Ucwaningo lwesayensi lwe-AI: Inhloso yangempela yokufunda ngomshini

• About the Author:
• Igama lomlobi

  I-Quantumrun Foresight
• Kwangathi 11, 2023

Ukuthuthukiswa kwemibono ngokwesiko bekubhekwa njengomsebenzi womuntu kuphela, njengoba kudinga ubuhlakani, ukuqondisisa, nokucabanga okujulile. Kodwa-ke, ngokuthuthuka kwezobuchwepheshe, ososayensi baya ngokuya bephendukela ekufundeni komshini (ML) ukuze bakhiqize izinto ezintsha ezitholakele. Ama-algorithms angakwazi ukuhlaziya amanani amakhulu edatha ngokushesha futhi akhombe amaphethini abantu okungenzeka bangakwazi ukuwabona.

Context

Kunokuba bancike emibonweni yabantu, abacwaningi bakhe ama-algorithms we-neural network ML ngomklamo ogqugquzelwe ubuchopho bomuntu, ophakamisa imibono emisha esekelwe kumaphethini wedatha. Ngenxa yalokho, izindawo eziningi zingase ziphendukele ku-ML ngokushesha ukuze kusheshiswe ukutholwa kwesayensi nokunciphisa ukuchema kwabantu. Endabeni yezinto zebhethri ezingahloliwe, ososayensi ngokwesiko bathembele kumasu okusesha egciniwe, ukumodela, nomuzwa wabo wamakhemikhali ukuze bakhombe ama-molecule asebenzayo. Ithimba elivela eNyuvesi yase-Liverpool ezinze e-UK liqashe i-ML ukuze lenze inqubo yokudala ibe lula. 

Okokuqala, abacwaningi badale inethiwekhi ye-neural ebeka phambili inhlanganisela yamakhemikhali ngokusekelwe emathubeni abo okukhiqiza into entsha ebalulekile. Ososayensi babe sebesebenzisa lezi zilinganiso ukuze baqondise izifundo zabo zaselabhorethri. Ngenxa yalokho, bathole izinketho ezine zezinto zebhethri ezisebenzayo ngaphandle kokuhlola yonke into esohlwini lwabo, zibasindise izinyanga zokuzama nephutha. Izinto ezisetshenziswayo ezintsha akuzona kuphela inkambu lapho i-ML ingasiza khona ucwaningo. Abacwaningi futhi basebenzisa amanethiwekhi e-neural ukuze baxazulule ukukhathazeka okubaluleke kakhulu kwezobuchwepheshe kanye nethiyori. Isibonelo, isazi sefiziksi eZurich's Institute for Theoretical Physics, u-Renato Renner, uthemba ukuthuthukisa incazelo ehlangene yokuthi umhlaba usebenza kanjani usebenzisa i-ML. 

Ukwengeza, amamodeli akhiqizayo e-AI athuthuke kakhulu njenge-OpenAI's ChatGPT avumela abacwaningi ukuthi bakhiqize ngokuzenzakalela idatha entsha, amamodeli, kanye nokuqagela. Lokhu kufezwa kufezwa ngamasu anjengamanethiwekhi adversarial akhiqizayo (ama-GAN), ama-autoencoder ahlukile (ama-VAE), namamodeli olimi asekelwe ku-transformer (njenge-Generative Pre-trained Transformer-3 noma i-GPT-3). Lawa mamodeli e-AI angasetshenziselwa ukukhiqiza amasethi edatha yokwenziwa, ukuklama nokuthuthukisa izakhiwo ezintsha ze-ML, futhi kuthuthukiswe imibono emisha yesayensi ngokuhlonza amaphethini nobudlelwano kudatha ebingaziwa ngaphambilini.

Umthelela ophazamisayo

Ososayensi bangase basebenzise i-AI ekhiqizayo ukusiza ngocwaningo. Ngekhono lokuhlaziya amaphethini nokubikezela imiphumela ngokusekelwe kulolo lwazi, lawa mamodeli angase axazulule imibono eyinkimbinkimbi yesayensi ehlale ingaxazululiwe isintu. Lokhu ngeke nje konge isikhathi nemali, kodwa kuzosiza nokuqonda komuntu isayensi ukuthi kudlulele ngale kwemingcele yayo yamanje. 

Ibhizinisi locwaningo nentuthuko (i-R&D) cishe lizothola kulula ukuqoqa uxhaso olufanele ngoba i-ML ingacubungula idatha ngokushesha. Ngenxa yalokho, ososayensi bazofuna usizo olwengeziwe ngokuqasha izisebenzi ezintsha noma ukusebenzisana namabhizinisi nezinkampani ezaziwayo ukuze bakhiqize imiphumela engcono. Uwonke umthelela wale ntshisekelo uzoba muhle, hhayi nje ekuthuthukisweni kwesayensi kuphela kodwa nakochwepheshe ngaphakathi kwemikhakha yesayensi. 

Kodwa-ke, okungase kube isivimbamgwaqo ukuthi izixazululo ezivela kulawa mamodeli aguquguqukayo zivame ukuba yinselele kubantu ukuba baziqonde, ikakhulukazi ukucabanga okuhilelekile. Ngenxa yokuthi imishini inikeza izimpendulo kuphela futhi ingasichazi isizathu sesixazululo, ososayensi bangase bahlale bengaqiniseki ngenqubo nesiphetho. Lokhu kungabonakali kwenza buthaka ukuzethemba emiphumeleni futhi kunciphisa inani lamanethiwekhi e-neural angasiza ekuhlaziyeni. Ngakho-ke, kuyodingeka ukuthi abacwaningi bakhe imodeli engazichaza yona.

Imithelela yocwaningo lwesayensi lwe-AI

Imithelela ebanzi yocwaningo lwesayensi ye-AI ingafaka:

• Izinguquko kumazinga okuqanjwa kwamaphepha ocwaningo, okuhlanganisa nokunikeza isikweletu sempahla yengqondo ku-AI. Ngokufanayo, amasistimu e-AI ngolunye usuku azoklonyeliswa njengabangase bazuze uMklomelo KaNobel, okungase kubangele izimpikiswano ezishubile mayelana nokuthi lawa ma-algorithms kufanele avunywe njengabaqambi.
• Ucwaningo olukhiqizwe yi-AI lungase luholele ezinhlotsheni ezintsha zesibopho sezomthetho kanye neminye imibuzo yezomthetho neyokuziphatha ehlobene nokusebenzisa i-AI nezinhlelo ezizimele ekutholweni kwesayensi.
• Ososayensi abasebenza ngamathuluzi ahlukahlukene e-AI akhiqizayo ukusheshisa intuthuko nokuhlolwa kwezokwelapha.
• Ukwenyuka kokusetshenziswa kwamandla okubangelwa amandla aphezulu ekhompuyutha adingekayo ukuze kusetshenziswe lawa ma-algorithms anemininingwane.
• Ososayensi besikhathi esizayo baqeqeshelwa ukusebenzisa i-AI namanye amathuluzi e-ML ekuhambeni komsebenzi wabo.
• Ohulumeni abakha amazinga omhlaba mayelana nemikhawulo kanye nezidingo zokwenza izivivinyo zesayensi ezikhiqizwe yi-AI.

Imibuzo okufanele icatshangelwe

• Uma ungusosayensi, isikhungo sakho noma ilabhorethri yakho ihlela kanjani ukufaka ucwaningo olusizwa yi-AI?
• Ucabanga ukuthi ucwaningo olwenziwa yi-AI luzoba nomthelela kanjani emakethe yemisebenzi yososayensi nabacwaningi?