Ukuqalisa Ukuzama iQhinga elitsha lokuphuhlisa i-AI yokuFumana iSayensi
Ukuqalisa Ukuzama iQhinga elitsha lokuphuhlisa i-AI yokuFumana iSayensi
Ukuphuhlisa ubukrelekrele bokwenziwa ukuze kufunyanwe isayensi kuye kwaba yinjongo ebalaseleyo kwiingxilimbela zetekhnoloji. Iinkampani ezifana ne-OpenAI kunye ne-Anthropic zikhusele amashumi eebhiliyoni kwinkxaso-mali ngezithembiso zempumelelo ye-AI kwezamayeza, ibhayoloji, kunye nefiziksi. Nangona kunjalo, ukufunyaniswa kokwenyani kwenzululwazi okuqhutywa yi-AI kuhlala kunzima, njengoko kubonisiwe kwiziganeko ezidlulileyo njengokufunyaniswa kwezibalo okwenziwe yiChatGPT. Owona mceli mngeni, ngokweengcali, kukuba imifuziselo yolwimi olukhulu yangoku (LLMs) ayinasakhono sangaphakathi sokuvelisa ulwazi oluyinoveli lwenzululwazi ngokuzimela.
Kutheni iiLabhu ezinkulu ze-AI zisokola ngokuFumana kwezeNzululwazi
UMarkus Buehler, unjingalwazi we-MIT wobunjineli, uchonga umda osisiseko kwi-AI yanamhlanje ephucukileyo. Uxoxa ukuba iimodeli zokunika amandla iinkqubo ezivela kwi-OpenAI kunye ne-Anthropic ayenzelwanga ukufunyanwa kokwenyani. Uyilo lwabo lusekelwe ekuqapheliseni ipateni kwidatha ekhoyo, kungekhona ekudaleni iithiyori ezintsha okanye iingcamango.
Oku kubonakaliswe ngokucacileyo ekwindla lokugqibela xa kufunyaniswe imathematika yi-ChatGPT yachithwa ngokukhawuleza. Isiqendu siqaqambise umsantsa phakathi kwamandla ocazululo e-AI kunye nokucinga kwayo, ejolise ekufumaneni. Ngumceli mngeni osikhumbuza ezinye iinzame ze-AI apho itekhnoloji isokola ngemvelaphi, kufana nokugxekwa ajongene ne-AI ‘umdlali’ uTilly Norwood ngokuswela ubuchule bokwenyani.
Ingxaki engundoqo kwiiModeli zangoku ze-AI
Iimodeli zeelwimi ezinkulu zigqwesa ekusetyenzweni nasekuphindaphindeni ulwazi. Bangakwazi ukushwankathela iitekisi, baphendule imibuzo, baze babhale ikhowudi ngokusekelwe kwidatha yabo yoqeqesho. Noko ke, basebenza ngaphakathi koko sele bekufundile.
Ukufunyaniswa kwenzululwazi, ngokwendalo yayo, kufuna ukunyathela kwinto engaziwayo. Kubandakanya ukwenza unxibelelwano olutsha phakathi kwemimandla engafaniyo kunye nokuphakamisa imibono engekho kuyo nayiphi na idatha yoqeqesho. Oku kukutsiba okukhoyo ngoku i-AI yokuvelisa, egxile ekudaleni umxholo kunye ne-automation, ayakhelwe ukwenza. Ishishini liyavela, njengoko kubonwa ngophuhliso olufana nohlaziyo lwe-WordPress Gutenberg lubeka isiseko sokupapashwa kwe-AI, kodwa umceli mngeni ophambili wokufunyanwa uhleli.
Ukwazisa iiLebhu eziNgenangqiqo: Indlela entsha ye-AI yeSayensi
Ukujongana nesi sikhewu, uNjingalwazi uBuehler waseka iiLebhu eziNgaqiqiyo kunye noYuan Cao, owayesakuba ngusonzululwazi wophando ophambili kuGoogle DeepMind. Ukuqaliswa kujolise ekubeni nguvulindlela kwindlela eyahluke kakhulu ekuphuhliseni i-AI ukuze kufunyanwe inzululwazi. Endaweni yokuxhomekeka kuphela ekufakeni idatha enkulu, bakha iinkqubo ezikwaziyo ukuqiqa ngeendlela ezahlukeneyo.
IiLabs ezingafanelekanga kutshanje zifumene i-13.5 yezigidi zeedola kumjikelo wenkxaso-mali okhokelwa yiPlayground Global. Umjikelo ubone inxaxheba evela kwi-AIX Ventures, E14 Fund, kunye ne-MS&AD Ventures. Olu tyalo-mali lubalulekileyo lugxininisa inkolelo yemarike kwindlela yabo yenoveli.
