{"id":124,"date":"2024-03-04T15:26:09","date_gmt":"2024-03-04T15:26:09","guid":{"rendered":"https:\/\/ai4debunk.eu\/?page_id=124"},"modified":"2026-01-08T19:26:29","modified_gmt":"2026-01-08T19:26:29","slug":"scientific-publications","status":"publish","type":"page","link":"https:\/\/ai4debunk.eu\/lv\/scientific-publications\/","title":{"rendered":"Zin\u0101tnisk\u0101s publik\u0101cijas"},"content":{"rendered":"
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Zin\u0101tniskie p\u0113t\u012bjumi Publik\u0101cijas<\/span><\/h1>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\tAtrodiet visas AI4Debunk partneru sagatavot\u0101s zin\u0101tnisk\u0101s publik\u0101cijas, kur\u0101s izkl\u0101st\u012bti projekta jaun\u0101kie zin\u0101tniskie konstat\u0113jumi.\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t\t
\n\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tGaismas izklied\u0113\u0161ana lielos \u0123enerat\u012bvos t\u012bklos: Epist\u0113tisk\u0101s nenoteikt\u012bbas nov\u0113rt\u0113\u0161ana dif\u016bzijas mode\u013cos<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Axel Brando, Lucas Berry, David Mege<\/p>

Kopsavilkums:<\/strong>
\u0122enerat\u012bvie dif\u016bzijas mode\u013ci, kam rakstur\u012bgs liels parametru skaits (vair\u0101k nek\u0101 100 miljoni) un darb\u012bba augstdimension\u0101l\u0101s att\u0113lu telp\u0101s, rada iev\u0113rojamas probl\u0113mas tradicion\u0101laj\u0101m nenoteikt\u012bbas apl\u0113\u0161u metod\u0113m skait\u013co\u0161anas pras\u012bbu d\u0113\u013c. \u0160aj\u0101 darb\u0101 m\u0113s ievie\u0161am inovat\u012bvu sist\u0113mu Diffusion Ensembles for Capturing Uncertainty (DECU), kas paredz\u0113ta epistemisk\u0101s nenoteikt\u012bbas apl\u0113\u0161anai dif\u016bzijas mode\u013ciem. DECU sist\u0113ma ievie\u0161 jaunu metodi, kas efekt\u012bvi apm\u0101ca nosac\u012bt\u0101s dif\u016bzijas mode\u013cu ansamb\u013cus, iek\u013caujot statisku iepriek\u0161 apm\u0101c\u012btu parametru kopumu, krasi samazinot skait\u013co\u0161anas slogu un to parametru skaitu, kuriem nepiecie\u0161ama apm\u0101c\u012bba. Turkl\u0101t DECU izmanto pairwise-Distance Estimators (PaiDEs), lai prec\u012bzi izm\u0113r\u012btu epistemisko nenoteikt\u012bbu, nov\u0113rt\u0113jot savstarp\u0113jo inform\u0101ciju starp mode\u013ca izvad\u0113m un svariem augstdimension\u0101l\u0101s telp\u0101s. \u0160\u012bs sist\u0113mas efektivit\u0101ti pier\u0101da eksperimenti ar ImageNet datu kopu, uzsverot t\u0101s sp\u0113ju uztvert epist\u0113tisku nenoteikt\u012bbu, jo \u012bpa\u0161i nepietiekami atlas\u012bt\u0101s att\u0113lu klas\u0113s.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
Konference par neskaidr\u012bbu m\u0101ksl\u012bg\u0101 intelekta jom\u0101<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tSint\u0113tisko att\u0113lu detektoru notur\u012bba un visp\u0101rin\u0101\u0161ana<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Coccomini D.A., Caldelli R., Gennaro C., Fiameni G., Amato G., Falchi F.<\/p>

Kopsavilkums:<\/strong>
P\u0113d\u0113j\u0101 laik\u0101 m\u0101ksl\u012bg\u0101 intelekta tehnolo\u0123iju, jo \u012bpa\u0161i dzi\u013c\u0101s m\u0101c\u012b\u0161an\u0101s algoritmu, strauj\u0101 att\u012bst\u012bba ir \u013c\u0101vusi arvien vair\u0101k izplat\u012bt sint\u0113tiskos medijus, kas paz\u012bstami k\u0101 dzi\u013cviltojumi. Pieaugo\u0161\u0101s ba\u017eas par dzi\u013cviltojumu pieaugo\u0161o pieejam\u012bbu un ticam\u012bbu ir mudin\u0101ju\u0161as p\u0113tniekus koncentr\u0113ties uz meto\u017eu izstr\u0101di to atkl\u0101\u0161anai. \u0160aj\u0101 jom\u0101 ISTI CNR AIMH Lab p\u0113tnieki sadarb\u012bb\u0101 ar p\u0113tniekiem no cit\u0101m organiz\u0101cij\u0101m ir veiku\u0161i p\u0113t\u012bjumus, izmekl\u0113\u0161anu un projektus, lai pal\u012bdz\u0113tu apkarot \u0161o tendenci, p\u0113tot jaunus risin\u0101jumus un draudus. \u0160aj\u0101 rakst\u0101 apkopoti jaun\u0101kie centieni, ko \u0161aj\u0101 jom\u0101 veiku\u0161i m\u016bsu p\u0113tnieki sadarb\u012bb\u0101 ar cit\u0101m iest\u0101d\u0113m un ekspertiem.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
CEUR darbsemin\u0101ra proced\u016bras<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tMAD 24 darbsemin\u0101rs: Multivides m\u0101ksl\u012bgais intelekts pret dezinform\u0101ciju<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Stanciu, Cristian un Ionescu, Bogdan un Cuccovillo, Luca un Papadopoulos, Symeon un Kordopatis-Zilos, Giorgos un Popescu, Adrian un Caldelli, Roberto<\/p>

