{"id":160836,"date":"2026-02-13T11:12:14","date_gmt":"2026-02-13T09:12:18","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/google-predstavila-gemini-3-deep-think-moshhnyj-ii-dlya-nauki-kotoryj-vsyo-eshhyo-podgonyaet-zadachi-pod-otvet"},"modified":"2026-02-13T13:33:11","modified_gmt":"2026-02-13T11:33:11","slug":"google-predstavila-gemini-3-deep-think-moshhnyj-ii-dlya-nauki-kotoryj-vsyo-eshhyo-podgonyaet-zadachi-pod-otvet","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/google-predstavila-gemini-3-deep-think-moshhnyj-ii-dlya-nauki-kotoryj-vsyo-eshhyo-podgonyaet-zadachi-pod-otvet","title":{"rendered":"Google prezantoi Gemini 3 Deep Think \u2014 nj\u00eb AI t\u00eb fuqish\u00ebm p\u00ebr shkenc\u00ebn, i cili ende \"p\u00ebrshtat\" detyrat me p\u00ebrgjigjen","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Kompania Google l\u00ebshoi nj\u00eb p\u00ebrdit\u00ebsim t\u00eb madh Gemini 3 Deep Think<\/a> \u2014 nj\u00eb platform\u00eb e bazuar n\u00eb inteligjenc\u00ebn artificiale q\u00eb mendon, e cila ka p\u00ebr q\u00ebllim t\u00eb zgjidh\u00eb problemet aktuale n\u00eb fush\u00ebn e shkenc\u00ebs dhe inxhinieris\u00eb.<\/p>\n

<\/center>\"Google<\/p>\n

Zhvillimi i versionit t\u00eb ri Google Gemini 3 Deep Think \u00ebsht\u00eb realizuar n\u00eb bashk\u00ebpunim me shkenc\u00ebtar\u00ebt \u2014 sh\u00ebrbimi \u00ebsht\u00eb krijuar p\u00ebr t\u00eb zgjidhur probleme t\u00eb nd\u00ebrlikuara k\u00ebrkimore, t\u00eb cilat mund t\u00eb ken\u00eb aspekte pa kufij t\u00eb qart\u00eb ose nj\u00eb zgjidhje t\u00eb vetme t\u00eb sakt\u00eb, dhe t\u00eb dh\u00ebnat p\u00ebrkat\u00ebse mund t\u00eb jen\u00eb jopjesore ose t\u00eb \u00e7rregullta. Platforma e p\u00ebrdit\u00ebsuar Deep Think p\u00ebrfaqson nj\u00eb p\u00ebrpjekje p\u00ebr t\u00eb kaluar p\u00ebrtej teoris\u00eb abstrakte dhe p\u00ebr t'u futur n\u00eb fush\u00ebn e aplikimit praktik. Rezhimi i ri \u00ebsht\u00eb i disponuesh\u00ebm direkt n\u00eb aplikacionin Gemini, por mund ta shfryt\u00ebzojn\u00eb vet\u00ebm abonet e Google AI Ultra; kompania gjithashtu do ta hap\u00eb p\u00ebr her\u00eb t\u00eb par\u00eb aksesin n\u00eb sh\u00ebrbimin p\u00ebrmes API Gemini \u2014 inxhinier\u00ebt, hulumtuesit dhe nd\u00ebrmarrjet do t'u duhet t\u00eb dor\u00ebzojn\u00eb nj\u00eb k\u00ebrkes\u00eb p\u00ebrkat\u00ebse.<\/p>\n

N\u00eb provimin Humanity\u2019s Last Exam, platforma Google Gemini 3 Deep Think arriti 48.4% pa p\u00ebrdorimin e mjeteve t\u00eb jashtme; rezultati n\u00eb benchmarkun ARC-AGI-2 ishte nj\u00eb 84.6% pa precedent. N\u00eb testin e programimit Codeforces, ratingu Elo arriti 3455; n\u00eb detyrat e Olimpiad\u00ebs Nd\u00ebrkomb\u00ebtare t\u00eb Matemat\u00ebs 2025, sistemi tregoi nivelin e medaljes s\u00eb art\u00eb, t\u00eb nj\u00ebjtin rezultat arriti edhe n\u00eb zgjidhjen e detyrave t\u00eb olimpiadave n\u00eb kimi dhe fizik\u00eb. N\u00eb testin CMT-Benchmark, rezhimi Google Gemini 3 Deep Think demonstroi nj\u00eb nivel t\u00eb lart\u00eb t\u00eb njohurive n\u00eb fizik\u00ebn teorike me nj\u00eb rezultat prej 50.5%.<\/p>\n