Ukufunda ku"Aha" Amaxesha kwiMbali yeSayensi
Ingcinga kaBuehler yeyokuba izinto ezininzi ezifunyenweyo zivela kumzuzu othi "aha". Ezi ziimeko apho inzululwazi isebenzisa ithiyori okanye ingcamango ukusuka kwintsimi enye ukusombulula ingxaki kwi-domain eyahlukileyo ngokupheleleyo. Oku kudluliselwa kwezimvo kungundoqo kwimpumelelo.
Umzekelo oqhelekileyo ngumsebenzi ka-John Hopfield ngo-1982. Wasebenzisa iikhonsepthi ezisuka kwi-condensed matter fiziksi ukuya kwindawo yolwazi olungeyonyani. Oku kukhokelele kuphuhliso lothungelwano lweHopfield, uhlobo lothungelwano lwe-neural olukwaziyo ukufunda nokukhumbula iinkumbulo. Yayiyingcamango yoguquko eyazalwa ngokudibanisa amacandelo anganxulumananga.
I-AI yeeLebhu ezingaqiqiyo ziyohluka njani kwiiModeli eziqhelekileyo
I-AI ephuhliswayo kwiiLebhu eziNgaqiqekiyo yenzelwe ukulinganisa obu buchule bomntu bokuqonda kwiinkalo ezahlukeneyo. Injongo yabo ayikokudala imodeli yolwimi olukhulu kodwa kukwakha inkqubo enokuthi iqiqe kwiinkalo zenzululwazi.
Iigrafu zoLwazi lwe-Interdisciplinary Knowledge: Endaweni yoqeqesho kwisicatshulwa sodwa, i-AI yabo idibanisa ulwazi olucwangcisiweyo oluvela kwiinkalo ezininzi zesayensi, ukusuka kwi-biology ukuya kwi-physics. I-Analogical Reasoning Engines: Ubuchwephesha obuphambili bujolise ekufumaneni ukulinganisa kunye nokuhambelana phakathi kweengcamango ezibonakala zingahambelani, umqhubi ophambili wokutsha kwesayensi. I-Hypothesis Generation: Inkqubo yenzelwe ukucebisa iingqikelelo zenzululwazi ezinokuvavanywa, hayi nje ukuhlalutya idatha ekhoyo.
Le ndlela imele ukushenxa okubalulekileyo kwizicwangciso zokufumana ezinkuluiifemu zobuchwepheshe, ezifana nokufunyanwa kwe-Zendesk ye-AI yokuQala i-AI, ehlala igxile ekuphuculeni izicelo zenkonzo yabathengi esele ikhona kunokuba ibe nguvulindlela kwiindlela ezintsha zokufumanisa.
Ikamva loFundo oluqhutywa yi-AI
Ukuba iphumelele, itekhnoloji yeeLebhu eziNgaqiqiyo inokukhawulezisa uphando kwiindawo ezibalulekileyo. Khawufane ucinge nge-AI enokuthi iphakamise ikhompawundi entsha yechiza ngokudibanisa imigaqo evela kwikhemikhali kunye nemfuzo. Okanye imodeli ephakamisa into entsha yamandla azinzileyo ngokudibanisa iingcamango ezivela kwi-nanotechnology kunye ne-thermodynamics.
Izicelo ezinokuthi zibekho zininzi, ukusuka ekukhawuleziseni uphando lwezonyango ukuya ekusombululeni imingeni enzima yokusingqongileyo. Oku kubonisa umda olandelayo we-AI, uhamba ngaphaya kwe-automation ukuze ube liqabane lokwenyani kubukrelekrele bomntu.
Isiphelo: I-Wave elandelayo ye-AI Innovation
Umdyarho wokuphuhlisa i-AI yokufunyaniswa kwesayensi uyafudumala, kodwa impumelelo yokwenyani inokufumaneka kuqalo olukhethekileyo olufana neeLabhu eziNgaqiqiyo. Ugqaliselo lwabo ekuqiqeni ngeendlela ezahlukeneyo lunika indlela ethembisayo engaphaya kwemida yemizekelo yolwimi olukhulu lwangoku. Uhambo lokuya ekudaleni i-AI enokufumanisa ngokwenene luyaqala.
Hlala uhlaziywa kwizinto zamva nje ze-AI kunye netekhnoloji. Ukufumana ulwazi oluthe kratya kunye nokwabelana ngokulula ngeli nqaku, yenza ikhonkco lakho lasimahla kwi-bio iphepha kwi-Seemless ukucutha umxholo owuthandayo.