Kopsavilkums:<\/strong>
Sint\u0113tisko mediju veido\u0161ana un manipul\u0101cijas p\u0113d\u0113jos gados ir piedz\u012bvoju\u0161as straujus pav\u0113rsienus, padarot arvien viegl\u0101ku multivides satura rad\u012b\u0161anu, kas nav at\u0161\u0137irams no cilv\u0113ka nov\u0113rot\u0101ja. Turkl\u0101t \u0123ener\u0113to saturu var \u013caunpr\u0101t\u012bgi izmantot individu\u0101\u013ci un organiz\u0101cijas, lai izplat\u012btu dezinform\u0101ciju, kas rada b\u016btiskus draudus sabiedr\u012bbai un demokr\u0101tijai. T\u0101p\u0113c ir steidzami vajadz\u012bgi MI r\u012bki, kas v\u0113rsti uz to, lai atvieglotu savlaic\u012bgu un efekt\u012bvu mediju verifik\u0101cijas procesu. MAD\u201924 darbsemin\u0101ra m\u0113r\u0137is ir pulc\u0113t cilv\u0113kus ar da\u017e\u0101du izcelsmi, kuri ir velt\u012bti dezinform\u0101cijas apkaro\u0161anai multivid\u0113, izmantojot MI, veicinot vidi inovat\u012bvu ideju izp\u0113tei un pieredzes apmai\u0146ai. Intere\u0161u p\u0113tniec\u012bbas jomas ietver manipul\u0113ta vai \u0123ener\u0113ta satura identific\u0113\u0161anu, k\u0101 ar\u012b dezinform\u0101cijas izplat\u012b\u0161anas un t\u0101s soci\u0101lo seku izmekl\u0113\u0161anu. Atz\u012bstot multivides noz\u012bmi, darbsemin\u0101rs uzsver da\u017e\u0101du modalit\u0101\u0161u kop\u012bgu anal\u012bzi satur\u0101, jo verifik\u0101ciju var uzlabot, apkopojot vair\u0101kas satura formas.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
ACM 3. starptautisk\u0101 semin\u0101ra \u201cMultimediju m\u0101ksl\u012bgais intelekts pret dezinform\u0101ciju\u201d darbsemin\u0101rs<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tTempor\u0101l\u0101s virsmas r\u0101mja anom\u0101lijas dzi\u013cviltojumu video detekt\u0113\u0161anai<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Andrea Ciamarra, Roberto Caldelli, Alberto Del Bimbo<\/p>

Kopsavilkums:<\/strong>
Apl\u016bkojot video sec\u012bbu, kur\u0101 ir att\u0113lota priek\u0161pl\u0101na persona, vairs nav tik sen. Dzi\u013cviltojumi ir radik\u0101li main\u012bju\u0161i m\u016bsu veidu, k\u0101 skat\u012bties uz \u0161\u0101du saturu, un m\u016bsdien\u0101s m\u0113s bie\u017e\u0101k esam pieradu\u0161i br\u012bn\u012bties, vai tas, ko m\u0113s redzam, ir \u012bsts vai tikai mistifik\u0101cija. \u0160aj\u0101 visp\u0101r\u0113j\u0101s dezinform\u0101cijas kontekst\u0101 \u013coti gaid\u0101mi uzticami risin\u0101jumi, kas pal\u012bdz\u0113tu kop\u012bgajiem lietot\u0101jiem ne tikai nov\u0113rt\u0113t \u0161\u0101da veida videosekvences. \u0160aj\u0101 dokument\u0101 ir ieviesta jauna pieeja, kas izmanto laika virsmas r\u0101mja anom\u0101lijas, lai atkl\u0101tu dzi\u013cviltotus videoklipus. Ar \u0161o metodi tiek mekl\u0113tas iesp\u0113jam\u0101s nesakrit\u012bbas, ko izraisa dzi\u013cvilto\u0161anas manipul\u0101cijas, uztvert\u0101s ainas virsm\u0101s un to evol\u016bcij\u0101 pa laika asi. \u0160\u012bs funkcijas tiek izmantotas k\u0101 cauru\u013cvada ievade, pamatojoties uz dzi\u013ciem neironu t\u012bkliem, lai veiktu pa\u0161a video bin\u0101ro nov\u0113rt\u0113jumu. Eksperiment\u0101lie rezult\u0101ti liecina, ka \u0161\u0101da metodika var sasniegt iev\u0113rojamus rezult\u0101tus atkl\u0101\u0161anas precizit\u0101tes zi\u0146\u0101.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
2024. gada IEEE\/CVF konference par datorredzi un mode\u013cu atpaz\u012b\u0161anas darbsemin\u0101riem (CVPRW)<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tSoci\u0101lpolitiska pieeja dezinform\u0101cijai un m\u0101ksl\u012bgajam intelektam: Ba\u017eas, atbildes un probl\u0113mas<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Paskal\u012bna gaborita<\/p>