<\/center>\"Google<\/p>\n

Modeli Gemini 3 Deep Think u shnd\u00ebrrua n\u00eb baz\u00ebn e agjentit t\u00eb AI, i cili n\u00eb laboratorin Google DeepMind mori emrin Aletheia<\/a>. Ai p\u00ebrfshin nj\u00eb mjet p\u00ebr t\u00eb verifikuar hipotezat, t\u00eb shqiptuara n\u00eb gjuh\u00ebn natyrore, p\u00ebr t\u00eb identifikuar t\u00eb metat n\u00eb zgjidhjet e propozuara; parashikohet nj\u00eb proces iterativ i gjenerimit dhe rregullimit t\u00eb zgjidhjeve. E r\u00ebnd\u00ebsishme, agjenti mund t\u00eb pranoj\u00eb paaft\u00ebsin\u00eb e tij p\u00ebr t'u p\u00ebrgjigjur k\u00ebrkes\u00ebs. P\u00ebr t\u00eb trajtuar hulumtimet komplekse, ai i drejtohet sh\u00ebrbimit t\u00eb k\u00ebrkimit Google dhe mjeteve t\u00eb navigimit t\u00eb uebit \u2014 kur p\u00ebrgatit p\u00ebrmbledhje p\u00ebr literatur\u00ebn e publikuar, ai nuk jep lidhje joekzistuese dhe p\u00ebrpiqet t\u00eb shmang\u00eb pasakt\u00ebsit\u00eb n\u00eb llogaritje.<\/p>\n

Niveli aktual dhe perspektiv t\u00eb arritjeve t\u00eb Aletheia zhvilluesit e Google e shp\u00ebrndan\u00eb n\u00eb pes\u00eb grada. N\u00eb nj\u00eb m\u00ebnyr\u00eb pothuajse t\u00eb plot\u00eb autonome, agjenti AI p\u00ebrcaktoi rrug\u00ebt p\u00ebr zgjidhjen e tre problemeve t\u00eb hapura, t\u00eb p\u00ebrformuluara nga matematicienti P\u00e1l Erd\u0151s \u2014 ky k\u00ebrkim u vler\u00ebsua n\u00eb nivelin zero, i cili korrespondon me \"novitet t\u00eb pa r\u00ebnd\u00ebsish\u00ebm\". N\u00eb t\u00eb nj\u00ebjt\u00ebn m\u00ebnyr\u00eb, ai propozoi nj\u00eb zgjidhje p\u00ebr nj\u00eb problem tjet\u00ebr nga e nj\u00ebjta grup, duke arritur nivelin e par\u00eb ose \"novitet minimal\". N\u00eb nivelin e dyt\u00eb, \"t\u00eb p\u00ebrshtatshme p\u00ebr botim\", Aletheia tregoi rezultate n\u00eb m\u00ebnyr\u00eb autonome, n\u00eb m\u00ebnyr\u00eb bashk\u00ebpunimi me njeriun dhe n\u00eb m\u00ebnyr\u00eb si nj\u00eb mjet ndihm\u00ebs. Niveli i tret\u00eb (\"p\u00ebrparim t\u00eb r\u00ebnd\u00ebsish\u00ebm\") dhe i kat\u00ebrt (\"p\u00ebrparim t\u00eb sh\u00ebnuar\") akoma nuk i jan\u00eb n\u00ebnshtruar agjentit AI.<\/p>\n

Sipas detyr\u00ebs s\u00eb zhvilluesve, Aletheia analizoi 700 probleme matematikore t\u00eb pazgjidhura deri tani t\u00eb Erd\u00f6s dhe arriti t\u00eb zgjidhte 13 prej tyre. Megjithat\u00eb, p\u00ebr 9 nga k\u00ebto, ekzistonin zgjidhje tashm\u00eb t\u00eb njohura, nd\u00ebrsa vet\u00ebm 4 probleme duken si t\u00eb zgjidhura p\u00ebr her\u00eb t\u00eb par\u00eb. P\u00ebr m\u00eb tep\u00ebr, nga 212 zgjidhje q\u00eb ofroi AI, vet\u00ebm 6.5 % ishin 't\u00eb sakta n\u00eb p\u00ebrmbajtje'. T\u00eb tjerat ose kishin mang\u00ebsi themelore (68.5 %), ose zgjidhnin versione t\u00eb interpretuara gabimisht t\u00eb problemeve origjinale (31.5 %).<\/p>\n

Si rezultat, zhvilluesit pranuan se AI i tyre tregon ' nj\u00eb prirje p\u00ebr ta kuptuar gabimisht pyetjen n\u00eb nj\u00eb m\u00ebnyr\u00eb q\u00eb e b\u00ebn m\u00eb t\u00eb leht\u00eb p\u00ebrgjigjen' dhe mbetet 'jasht\u00ebzakonisht i prirur ndaj gabimeve krahasuar me njer\u00ebzit'. Me fjal\u00eb t\u00eb tjera, nuk \u00ebsht\u00eb ende n\u00eb gjendje t\u00eb z\u00ebvend\u00ebsoj\u00eb matematicien\u00ebt.<\/p>\n

Burimet:<\/strong><\/p>\n