
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 I-OpenAI   imigaqo yabo yokuzama nge-14 kaFebruwari ye  , umngeni owenziwe ngabaphengululi abali-11 abaphambili kwezibalo abavela kumaziko emfundo afana neHarvard, Yale, Stanford, neMIT. yanyatshile First Proof Umngeni yimiba eli-10 yezibalo ezingapapashwanga, ekumgangatho ophakamileyo wophando eyenzelwe ngokukodwa ukuba ingakwazi ukufikelela kwiimodeli zobukrekrele intlekele (AI). Injongo kukuqonda ukuba i-AI iyayenza na ngokwenene   xa isombulula ingxaki, kunokuba   indlela yokusombulula ingxaki yezibalo eyaziwayo. ukuqonda ikhumbula Izibonelelo zesi sivivinyo esinzima zazixubene, kodwa ziyamangalisa. Ngokweziphumo zengxelo ezivela kubachwepheshe, i-OpenAI ikholelwa ukuba ubuncinci ezintlanu zeemodeli zayo zokuzama ubufakazi (imiba 4, 5, 6, 9, kunye ne-10) zinethuba eliphezulu lokuba zilungile, kwaye ezinye izakubuyekezwa. Inzululwazi ephambili uJakub Pachocki kamva waphucula eli nani laba lixhoba, ebhala wathi: "Ngokweziphumo zengxelo ezivela kubachwepheshe, sikholelwa ukuba ubuncinci izisombululo ezithandathu (2, 4, 5, 6, 9, 10) zinethuba eliphezulu lokuba zilungile, kwaye ezinye izakujongeka zinjengokuphumelelayo." Waye   ukuba umzamo "wawungumsebenzi webala owenziwa kwiveki ngamaqela enye yeemodeli esizilolayo, nto leyo eyenza ukuba imeko esiyisebenzisileyo ifune ukulungiswa." wayongeza https://x.com/merettm/status/2022517085193277874?s=20&embedable=true Nangona kunjalo, abaphengululi kwezibalo sele bekhankanyile imingxuma kweminye yale mithandathu, kwaye isigqibo sokugqibela siya kuxhomekeka kuhlolisiso olusemthethweni loluntu. Akukho xesha limiselweyo loluhlolisiso loluntu olwaziswe esidlangalaleni. Umzamo wawuyisiphumo somsebenzi wokuzikhethela wentsebenziswano, apho iimodeli zakutshanje ze-OpenAI zisebenze kunye nabaphengululi kwezibalo abanamava kwiindawo ezifanelekileyo. Kubalulekile, imigaqo ye-First Proof ifuna ukuba zonke izimvo zezibalo zivela kwi-AI ngokuzimeleyo; igalelo lomntu kwizinto zezibalo aluvumelekanga, nto leyo eyenza intsebenziswano kunye nabachwepheshe ibe yinto yokujongwa. Iqela le-First Proof ngokwalo lifumanise ukuba, ngaphandle kwezithintelo, iinkqubo ze-AI, kubandakanya i-GPT 5.2 Pro kunye ne-Gemini 3.0 Deepthink, ziye zavakalisa ngokuqinisekileyo izisombululo kuyo yonke imiba eli-10, kodwa ezimbini kuphela eziye zathathelwa ingqalelo kuhlolisiso lwengcali. Lo mzamo ungaphakathi kwi-OpenAI ezama ukwenza inkqubela phambili kwi-sayensi kunye ne-numerical reasoning. NgoJulayi 2025, enye yeemodeli zayo yafumana imbasa yegolide kwi   ngamanqaku angama-35 kwangama-42. International Mathematical Olympiad Kutshanje, i-GPT-5.2 yancedisa abaphandi ukuba basombulule ingxaki eyimfihlo yetheoretiki kwisitatistiki, apho indima yomntu yavelisela ukuqinisekiswa nokubhala endaweni yokwakhiwa kwezibalo. Yintoni iFirst Proof? I-First Proof yindawo yokulinganisa eyenziwe ngabaphengululi abali-11 kwezibalo, kubandakanya amagama adumileyo avela eHarvard, Yale, Columbia, kunye ne-EPFL, ukuvavanya ukuba i-AI ingakwazi na ukusombulula ngokuzimeleyo iintlobo zemiba ephuma ngokwemvelo kuphando lwezemfundo. Ngokungafani nezibalo zokhuphiswano (cinga imiba ye-Olympiad eneziphumo ezicacileyo zezibalo), ezi zizisombululo ezincinci, ezingumsebenzi ezifumaneka ngabaphengululi befuna iziphumo ezinkulu. Imiba iquka iindawo ezifana ne-algebraic topology, i-spectral graph theory, i-symplectic geometry, kunye ne-numerical linear algebra. Okubalulekileyo, akukho nalunye isisombululo esake sapapashwa kwi-intanethi ngaphambi komngeni, nto leyo eyathintela nalo lonke ithuba lokungcola kwedatha yoqeqesho lwe-AI. Izisombululo zabhalwa ngemfihlelo zaza zakhutshwa nge-13 kaFebruwari, zinika abathathi-nxaxheba iveki enye ukuba baphumelele izisombululo. Umngeni ufumana igama lawo kwisifaniso sokubhaka: ubungqina bokuqala yinkqubo yokuqala yokusondela kwenhlama, inyathelo lokuqala phambi kokuba nantoni na ifakwe kwimveliso yokugqibela. Isifundo esivela kwi-First Proof ayisiyokuba i-AI yasilela. Yile nto ibango lokuphumelela ingxaki yezibalo sele lenyuke kakhulu. 
 
 UJuan Pablo Aguirre Osorio, Imicama, The Sociable

This story contains new, firsthand information uncovered by the writer.

Follow us on Twitter 

Read My Stories

Le audio iveliswe ngolwimi lokuqala lwebali!

I-OpenAI Iyatyhala iinethiwekhi zeNeural ukuya kuthi ga kwinqanaba eliphezulu loBalo.

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps