Ukuqala Ukuzama Iqhinga Elisha Lokuthuthukisa I-AI Yokutholwa Kwesayensi
Ukuqala Ukuzama Iqhinga Elisha Lokuthuthukisa I-AI Yokutholwa Kwesayensi
Ukuthuthukisa ubuhlakani bokwenziwa bokutholwa kwesayensi sekuwumgomo oyingqopha-mlando weziqhwaga zobuchwepheshe. Izinkampani ezifana ne-OpenAI ne-Anthropic zithole amashumi ezigidigidi zoxhaso ngezithembiso zempumelelo ye-AI kwezokwelapha, i-biology, kanye ne-physics. Kodwa-ke, ukutholwa kwesayensi kweqiniso okuqhutshwa yi-AI kuhlala kungaqondakali, njengoba kuboniswe izehlakalo ezedlule njengokuthonywa kwezibalo okukhiqizwe yi-ChatGPT. Inselele enkulu, ngokochwepheshe, ukuthi izinhlobo zezilimi ezinkulu zamanje (LLMs) azinawo amandla angaphakathi okukhiqiza ulwazi olusha lwesayensi ngokuzimela.
Kungani Amalebhu Amakhulu E-AI Ezabalaza Nokutholwa Kwesayensi
U-Markus Buehler, uprofesa wobunjiniyela we-MIT, uveza umkhawulo oyisisekelo ku-AI yanamuhla ethuthukisiwe. Uphikisa ngokuthi amamodeli anika amandla amasistimu avela ku-OpenAI ne-Anthropic awaklanyelwe ukutholwa kwangempela. Izakhiwo zabo zisekelwe ekubonweni kwephethini kusukela kudatha ekhona, hhayi ekudaleni imibono emisha noma okucatshangwayo.
Lokhu kufanekiselwa kahle ngekwindla edlule lapho okutholwayo kwezibalo okushiwo yi-ChatGPT kukhishwa ngokushesha. Isiqephu sigqamise igebe phakathi kwamandla okuhlaziya e-AI kanye nokucabanga kwayo kokudala, okugxile ekutholeni. Kuyinselelo esikhumbuza eminye imizamo ye-AI lapho ubuchwepheshe budonsa kanzima khona, kufana nokugxekwa ‘umlingisi’ we-AI u-Tilly Norwood ngokuntula ubuhlakani bangempela.
Inkinga Eyinhloko Ngamamodeli Amanje E-AI
Amamodeli olimi amakhulu ahamba phambili ekucubunguleni nasekubuyiseleni ulwazi. Bangakwazi ukufingqa imibhalo, baphendule imibuzo, futhi babhale nekhodi ngokusekelwe kudatha yabo yokuqeqeshwa. Nokho, basebenza ngaphakathi kwemingcele yalokho asebekufundile kakade.
Ukutholwa kwesayensi, ngokwemvelo yakho, kudinga ukungena kokungaziwa. Kubandakanya ukwakha ukuxhumana okusha phakathi kwezinkambu ezihlukene kanye nokuphakamisa imibono engekho kunoma iyiphi idathasethi yokuqeqeshwa. Lokhu ukugxuma i-AI yamanje ekhiqizayo, egxile ekudalweni kokuqukethwe nokuzenzakalela, ayakhelwanga ukuzenza. Imboni iyathuthuka, njengoba kubonakala ngentuthuko efana nesibuyekezo se-WordPress Gutenberg esibeka isisekelo sokushicilelwa kwe-AI, kodwa inselelo eyinhloko yokutholwa isekhona.
Sethula Amalebhu Angenangqondo: Indlela Entsha ye-AI Yesayensi
Ukuze kubhekwane naleli gebe, uSolwazi Buehler wasungula i-Unreasonable Labs kanye no-Yuan Cao, owayengusosayensi wocwaningo lwezisebenzi eziphezulu kwa-Google DeepMind. Ukuqaliswa kuhlose ukuphayona indlela ehluke kakhulu ekuthuthukiseni i-AI ukuze kutholwe isayensi. Esikhundleni sokuthembela kuphela ekufakweni kwedatha okukhulu, bakha amasistimu akwazi ukucabanga ngemikhakha eyahlukene.
Ama-Labs Angenangqondo asanda kuthola amaRandi ayizigidi ezingu-13.5 emzuliswaneni wokuxhasa ngemali oholwa yi-Playground Global. Umjikelezo ubone iqhaza elivela ku-AIX Ventures, E14 Fund, kanye ne-MS&AD Ventures. Lokhu kutshalwa kwezimali okubalulekile kugcizelela inkolelo yemakethe endleleni yabo yenoveli.