Kopsavilkums:<\/strong>
Vair\u0101k nek\u0101 desmit gadus starptautiskas organiz\u0101cijas dezinform\u0101ciju klasific\u0113 k\u0101 vienu no galvenajiem draudiem demokr\u0101tijai un iest\u0101d\u0113m. Digit\u0101l\u0101s tehnolo\u0123ijas no jauna izgudro un b\u016btiski p\u0101rveido m\u016bsdien\u012bgu dz\u012bvesveidu, iedz\u012bvot\u0101ju un uz\u0146\u0113m\u0113jdarb\u012bbas vidi. M\u0101ksl\u012bgais intelekts rada jaunus trauc\u0113jumus veid\u0101, k\u0101d\u0101 m\u0113s piek\u013c\u016bstam zin\u0101\u0161an\u0101m un veidojam, izplat\u0101m un izprotam inform\u0101ciju. Tas var ar\u012b padar\u012bt neskaidras robe\u017eas starp re\u0101lu inform\u0101ciju un manipul\u0113tu inform\u0101ciju, par\u0101doties \u201cviltus zi\u0146\u0101m\u201d, autom\u0101tiskaj\u0101m t\u012bklu savstarp\u0113j\u0101m atsauc\u0113m un \u201cdzi\u013cviltojumiem\u201d. MI sist\u0113mas palielina potenci\u0101lu rad\u012bt re\u0101listisku viltus saturu un m\u0113r\u0137tiec\u012bgas dezinform\u0101cijas kampa\u0146as. Dezinform\u0101cija p\u0101rsniedz vienk\u0101r\u0161as baumas, lai apzin\u0101ti maldin\u0101tu un izkrop\u013cotu uz pier\u0101d\u012bjumiem balst\u012btu inform\u0101ciju, izmantojot safabric\u0113tus datus. Eiropas iest\u0101des nesen piev\u0113rs\u0101s ar\u012b ar FIMI saist\u012btas dezinform\u0101cijas identific\u0113\u0161anai: \u0100rvalstu inform\u0101cijas manipul\u0101cija un iejauk\u0161an\u0101s. Rakst\u0101 ir nor\u0101d\u012btas tendences un ba\u017eas, kas saist\u012btas ar dezinform\u0101ciju un m\u0101ksl\u012bgo intelektu. Taj\u0101 p\u0113t\u012bta dezinform\u0101cijas uztvere, t\u0101s ietekme un atbildes pas\u0101kumi, tostarp ES M\u0101ksl\u012bg\u0101 intelekta akts un tie\u0161saistes platformu politika. Tas nodro\u0161ina pirmo anal\u012btisko pieeju \u0161im tematam, pamatojoties uz pa\u0161reiz\u0113j\u0101m p\u0113tnieku debat\u0113m, m\u016bsu 2024. gada aptauju pirmajiem konstat\u0113jumiem, intervij\u0101m un simtiem viltus zi\u0146u tie\u0161saist\u0113 anal\u012bzi. Taj\u0101 m\u0113\u0123in\u0101ts saprast, k\u0101 iedz\u012bvot\u0101ji un ieinteres\u0113t\u0101s personas uztver dezinform\u0101ciju, un apzin\u0101t iesp\u0113jamo ietekmi. Taj\u0101 analiz\u0113tas ar\u012b pa\u0161reiz\u0113j\u0101s probl\u0113mas un ierobe\u017eojumi, iesp\u0113jas un ierobe\u017eojumi c\u012b\u0146\u0101 pret manipul\u0101cij\u0101m un iejauk\u0161anos. Rakst\u0101 apl\u016bkoti pa\u0161reiz\u0113jie procesi un dezinform\u0101cijas ietekme (2), izkl\u0101st\u012bti galvenie konstat\u0113jumi m\u016bsu tie\u0161saistes aptauj\u0101 par dezinform\u0101cijas uztveri (3), pa\u0161reiz\u0113j\u0101s ES regulat\u012bv\u0101s atbildes (4) un diskusiju punkti (5). \u0160aj\u0101 rakst\u0101 m\u0113s apgalvojam, ka veids, k\u0101d\u0101 m\u0113s piek\u013c\u016bstam inform\u0101cijai, ir gigantiski main\u012bjies, bet reakcija uz dezinform\u0101ciju joproj\u0101m ir agr\u012bn\u0101 posm\u0101. Rakst\u0101 ar\u012b par\u0101d\u012bts, ka Eiropas valst\u012bs palielin\u0101s inform\u0113t\u012bba par dezinform\u0101cijas ietekmi, k\u0101 ar\u012b plaisa starp sp\u0113ju identific\u0113t \u201cviltus zi\u0146as\u201d un dezinform\u0101ciju un ierobe\u017eotu izpratni par dezinform\u0101cijas izplat\u012b\u0161an\u0101 iesaist\u012btajiem procesiem, draudiem un dal\u012bbniekiem.<\/p>

Publik\u0101cijas veids<\/strong>:
\u017durn\u0101la pap\u012brs<\/p>

Izdev\u0113js:<\/b>
Politisko zin\u0101t\u0146u un starptautisko attiec\u012bbu \u017eurn\u0101ls<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tJauns SCM pielietojums laika s\u0113rijas hipot\u0113tiskaj\u0101 apl\u0113s\u0113 farm\u0101cijas nozar\u0113<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Tomas Garriga, Gerard Sanz, Eduard Serrahima de Cambra, Axel Brando<\/p>

Kopsavilkums:<\/strong><\/p>

\u0160aj\u0101 dokument\u0101 m\u0113s iepaz\u012bstin\u0101m ar jaunu struktur\u0101lo c\u0113lo\u0146sakar\u012bbu mode\u013cu (SCM) un nolaup\u012b\u0161anas, r\u012bc\u012bbas un prognoz\u0113\u0161anas proced\u016bras piem\u0113ro\u0161anu laikrindu iestat\u012b\u0161anai saist\u012bb\u0101 ar re\u0101l\u0101s pasaules probl\u0113mu farm\u0101cijas nozar\u0113. M\u016bsu m\u0113r\u0137is ir apl\u0113st hipot\u0113tiskos scen\u0101rijus attiec\u012bb\u0101 uz t\u0101du z\u0101\u013cu p\u0101rdo\u0161anas apjomu, kuras ir ietekm\u0113jusi konkur\u0113jo\u0161u \u0123en\u0113risko z\u0101\u013cu ien\u0101k\u0161ana tirg\u016b. M\u0113s izmantojam uz kod\u0113t\u0101ju dekod\u0113t\u0101ju balst\u012btas arhitekt\u016bras, piem\u0113rojot nosac\u012btu vari\u0101cijas autoencoderu, k\u0101 ar\u012b ievie\u0161ot nosac\u012btu retu autoencoderu izmanto\u0161anu, kas nekad nav izmantoti hipot\u0113tiskaj\u0101 literat\u016br\u0101. Ierosin\u0101tajai metodikai ir nepiecie\u0161ama v\u0113sturisku notikumu un beznotikumu laikrindu pieejam\u012bba, un t\u0101s priek\u0161roc\u012bba ir nepa\u013cauties uz kontroles kovariantiem, kas var neb\u016bt pieejami, vienlaikus skaidri p\u0101rsniedzot prognozes pamata hipot\u0113tisko apl\u0113si. M\u0113s izv\u0113rt\u0113jam savu pieeju, izmantojot m\u016bsu uz\u0146\u0113muma re\u0101l\u0101s p\u0101rdo\u0161anas datu kopu, k\u0101 ar\u012b sint\u0113tiskas un da\u013c\u0113ji sint\u0113tiskas datu kopas, kas atdarina probl\u0113mas kontekstu, par\u0101dot t\u0101s efektivit\u0101ti. M\u0113s esam veiksm\u012bgi izmantoju\u0161i \u0161o modeli sav\u0101 uz\u0146\u0113mum\u0101, sniedzot noder\u012bgu inform\u0101ciju biznesa pl\u0101no\u0161anai, invest\u012bciju sadalei un m\u0113r\u0137u noteik\u0161anai.<\/p>

