Bonisa i-HN: 1v1 umdlalo wokubhala ii-LLMs ezisokola ngawo

Umda woMntu: Umdlalo olula okhuba oomatshini

IiModeli zeeLwimi ezinkulu ezifana ne-GPT-4 kunye noClaude ziye zasebenza ngokumangalisayo ekuveliseni ikhowudi, ekusombululeni iipuzzle ze-algorithmic ezintsonkothileyo, kunye nokulungisa ii-codebases zonke. Bangamaqabane aboyikekayo kuphuhliso. Kodwa kwenzeka ntoni xa ukhulula imiceli mngeni emikhulu kwaye ubonise iphazili ebonakala ilula, intloko ukuya-entloko? Iprojekthi entsha ethi "Bonisa i-HN" ​​ityhila indawo eyimfama enomdla: umdlalo wekhowudi we-1v1 apho intuition yomntu ihlala igqitha amandla e-AI ekrwada. Lo msantsa uqaqambisa inqaku elibalulekileyo kumashishini: ukugqwesa kokwenyani akukho malunga nokuzenzela imisebenzi nge-AI; imalunga nokudala iinkqubo ezikrelekrele ezisebenzisa amandla abantu kunye noomatshini. Amaqonga afana ne-Mewayz akhelwe phezu kwalo mgaqo uchanekileyo, ukudibanisa i-AI njengesixhobo esinamandla ngaphakathi kwesakhelo sokusebenza esiphakathi kwabantu.

Umdlalo: Idabi leWits kunye neNgqiqo

Umdlalo ngokwawo unenkohliso ngokuthe ngqo. Abadlali ababini-nokuba ngaba ngumntu, i-AI, okanye umxube-uboniswa ngegridi kunye nesethi yemithetho yokubeka amathokheni. Injongo kukuba ngumdlali wokugqibela okwaziyo ukwenza intshukumo esemthethweni. Nangona imithetho ilula ngokwaneleyo ukuba umntwana aqonde, iqhinga elivelayo linzima. Ifuna ukucinga iintshukumo ezininzi eziya phambili, ukulindela ukhetho lomchasi, kunye neqhinga lokuhlengahlengisa kubhabho. Oku kudityaniswa kokulula kunye nobunzulu beqhinga yinto edala isaqhwithi esigqibeleleyo kwiiLLMs. Bayakwazi ukucengceleza imigaqo ngokugqibeleleyo, kodwa bayasokola ngomdlalo oguquguqukayo, oguqukayo oza ngokwemvelo ebantwini.

Kutheni iiLLMs ziFalter kwi-Head-to-Head Setting

Xa zingquzulana okanye ngokuchasene nabantu, iiLLM zangoku zibonakalisa ubuthathaka obuthile kolu hlobo lomdlalo. Ukusebenza kwabo kukholisa ukuthotywa ngexesha lomdlalo, hayi ngenxa yokunqongophala kwamandla okubala, kodwa ngenxa yokunqongophala kwengqondo yobuchule bokwenene.

• Ukuthembela ngokuPheleleyo kuthelekiso lwePatheni: ii-LLMs ziqeqeshelwa iiseti zedatha ezinkulu, zizenza zigqwese ekufumaneni iipateni zeenkcukacha-manani. Nangona kunjalo, kwimeko yomdlalo wenoveli, bahlala besilela ukuqulunqa iqhinga elitsha lokwenene, endaweni yoko babuyela umva kwiintshukumo "ezikhangeleka" ezichanekileyo ngokusekelwe kuqeqesho lwabo.
• Ukungabikho kweSicwangciso soMdlalo oBambiseneyo: Ngoxa i-AI inokuvavanya intshukumo enye, inzima ukugcina isicwangciso esihambelanayo, sexesha elide kulo lonke umdlalo. Abadlali abangabantu banokubeka imigibe kwaye badlale umdlalo okhuselayo okanye ohlaselayo; i-AI idla ngokudlala ngokuqhubekayo, ishukuma ngokushukuma.
• Ukungakwazi ukwenza iModeli yoMchasi onamandla: Inxalenye ephambili yawo nawuphi na umdlalo we-1v1 kukuqonda kunye nokuxela kwangaphambili ingqondo yomchasi wakho. Ii-LLM zine "ithiyori yengqondo" engalunganga kwaye ayinakukwazi ukulinganisa ngokufanelekileyo ubuqhetseba bomntu ochasayo okanye enye i-AI yokutshintsha kwiqhinga.