Ukufunda kokuthi "Aha" Izikhathi Emlandweni Wesayensi
I-hypothesis ka-Buehler ukuthi izinto eziningi ezinhle ezitholakele zivela ezikhathini ze-"aha". Lezi yizimo lapho usosayensi esebenzisa ithiyori noma umqondo osuka endaweni eyodwa ukuze axazulule inkinga esizindeni esihluke ngokuphelele. Lokhu kuhlukahlukana kwemibono kuyisihluthulelo sempumelelo.
Isibonelo sakudala umsebenzi ka-John Hopfield wango-1982. Wasebenzisa imiqondo esuka ku-condensed matter physics ukuya emkhakheni osanda kuqala ngaleso sikhathi wobuhlakani bokwenziwa. Lokhu kwaholela ekuthuthukisweni kwamanethiwekhi e-Hopfield, uhlobo lwenethiwekhi ye-neural ekwazi ukufunda nokukhumbula izinkumbulo. Kwakuwumbono woguquko owazalwa ngokuxhumanisa iziyalo ezingahlobene.
Ihluke kanjani I-AI Yamalebhu Angenangqondo Kumamodeli Avamile
I-AI ethuthukiswayo kuma-Unreasonable Labs yakhelwe ukulingisa leli khono lomuntu lokuqonda kwezinhlaka ezahlukene. Umgomo wabo awukona ukwakha imodeli yolimi enkudlwana kodwa ukwakha uhlelo olungabonisana nezizinda zesayensi.
I-Interdisciplinary Knowledge Graphs: Esikhundleni sokuqeqeshwa ngombhalo wodwa, i-AI yabo ihlanganisa ulwazi oluhlelekile olusuka emikhakheni eminingi yesayensi, kusukela kubhayoloji kuya ku-physics. I-Analogical Reasoning Engines: Ubuchwepheshe obuyinhloko bugxile ekutholeni ama-analogies nokufana phakathi kwemiqondo ebonakala ingahlobene, umshayeli oyinhloko wokuqamba okusha kwesayensi. I-Hypothesis Generation: Uhlelo luklanyelwe ukuphakamisa imibono yesayensi engahlolwa, hhayi nje ukuhlaziya idatha ekhona.
Le ndlela imele ukusuka okuphawulekayo kumasu okuthola amakhudlwanaamafemu ezobuchwepheshe, afana nokutholwa kwe-Zendesk kwe-AI yokuqala ye-Forethought, evame ukugxila ekucwengiseni izinhlelo zokusebenza ezikhona zesevisi yamakhasimende kunokuvula izindlela ezintsha zokutholwa.
Ikusasa Lokutholwa Okuqhutshwa yi-AI
Uma kuphumelela, ubuchwepheshe Bamalebhu Angenangqondo bungasheshisa ucwaningo ezindaweni ezibucayi. Cabanga nge-AI engaphakamisa inhlanganisela yezidakamizwa entsha ngokuhlanganisa izimiso ezivela kumakhemikhali kanye nezakhi zofuzo. Noma imodeli ephakamisa into entsha yamandla asimeme ngokuxhumanisa imiqondo evela ku-nanotechnology kanye ne-thermodynamics.
Izicelo ezingase zibe khona ziningi, kusukela ekusheshiseni ucwaningo lwezokwelapha kuya ekuxazululeni izinselele eziyinkimbinkimbi zemvelo. Lokhu kumelela umngcele olandelayo we-AI, udlulela ngale kwe-automation ukuze ube uzakwethu weqiniso ekuhlakanipheni kwabantu.
Isiphetho: Igagasi Elilandelayo Le-AI Innovation
Umjaho wokuthuthukisa i-AI yokutholwa kwesayensi uyakhula, kodwa impumelelo yangempela ingase ilele ngeziqalo ezikhethekile ezifana namalebhu angenangqondo. Ukugxila kwabo ekucabangeni kwemikhakha eyahlukene kunikeza indlela ethembisayo ngale kwemikhawulo yamamodeli wezilimi ezinkulu zamanje. Uhambo lokudala i-AI engathola ngempela lusanda kuqala.
Hlala unolwazi ngezinto ezintsha zakamuva ku-AI nobuchwepheshe. Ukuze uthole imininingwane eyengeziwe kanye nokwabelana kalula ngalesi sihloko, dala ikhasi lakho lamahhala lesixhumanisi ku-bio kokuthi Seemless ukuze uhlele okuqukethwe kwakho okuthandayo.