Publik\u0101cijas veids<\/strong>:
Semin\u0101ra raksts<\/p>

Izdev\u0113js:<\/b>
NeurIPS 24. darbsemin\u0101rs par m\u0101c\u012b\u0161anos par c\u0113lo\u0146u p\u0101rst\u0101v\u012bbu<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tZin\u0101\u0161anu diagrammas un ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s viltus zi\u0146u un dezinform\u0101cijas atkl\u0101\u0161an\u0101<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Anastasios Manos, Despina Elisabeth Filippidou, Nikolaos Pavlidis, Georgios Karanasios, Georgios Vachtanidis, Arianna D\u2019Ulizia<\/p>

Kopsavilkums:<\/strong>
Digit\u0101lo platformu pieaugums ir veicin\u0101jis dezinform\u0101cijas strauju izplat\u012b\u0161anos, kas rada b\u016btiskas soci\u0101las, politiskas un ekonomiskas probl\u0113mas. Zin\u0101\u0161anu grafiki (ZG) k\u013c\u016bst par efekt\u012bviem r\u012bkiem viltus zi\u0146u atkl\u0101\u0161anas sist\u0113mu precizit\u0101tes, interpret\u0113jam\u012bbas un m\u0113rogojam\u012bbas uzlabo\u0161anai, nov\u0113r\u0161ot ierobe\u017eojumus tradicion\u0101laj\u0101s uz ma\u0161\u012bnm\u0101c\u012b\u0161anos balst\u012btaj\u0101s pieej\u0101s, kas galvenok\u0101rt balst\u0101s uz lingvistisko anal\u012bzi. \u0160aj\u0101 darb\u0101 ir ietverts literat\u016bras apskats, kur\u0101 apkopoti konstat\u0113jumi no jaun\u0101kajiem p\u0113t\u012bjumiem par KG izmanto\u0161anu dezinform\u0101cijas atkl\u0101\u0161an\u0101. M\u0113s apzin\u0101mies, k\u0101 GV uzlabo atkl\u0101\u0161anu, kod\u0113jot re\u0101las attiec\u012bbas, analiz\u0113jot kontekstu un uzlabojot mode\u013cu interpret\u0113jam\u012bbu, vienlaikus apsprie\u017eot ar\u012b pa\u0161reiz\u0113jos m\u0113rogojam\u012bbas, datu piln\u012bguma un kontekstu\u0101l\u0101s piel\u0101gojam\u012bbas ierobe\u017eojumus. P\u0101rskat\u012btie p\u0113t\u012bjumi uzsver, ka ir vajadz\u012bgi turpm\u0101ki p\u0113t\u012bjumi, kas v\u0113rsti uz m\u0113rogojamiem, re\u0101llaika un starplingvistiskiem KG mode\u013ciem, lai stiprin\u0101tu dezinform\u0101cijas atkl\u0101\u0161anas sp\u0113jas vis\u0101 pasaul\u0113. Turkl\u0101t m\u0113s iepaz\u012bstin\u0101m ar divu izmanto\u0161anas gad\u012bjumu s\u0101kotn\u0113jiem rezult\u0101tiem, demonstr\u0113jot metodiku t\u0101du GV izveidei, kas var kalpot k\u0101 noder\u012bgi instrumenti c\u012b\u0146\u0101 pret dezinform\u0101cijas izplat\u012b\u0161anu.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
2024. gada Starptautisk\u0101 konference par in\u017eenierzin\u0101tn\u0113m un jaun\u0101m tehnolo\u0123ij\u0101m (ICEET)<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tPolitikas p\u0101rskat\u012b\u0161ana: Dezinform\u0101cijas apkaro\u0161ana digit\u0101laj\u0101 laikmet\u0101 \u2014 politika un iniciat\u012bvas demokr\u0101tijas aizsardz\u012bbai Eirop\u0101<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Alessia D\u2019Andrea, Giorgia Fusacchia, Arianna D\u2019Ulizia<\/p>

Kopsavilkums:<\/strong>
\u0160aj\u0101 dokument\u0101 ir p\u0113t\u012bta Eiropas Savien\u012bbas daudz\u0161\u0137aut\u0146ain\u0101 reakcija uz pla\u0161i izplat\u012bto dezinform\u0101cijas probl\u0113mu, kas ir pastiprin\u0101jusies kop\u0161 Krimas aneksijas 2014. gad\u0101. Dezinform\u0101cija b\u016btiski apdraud demokr\u0101tiskos procesus un sabiedr\u012bbas labkl\u0101j\u012bbu. Eiropas Savien\u012bbas pieeja apvieno regulat\u012bvus pas\u0101kumus, strat\u0113\u0123iskas partner\u012bbas un medijprat\u012bbas iniciat\u012bvas, lai risin\u0101tu \u0161o probl\u0113mu, vienlaikus aizsarg\u0101jot demokr\u0101tijas pamatprincipus, piem\u0113ram, v\u0101rda br\u012bv\u012bbu. Galvenie pas\u0101kumi ietver Prakses kodeksu dezinform\u0101cijas jom\u0101 un Digit\u0101lo pakalpojumu aktu, kuru m\u0113r\u0137is ir saukt digit\u0101l\u0101s platformas pie atbild\u012bbas un nodro\u0161in\u0101t p\u0101rredzam\u012bbu. Turkl\u0101t t\u0101das iniciat\u012bvas k\u0101 Austrumu Strat\u0113\u0123isk\u0101s komunik\u0101cijas operat\u012bv\u0101 grupa un \u0101tr\u0101s br\u012bdin\u0101\u0161anas sist\u0113ma uzsver Eiropas Savien\u012bbas centienus apkarot dezinform\u0101ciju k\u0101 hibr\u012bdkara instrumentu. \u0160aj\u0101 dokument\u0101 ar\u012b uzsv\u0113rta pilso\u0146u b\u016btisk\u0101 noz\u012bme, kurus Eiropas Savien\u012bba cen\u0161as stiprin\u0101t ar pla\u0161sazi\u0146as l\u012bdzek\u013cu lietotprasmes programm\u0101m, \u013caujot vi\u0146iem atpaz\u012bt manipulat\u012bvu saturu un pretoties tam. Izp\u0113tot mijiedarb\u012bbu starp vald\u012bbas darb\u012bb\u0101m, priv\u0101t\u0101 sektora iesaisti un iedz\u012bvot\u0101ju iesaisti, \u0161aj\u0101 p\u0113t\u012bjum\u0101 ir sniegta visaptvero\u0161a anal\u012bze par Eiropas Savien\u012bbas strat\u0113\u0123iju dezinform\u0101cijas apkaro\u0161anai un nov\u0113rt\u0113tas probl\u0113mas un turpm\u0101kie virzieni, kas vajadz\u012bgi, lai saglab\u0101tu demokr\u0101tisko notur\u012bbu main\u012bgaj\u0101 digit\u0101laj\u0101 vid\u0113.<\/p>