"Ukubukela umdlalo we-AI kufana nokubukela i-grandmaster idlala i-novice oye wakhumbula iwaka leentshukumo zokuvula kodwa akayiqondi ingongoma yomdlalo. Inobuchule obuchwephesha kodwa ilahlekile ngokweqhinga. "

Iimpembelelo kubukrelekrele obongezelelweyo

Olu vavanyo alukho malunga nokungqina ukuba abantu "bakrelekrele" kune-AI. Endaweni yoko, ibonisa ngamandla ukuba ubukrelekrele buneenkalo ezininzi. Ii-LLMs zinobuchule obumangalisayo bokubala kunye nokuvelisa, kodwa abanalo ukucinga okucacileyo, okucwangcisiweyo, kunye nokuguquguqukayo okuchaza ukusonjululwa kwengxaki yomntu. Lo ngumbono obalulekileyo kumashishini abalekayo ukuba amkele i-AI. Ukutshintsha ngokulula abantu abathatha izigqibo nge-AI kunokukhokelela kwiinkqubo eziqinileyo, ezinqabileyo. Amandla okwenene aleleukwandiswa-ukusebenzisa i-AI ukusingatha ukuphakamisa i-data-heavy lifting ngelixa ugcina abantu kwi-loop yokulawula iqhinga kunye nokusombulula iingxaki .

Ukwakha iiNkqubo eziNcinci ngeMewayz

Apha kulapho ishishini lemodyuli le-OS elifana neMewayz lixabiseke kakhulu. IMewayz ayenzelwanga ukuba yi-AI ezimeleyo eqhuba inkampani yakho. Yenzelwe ukuba ibe ngumaleko weokhestra odibanisa ubuchule bomntu beqela lakho kunye nesakhono esinamandla sezixhobo ze-AI. Kanye njengokuba umdlali ongumntu unokubetha i-AI kulo mdlalo wekhowudi ngokusebenzisa i-intuition kunye neqhinga, ishishini elisebenzisa i-Mewayz linokufikelela kwiziphumo eziphezulu ngokulawula izixhobo ze-AI ngobukrelekrele. Iqonga likuvumela ukuba uzenzele ngokuphindaphindiweyo ukuhamba komsebenzi ngoncedo lwe-AI ngelixa uqinisekisa ukuba izigqibo ezintsonkothileyo, ezicwangcisiweyo ziyaziswa sisigwebo somntu. Kumalunga nokudala inkqubo ye-symbiotic apho icandelo ngalinye-umntu kunye nomatshini-wenza eyona nto liyenzayo, ekhokelela kumbutho omelele ngakumbi kwaye oguquguqukayo.

Imibuzo Ebuzwa Rhoqo

Umda woMntu: Umdlalo olula okhuba oomatshini

IiModeli zeeLwimi ezinkulu ezifana ne-GPT-4 kunye noClaude ziye zasebenza ngokumangalisayo ekuveliseni ikhowudi, ekusombululeni iipuzzle ze-algorithmic ezintsonkothileyo, kunye nokulungisa ii-codebases zonke. Bangamaqabane aboyikekayo kuphuhliso. Kodwa kwenzeka ntoni xa ukhulula imiceli mngeni emikhulu kwaye ubonise iphazili ebonakala ilula, intloko ukuya-entloko? Iprojekthi entsha ethi "Bonisa i-HN" ​​ityhila indawo eyimfama enomdla: umdlalo wekhowudi we-1v1 apho intuition yomntu ihlala igqitha amandla e-AI ekrwada. Lo msantsa uqaqambisa inqaku elibalulekileyo kumashishini: ukugqwesa kokwenyani akukho malunga nokuzenzela imisebenzi nge-AI; imalunga nokudala iinkqubo ezikrelekrele ezisebenzisa amandla abantu kunye noomatshini. Amaqonga afana ne-Mewayz akhelwe phezu kwalo mgaqo uchanekileyo, ukudibanisa i-AI njengesixhobo esinamandla ngaphakathi kwesakhelo sokusebenza esiphakathi kwabantu.

