kwamanethiwekhi e-neural ajulile imboni yokumaketha, lapho i-AI icubungula izigidi zolwazi lwamakhasimende ukuze ikhombe ngokunembile amaqembu athile okungenzeka ukuthi athenge umkhiqizo noma isevisi. Ngenxa yokunemba okuphezulu okwandayo kwalokhu kuhlaziywa kwedatha, imikhankaso yokumaketha isibe yimpumelelo kakhulu nge-hypertargeting (ukuhlonza amasethi amancane amakhasimende kanye nokuwathumela imilayezo eyenziwe ngokwezifiso kakhulu). 

Okunye ukusetshenziswa okuvelayo isofthiwe yokuqaphela ubuso, indawo yenkulumompikiswano ehlobene nokuvikeleka ku-inthanethi kanye nobumfihlo bedatha. Ukuqashelwa kobuso okwamanje kusetshenziswa kusukela ekuqinisekiseni uhlelo lokusebenza kuya emthethweni futhi kunikwe amandla amanethiwekhi ajulile e-neural acubungula amarekhodi amaphoyisa kanye nezithombe ozishuthe zona ezithunyelwe ngabasebenzisi. Izinsizakalo zezezimali zingomunye umkhakha ozuza kakhulu kumanethiwekhi ajulile e-neural, esebenzisa i-AI ukubikezela ukunyakaza kwezimakethe, ukuhlaziya izicelo zemalimboleko, nokuhlonza ukukhwabanisa okungenzeka.

Amanethiwekhi ajulile we-neural angaphinda ahlaziye izithombe zezokwelapha, njengama-x-reyi kanye ne-magnetic resonance imaging (MRI), ukusiza ukuhlonza izifo nokubikezela imiphumela yesiguli. Angaphinda asetshenziselwe ukuhlaziya amarekhodi ezempilo e-elekthronikhi ukukhomba amathrendi nezici zobungozi ezimweni ezithile. Amanethiwekhi e-Neural nawo anamandla okusetshenziswa ekutholweni kwezidakamizwa, emithini yomuntu siqu, nasekulawuleni impilo yabantu. Kodwa-ke, kubalulekile ukuqaphela ukuthi ama-ANN kufanele asize ekuthathweni kwezinqumo zezokwelapha kunokuthatha indawo yobuchwepheshe nokwahlulela kochwepheshe bezokwelapha abaqeqeshiwe.

Izinhlelo zokusebenza zamanethiwekhi e-neural ajulile

Izinhlelo zokusebenza ezibanzi zamanethiwekhi e-neural ajulile zingafaka:

• Ama-algorithms aya ngokuya aba yinkimbinkimbi ngamasethi edatha ayinkimbinkimbi kakhulu kanye nobuchwepheshe obungcono, okuholela emisebenzini yezinga eliphezulu efana nokuhlinzeka ngezinsizakalo zokubonisana nezeluleko zokutshala izimali. Ngo-2022, ama-algorithms anamandla alungele umthengi, njenge-Open AI's ChatGPT abonise amandla, ukuguquguquka, kanye nokusebenziseka kwesistimu ye-AI eqeqeshwe kumadathasethi amakhulu ngokwanele. (Abasebenzi bamakhola amhlophe emhlabeni wonke bathole ukuthuthumela okuhlangene.)
• Ubuhlakani bokwenziwa busetshenziswa ngokwandayo kwezempi ukuhlinzeka ngolwazi lwesikhathi sangempela kanye nobuhlakani bokusekela amasu empi.
• Amanethiwekhi ajulile e-neural avumela i-Metaverse ukuthi idale i-ecosystem yedijithali eyinkimbinkimbi ehlanganiswe ulwazi lwesikhathi sangempela olufana nezibalo zabantu, ukuziphatha kwamakhasimende, nezibikezelo zomnotho.
• Ama-ANN aqeqeshelwa ukubona amaphethini kudatha ayinkomba yomsebenzi wokukhwabanisa, futhi asetshenziselwa ukumaka imisebenzi esolisayo emikhakheni efana nezezimali ne-e-commerce.
• Amanethiwekhi ajulile we-neural asetshenziswa ukubona izinto, abantu, nezigcawu ezithombeni nakumavidiyo. Le ndlela isetshenziswa ezinhlelweni ezifana nezimoto ezizishayelayo, izinhlelo zokuphepha, kanye nokumaka ezinkundleni zokuxhumana.

Imibuzo okufanele icatshangelwe

• Ucabanga ukuthi amanethiwekhi ajulile we-neural azowushintsha kanjani umphakathi eminyakeni emithathu ezayo?
• Yiziphi izinselele ezingase zibe nobungozi?