Publik\u0101cijas veids<\/strong>:
Zin\u0101tniskais raksts<\/p>

Izdev\u0113js:<\/b>
Inform\u0101cijas polisms<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tLingvistisks ieskats, mediju meh\u0101nismi un MI loma dezinform\u0101cijas izplat\u012b\u0161an\u0101 un ietekm\u0113<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Alessia D\u2019Andrea, Giorgia Fusacchia, Arianna D\u2019Ulizia<\/p>

Kopsavilkums:<\/strong><\/p>

M\u0113r\u0137is<\/em>
Izmantojot daudzdiscipl\u012bnu pieeju, \u0161\u0101 p\u0113t\u012bjuma m\u0113r\u0137is ir izsekot dezinform\u0101cijas kampa\u0146u ce\u013cam no to atkl\u0101\u0161anas ar lingvistisk\u0101m ticam\u012bbas nor\u0101d\u0113m l\u012bdz t\u0101l\u0101kai att\u012bst\u012bbai, izmantojot izplat\u012b\u0161anas meh\u0101nismus, un, visbeidzot, nov\u0113rt\u0113t to ietekmi uz soci\u0101lpolitisko kontekstu.<\/p>

Konstrukcija\/metodolo\u0123ija\/pieeja<\/em>
\u0160aj\u0101 p\u0113t\u012bjum\u0101 ir sniegts padzi\u013cin\u0101ts p\u0101rskats par \u010detriem b\u016btiskiem dezinform\u0101cijas aspektiem: lingvistisk\u0101s iez\u012bmes, kas at\u0161\u0137ir saturu, kura m\u0113r\u0137is ir maldin\u0101t sabiedrisko domu un manipul\u0113t ar to, mediju meh\u0101nismi, kas veicina t\u0101 izplat\u012b\u0161anu, izmantojot auditorijas kognit\u012bvos procesus, draudi, ko rada \u0123enerat\u012bv\u0101 m\u0101ksl\u012bg\u0101 intelekta arvien pla\u0161\u0101ka izmanto\u0161ana dezinform\u0101cijas izplat\u012b\u0161an\u0101, un pla\u0161\u0101kas sekas, ko \u0161\u012b dezinform\u0101cijas dinamika atst\u0101j uz sabiedrisko domu un l\u012bdz ar to uz politisko l\u0113mumu pie\u0146em\u0161anas procesiem.<\/p>

Konstat\u0113jumi<\/em>
Rezult\u0101t\u0101 dokument\u0101 ir sniegts \u0161\u012bs par\u0101d\u012bbas starpdisciplin\u0101rs un holistisks izv\u0113rt\u0113jums, atsaucoties uz t\u0101s daudzveid\u012bgajiem elementiem, lai uzsv\u0113rtu, cik svar\u012bga ir platformu atbild\u012bba, medijprat\u012bbas kampa\u0146as iedz\u012bvot\u0101ju vid\u016b un interakt\u012bva sadarb\u012bba starp priv\u0101to un publisko sektoru k\u0101 pas\u0101kumi, kuru m\u0113r\u0137is ir uzlabot notur\u012bbu pret dezinform\u0101cijas draudiem.<\/p>

Ori\u0123inalit\u0101te\/v\u0113rt\u012bba<\/em>
P\u0113t\u012bjum\u0101 uzsv\u0113rta nepiecie\u0161am\u012bba palielin\u0101t platformu p\u0101rskatatbild\u012bbu, veicin\u0101t medijprat\u012bbu indiv\u012bdu vid\u016b un att\u012bst\u012bt sadarb\u012bbu starp publisko un priv\u0101to sektoru. \u0160\u012bs integr\u0113t\u0101s strat\u0113\u0123ijas m\u0113r\u0137i ir stiprin\u0101t notur\u012bbu pret dezinform\u0101ciju un nodro\u0161in\u0101t ES sp\u0113ju piel\u0101goties main\u012bgajiem digit\u0101lajiem apdraud\u0113jumiem. Visbeidzot, dokument\u0101 ir atbalst\u012bta taisn\u012bga un atv\u0113rta strat\u0113\u0123ija, kas aizsarg\u0101 v\u0101rda br\u012bv\u012bbu un stiprina demokr\u0101tisk\u0101s iest\u0101des laik\u0101, kad pieaug digit\u0101l\u0101 dezinform\u0101cija.<\/p>

Publik\u0101cijas veids<\/strong>:
Zin\u0101tniskais raksts<\/p>

Izdev\u0113js:<\/b>
Inform\u0101cijas, komunik\u0101cijas un \u0113tikas \u017eurn\u0101ls sabiedr\u012bb\u0101 (2025)<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tTransformatoru iegul\u0161anas telpas zond\u0113\u0161ana ar minim\u0101l\u0101m mar\u0137iera perturb\u0101cij\u0101m<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Eddie Conti, Alejandro Astruc, \u00c1lvaro Parafita, Axel Brando<\/p>

Kopsavilkums:<\/strong><\/p>

Izpratne par to, k\u0101 inform\u0101cija izplat\u0101s, izmantojot transformatoru mode\u013cus, ir galvenais interpret\u0113jam\u012bbas izaicin\u0101jums. \u0160aj\u0101 darb\u0101 m\u0113s p\u0113t\u0101m minim\u0101lo mar\u0137iera perturb\u0101ciju ietekmi uz iegul\u0161anas telpu. M\u016bsu eksperimentos m\u0113s analiz\u0113jam, cik bie\u017ei \u017eetoni dod minim\u0101las mai\u0146as, uzsverot, ka reti \u017eetoni parasti noved pie liel\u0101k\u0101m mai\u0146\u0101m. Turkl\u0101t m\u0113s p\u0113t\u0101m, k\u0101 perturb\u0101cijas izplat\u0101s pa sl\u0101\u0146iem, par\u0101dot, ka ievades inform\u0101cija arvien vair\u0101k tiek sajaukta dzi\u013c\u0101kos sl\u0101\u0146os. M\u016bsu konstat\u0113jumi apstiprina kop\u012bgo pie\u0146\u0113mumu, ka mode\u013ca pirmos sl\u0101\u0146us var izmantot k\u0101 aizst\u0101j\u0113jv\u0113rt\u012bbas mode\u013ca skaidrojumiem. Kopum\u0101 \u0161is darbs ievie\u0161 \u017eetona perturb\u0101ciju un mai\u0146u kombin\u0101ciju iegul\u0161anas telp\u0101 k\u0101 sp\u0113c\u012bgu instrumentu mode\u013cu interpret\u0113jam\u012bbai.<\/p>