Umdlalo: Idabi leWits kunye neNgqiqo

Umdlalo ngokwawo unenkohliso ngokuthe ngqo. Abadlali ababini-nokuba ngaba ngumntu, i-AI, okanye umxube-uboniswa ngegridi kunye nesethi yemithetho yokubeka amathokheni. Injongo kukuba ngumdlali wokugqibela okwaziyo ukwenza intshukumo esemthethweni. Nangona imithetho ilula ngokwaneleyo ukuba umntwana aqonde, iqhinga elivelayo linzima. Ifuna ukucinga iintshukumo ezininzi eziya phambili, ukulindela ukhetho lomchasi, kunye neqhinga lokuhlengahlengisa kubhabho. Oku kudityaniswa kokulula kunye nobunzulu beqhinga yinto edala isaqhwithi esigqibeleleyo kwiiLLMs. Bayakwazi ukucengceleza imigaqo ngokugqibeleleyo, kodwa bayasokola ngomdlalo oguquguqukayo, oguqukayo oza ngokwemvelo ebantwini.

Kutheni iiLLMs ziFalter kwi-Head-to-Head Setting

Xa zingquzulana okanye ngokuchasene nabantu, iiLLM zangoku zibonakalisa ubuthathaka obuthile kolu hlobo lomdlalo. Ukusebenza kwabo kukholisa ukuthotywa ngexesha lomdlalo, hayi ngenxa yokunqongophala kwamandla okubala, kodwa ngenxa yokunqongophala kwengqondo yobuchule bokwenene.

Iimpembelelo kuBukrelekrele obongezelelweyo

Olu vavanyo alukho malunga nokungqina ukuba abantu "bakrelekrele" kune-AI. Endaweni yoko, ibonisa ngamandla ukuba ubukrelekrele buneenkalo ezininzi. Ii-LLMs zinobuchule obumangalisayo bokubala kunye nokuvelisa, kodwa abanalo ukucinga okucacileyo, okucwangcisiweyo, kunye nokuguquguqukayo okuchaza ukusonjululwa kwengxaki yomntu. Lo ngumbono obalulekileyo kumashishini abalekayo ukuba amkele i-AI. Ukutshintsha ngokulula abantu abathatha izigqibo nge-AI kunokukhokelela kwiinkqubo eziqinileyo, ezinqabileyo. Amandla okwenene alele ekwandiseni-ukusebenzisa i-AI ukusingatha ukunyusa i-data-heavy lifting ngelixa ugcina abantu kwi-loop yokulawula isicwangciso kunye nokusombulula iingxaki .

Ukwakha iiNkqubo eziNcinci ngeMewayz

Apha kulapho ishishini lemodyuli le-OS elifana neMewayz lixabiseke kakhulu. IMewayz ayenzelwanga ukuba yi-AI ezimeleyo eqhuba inkampani yakho. Yenzelwe ukuba ibe ngumaleko weokhestra odibanisa ubuchule bomntu beqela lakho kunye nesakhono esinamandla sezixhobo ze-AI. Kanye njengokuba umdlali ongumntu unokubetha i-AI kulo mdlalo wekhowudi ngokusebenzisa i-intuition kunye neqhinga, ishishini elisebenzisa i-Mewayz linokufikelela kwiziphumo eziphezulu ngokulawula izixhobo ze-AI ngobukrelekrele. Iqonga likuvumela ukuba uzenzele ngokuphindaphindiweyo ukuhamba komsebenzi ngoncedo lwe-AI ngelixa uqinisekisa ukuba izigqibo ezintsonkothileyo, ezicwangcisiweyo ziyaziswa sisigwebo somntu. Kumalunga nokudala inkqubo ye-symbiotic apho icandelo ngalinye-umntu kunye nomatshini-wenza eyona nto liyenzayo, ekhokelela kumbutho omelele ngakumbi kwaye oguquguqukayo.