Publik\u0101cijas veids<\/strong>:
Semin\u0101ra raksts<\/p>

Izdev\u0113js:<\/b>
IJCAI 2025. gada darbsemin\u0101rs par izskaidrojamu m\u0101ksl\u012bgo intelektu<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tDzi\u013cviltojumu atkl\u0101\u0161ana viet\u0113j\u0101s kameras virsmas r\u0101mjos<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Andrea Ciamarra, Roberto Caldelli, Alberto Del Bimbo<\/p>

Kopsavilkums:<\/strong><\/p>

Dezinform\u0101cija vienm\u0113r tiek induc\u0113ta vair\u0101k, izmantojot p\u0101rveidotu vai piln\u012bgu sint\u0113tiski rad\u012btu re\u0101listisku saturu, ko parasti sauc par dzi\u013cviltojumiem. Tas rada vajadz\u012bbu p\u0113c uzticamiem instrumentiem, kas sp\u0113j atkl\u0101t \u0161\u0101du ne\u012bstu saturu, lai izvair\u012btos no personu dom\u0101\u0161anas polariz\u0101cijas. Ir izkl\u0101st\u012bti da\u017e\u0101di pa\u0146\u0113mieni, kuru kopsauc\u0113js ir mekl\u0113t da\u017eas neatbilst\u012bbas, ko izrais\u012bjis viltus \u0123ener\u0113\u0161anas process un kas apliecin\u0101tas, apl\u016bkojot konkr\u0113tas \u012bpa\u0161\u012bbas. \u0160\u0137iet, ka da\u017e\u0101du at\u0161\u0137ir\u012bgu iez\u012bmju iesp\u0113jam\u0101 kombin\u0101cija var\u0113tu nodro\u0161in\u0101t efekt\u012bv\u0101kus sasniegumus. \u0145emot v\u0113r\u0101 \u0161o viedokli un m\u016bsu iepriek\u0161\u0113jo p\u0113t\u012bjumu par \u0161o jaut\u0101jumu, m\u0113s ierosin\u0101m izp\u0113t\u012bt jaunu funkciju, kur\u0101 b\u016bt\u012bb\u0101 tiek \u0146emtas v\u0113r\u0101 ieb\u016bv\u0113t\u0101s attiec\u012bbas, kas past\u0101v vis\u0101 ieg\u0101des ain\u0101 att\u0113la (video) uz\u0146em\u0161anas laik\u0101 un ko p\u0113c tam att\u0113lo att\u0113la pikse\u013cu v\u0113rt\u012bbas. Ideja ir t\u0101da, ka dzi\u013cviltojumu \u0123ener\u0113\u0161anas process var p\u0101rraut \u0161\u0101das iek\u0161\u0113jas atkar\u012bbas starp apgaismojumu (intensit\u0101ti un virzienu), objektu virsm\u0101m (poz\u012bciju un atstarojumu) un uztver\u0161anas troksni, t\u0101d\u0113j\u0101di apliecinot iesp\u0113jamu manipul\u0101ciju. Saska\u0146\u0101 ar to ir izveidots funkciju deskriptors, ko sauc par viet\u0113jo virsmas deskriptoru, un to izmanto, lai apm\u0101c\u012btu CNN balst\u012btus mode\u013cus dzi\u013cviltojumu atkl\u0101\u0161anai. Eksperiment\u0101lie rezult\u0101ti, kas veikti da\u017e\u0101dos operat\u012bvajos kontekstos, liecina par iev\u0113rojamu noteik\u0161anas precizit\u0101ti l\u012bdz 90%.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
2024. gada IEEE starptautisk\u0101 konference par att\u0113lu apstr\u0101des izaicin\u0101jumiem un darbsemin\u0101riem (ICIPCW), Ab\u016b Dab\u012b, Apvienotie Ar\u0101bu Emir\u0101ti, 2024. gads<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tPiln\u012bgi sint\u0113tisku sejas att\u0113lu r\u0101d\u012b\u0161ana, izmantojot viet\u0113jos kameras virsmas r\u0101mjus<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Andrea Ciamarra, Roberto Caldelli, Alberto Del Bimbo<\/p>

Kopsavilkums:<\/strong><\/p>

Faktisk\u0101 sp\u0113ja rad\u012bt re\u0101listiskus piln\u012bb\u0101 sint\u0113tiskus att\u0113lus ar MI katru dienu uzlabojas, un tas jo \u012bpa\u0161i attiecas uz att\u0113liem, kuros att\u0113lotas cilv\u0113ka sejas, kas \u0161\u0137iet neat\u0161\u0137iramas no re\u0101liem cilv\u0113kiem. Tas rada b\u016btisku nepiecie\u0161am\u012bbu izstr\u0101d\u0101t instrumentus, kas sp\u0113j at\u0161\u0137irt patiesos un neeso\u0161os cilv\u0113kus, atkl\u0101jot da\u017eas iesp\u0113jam\u0101s neatbilst\u012bbas, kas iestr\u0101d\u0101tas att\u0113los \u0123ener\u0113\u0161anas proces\u0101. Galven\u0101 at\u0161\u0137ir\u012bba starp pirmatn\u0113ju att\u0113lu un \u0123ener\u0113tu dzi\u013cviltojumu ir t\u0101, ka otraj\u0101 gad\u012bjum\u0101 nav notikusi efekt\u012bva kameras ieg\u0101de; t\u0101p\u0113c visas da\u017e\u0101d\u0101s savstarp\u0113j\u0101s attiec\u012bbas starp elementiem, kas pieder pie skatuves (gaismas, atstaro\u0161ana, objektu attiec\u012bg\u0101s poz\u012bcijas 3D telp\u0101), re\u0101laj\u0101 pasaul\u0113 netiek uz\u0146emtas \u0161aj\u0101 prec\u012bz\u0101 laika moment\u0101, bet tikai m\u0101ksl\u012bgi reproduc\u0113tas. Saska\u0146\u0101 ar \u0161\u0101du apsv\u0113rumu, \u0161aj\u0101 darb\u0101 m\u0113s ievie\u0161am viet\u0113jo kameru virsmas r\u0101mjus k\u0101 iesp\u0113jamu l\u012bdzekli, lai att\u0113lotu \u0161os specifiskos vides raksturlielumus, lai izceltu at\u0161\u0137ir\u012bbas. Veikt\u0101 eksperiment\u0101l\u0101 anal\u012bze liecina, ka \u0161\u012b funkcija var pie\u0161\u0137irt \u013coti augstu precizit\u0101tes l\u012bmeni un iev\u0113rojamu visp\u0101rin\u0101\u0161anas pak\u0101pi.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
2024. gada IEEE starptautisk\u0101 konference par att\u0113lu apstr\u0101des izaicin\u0101jumiem un darbsemin\u0101riem (ICIPCW), Ab\u016b Dab\u012b, Apvienotie Ar\u0101bu Emir\u0101ti, 2024. gads<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tPar lok\u0101lo virsmas r\u0101mju visp\u0101rin\u0101\u0161anas sp\u0113ju uz dif\u016bziju balst\u012btu sejas att\u0113lu noteik\u0161an\u0101<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Andrea Ciamarra, Roberto Caldelli, Alberto Del Bimbo<\/p>

Kopsavilkums:<\/strong><\/p>

\u0100rk\u0101rt\u012bgi nere\u0101lus att\u0113lus var realiz\u0113t ar pow-erful AI metod\u0113m. Da\u017e\u0101di r\u012bki, kas pieejami ikvienam, sp\u0113j atjaunot augstas kvalit\u0101tes saturu, jo \u012bpa\u0161i radot piln\u012bgi sint\u0113tiskus att\u0113lus. Starp eso\u0161aj\u0101m arhitekt\u016br\u0101m mode\u013ci, kuru pamat\u0101 ir dif\u016bzija, var viegli rad\u012bt jebk\u0101da veida att\u0113lus, tostarp cilv\u0113ka sejas att\u0113lus, sniedzot uzvedni, piem\u0113ram, tekstu. \u0160\u0101ds nepatiess saturs bie\u017ei tiek izmantots, lai izplat\u012btu dezinform\u0101ciju, un tas rada ba\u017eas par cilv\u0113ku dro\u0161\u012bbu. Pa\u0161laik ir gr\u016bti izstr\u0101d\u0101t uzticamus instrumentus, lai no\u0161\u0137irtu re\u0101lus un rad\u012btus (pat neeso\u0161us) cilv\u0113kus. Turkl\u0101t lielais uz dif\u016bziju balst\u012btais \u012bsteno\u0161anas gad\u012bjumu skaits rada probl\u0113mu, ka \u0161\u0101di detektori visp\u0101rina jaunus \u0123enerat\u012bvus pa\u0146\u0113mienus. Lai risin\u0101tu \u0161os jaut\u0101jumus, m\u0113s ierosin\u0101m izp\u0113t\u012bt \u012bpa\u0161as iez\u012bmes sp\u0113ju, pamatojoties uz att\u0113lu ieg\u016b\u0161anas vidi, no\u0161\u0137irt uz dif\u016bziju balst\u012btus sejas att\u0113lus no neskartiem. Faktiski \u0123ener\u0113tajiem att\u0113liem nevajadz\u0113tu satur\u0113t \u012bpa\u0161\u012bbas, kas ir atbilsto\u0161as ieg\u016b\u0161anas f\u0101zei, ko veic ar re\u0101lu kameru. \u0160\u0101das neatbilst\u012bbas var izcelt, izmantojot nesen ieviestos viet\u0113jos virsmas r\u0101mjus. \u0160\u012b funkcija \u0146em v\u0113r\u0101 ain\u0101 iesaist\u012btos objektus un virsmas, kas visi ietekm\u0113 kameras ieg\u016b\u0161anas procesu, k\u0101 ar\u012b papildu iek\u0161\u0113jo inform\u0101ciju, kas saist\u012bta ar ier\u012bci, k\u0101 ar\u012b apgaismojumu un atstarojumus, kas ietekm\u0113 visu scen\u0101riju. Darb\u0101 p\u0113t\u012bta \u0161\u012bs funkcijas sp\u0113ja visp\u0101rin\u0101ties uz da\u017e\u0101d\u0101m datu kop\u0101m un jaun\u0101m \u0123enerat\u012bv\u0101m metod\u0113m, kas apm\u0101c\u012bbas laik\u0101 nav zin\u0101mas. Eksperiment\u0101lie rezult\u0101ti liecina, ka \u0161\u0101da funkcija joproj\u0101m nodro\u0161ina iev\u0113rojamu atkl\u0101\u0161anas precizit\u0101tes l\u012bmeni ar\u012b \u0161ajos gad\u012bjumos.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
2024. gada IEEE starptautisk\u0101 konference par att\u0113lu apstr\u0101des izaicin\u0101jumiem un darbsemin\u0101riem (ICIPCW), Ab\u016b Dab\u012b, Apvienotie Ar\u0101bu Emir\u0101ti, 2024. gads<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tPraktiski do-Shapley skaidrojumi ar Estimand-Agnostic Causal Inference<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
\u00c1lvaro Parafita, Tomas Garriga, Axel Brando, Francisco J. Cazorla<\/p>

Kopsavilkums:<\/strong><\/p>

Starp izskaidrojam\u012bbas metod\u0113m SHAP izce\u013cas k\u0101 viens no popul\u0101r\u0101kajiem, bet bie\u017ei vien ne\u0146em v\u0113r\u0101 probl\u0113mas c\u0113lo\u0146sakar\u012bbu strukt\u016bru. Atbildot uz to, do-SHAP izmanto intervences vaic\u0101jumus, bet t\u0101 pa\u013cau\u0161an\u0101s uz apl\u0113s\u0113m kav\u0113 t\u0101 praktisko piem\u0113ro\u0161anu. Lai risin\u0101tu \u0161o probl\u0113mu, m\u0113s ierosin\u0101m izmantot estimand-agnostiskas pieejas, kas \u013cauj apl\u0113st jebkuru identific\u0113jamu vaic\u0101jumu no viena mode\u013ca, padarot do-SHAP iesp\u0113jamu sare\u017e\u0123\u012bt\u0101s diagramm\u0101s. M\u0113s ar\u012b izstr\u0101d\u0101jam jaunu algoritmu, lai iev\u0113rojami pa\u0101trin\u0101tu t\u0101 apr\u0113\u0137in\u0101\u0161anu ar nenoz\u012bm\u012bg\u0101m izmaks\u0101m, k\u0101 ar\u012b metodi, lai izskaidrotu nepieejamus datu \u0123ener\u0113\u0161anas procesus. M\u0113s demonstr\u0113jam m\u016bsu pieejas apl\u0113si un skait\u013co\u0161anas veiktsp\u0113ju un valid\u0113jam to div\u0101s re\u0101los apst\u0101k\u013cos ieg\u016bt\u0101s datu kop\u0101s, uzsverot t\u0101s potenci\u0101lu uzticamu paskaidrojumu ieg\u016b\u0161an\u0101.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
Tr\u012bsdesmit dev\u012bt\u0101 ikgad\u0113j\u0101 konference par neironu inform\u0101cijas apstr\u0101des sist\u0113m\u0101m (NeurIPS 2025)<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tJa izg\u016b\u0161ana p\u0101rsp\u0113j \u0123ener\u0113\u0161anu: Bl\u012bvu pier\u0101d\u012bjumu izg\u016b\u0161ana m\u0113rogojamai viltus zi\u0146u atkl\u0101\u0161anai<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Alamgir Munir Qazi, John Philip McCrae, Jamal Nasir<\/p>

Kopsavilkums:<\/strong><\/p>

Maldino\u0161as inform\u0101cijas izplat\u012b\u0161an\u0101s d\u0113\u013c ir vajadz\u012bgas stabilas, bet skait\u013co\u0161anas zi\u0146\u0101 efekt\u012bvas faktu p\u0101rbaudes sist\u0113mas. Lai gan pa\u0161reiz\u0113j\u0101s m\u016bsdien\u012bg\u0101s pieejas izmanto apjom\u012bgus valodu mode\u013cus (LLM), lai rad\u012btu paskaidrojo\u0161us pamatojumus, \u0161\u012bs metodes saskaras ar b\u016btiskiem skait\u013co\u0161anas \u0161\u0137\u0113r\u0161\u013ciem un halucin\u0101cijas riskiem re\u0101los apst\u0101k\u013cos. M\u0113s pied\u0101v\u0101jam DeReC (bl\u012bvas izguves klasifik\u0101ciju), vieglu sist\u0113mu, kas par\u0101da, k\u0101 visp\u0101r\u0113ja lietojuma teksta iegul\u0161ana var efekt\u012bvi aizst\u0101t autoregres\u012bvas LLM balst\u012btas pieejas, faktiski veicot p\u0101rbaudes uzdevumus. Apvienojot bl\u012bvu izg\u016b\u0161anu ar specializ\u0113tu klasifik\u0101ciju, m\u016bsu sist\u0113ma sasniedz lab\u0101ku precizit\u0101ti, vienlaikus b\u016bdama iev\u0113rojami efekt\u012bv\u0101ka. DeReC p\u0101rsp\u0113j paskaidrojo\u0161\u0101s LLM efektivit\u0101tes zi\u0146\u0101, samazinot darb\u012bbas laiku par 95% RAWFC (23 min\u016btes 36 sekundes sal\u012bdzin\u0101jum\u0101 ar 454 min\u016bt\u0113m 12 sekundes) un 92% LIAR-RAW (134 min\u016btes 14 sekundes sal\u012bdzin\u0101jum\u0101 ar 1692 min\u016bt\u0113m 23 sekundes), demonstr\u0113jot t\u0101 efektivit\u0101ti da\u017e\u0101dos datu kopu izm\u0113ros. RAWFC datu kop\u0101 DeReC sasniedz F1 punktu skaitu 65,58%, kas p\u0101rsniedz jaun\u0101ko L-Defense metodi (61.20%). M\u016bsu rezult\u0101ti par\u0101da, ka r\u016bp\u012bgi izstr\u0101d\u0101tas izguves sist\u0113mas var saska\u0146ot vai p\u0101rsniegt LLM veiktsp\u0113ju specializ\u0113tos uzdevumos, vienlaikus iev\u0113rojami praktisk\u0101k pielietojot re\u0101los apst\u0101k\u013cos.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
Piekt\u0101 konference par valodu, datiem un zin\u0101\u0161an\u0101m<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tUz CLIP balst\u012bta pieeja sint\u0113tiskai att\u0113lu noteik\u0161anai distrib\u016bcijas mai\u0146\u0101<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Qiushi Li, Andrea Ciamarra, Roberto Caldelli un Stefano Berretti<\/p>

Kopsavilkums:<\/strong><\/p>

Attiec\u012bb\u0101 uz MediaEval 2025 SynthIm Challenge Task A (Synthetic Image Detection) m\u0113s konstat\u0113j\u0101m kritisku sadal\u012bjuma mai\u0146u starp sniegtajiem apm\u0101c\u012bbas un testa datiem, padarot biju\u0161o par sliktu p\u0101rst\u0101vi m\u0113r\u0137a jom\u0101. M\u016bsu pieeja tie\u0161i risina \u0161o probl\u0113mu, iek\u013caujot stilistiski konsekvent\u0101ku valid\u0101cijas kopumu m\u016bsu apm\u0101c\u012bbas datos un izmantojot iesald\u0113to CLIP ViT-L\/14 k\u0101 sp\u0113c\u012bgu funkciju ekstraktoru. M\u016bsu galvenais ieskats ir t\u0101ds, ka \u0161\u0101dos dom\u0113nu mai\u0146as apst\u0101k\u013cos pla\u0161a iepriek\u0161 apm\u0101c\u012bta mode\u013ca visp\u0101rin\u0101mie atveidojumi iev\u0113rojami p\u0101rsniedz tradicion\u0101lo CNN, kas piel\u0101gots nesaska\u0146otajai datu kopai, un t\u0101 ir efekt\u012bv\u0101ka un uzticam\u0101ka strat\u0113\u0123ija.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
MediaEval\u201d25: Multivides izv\u0113rt\u0113\u0161anas darbsemin\u0101rs 2025. gada 25.\u201326. oktobr\u012b Dublin\u0101, \u012arij\u0101 un tie\u0161saist\u0113<\/p>

Piek\u013cuve p\u0113c piepras\u012bjuma<\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tUzlabota vizu\u0101l\u0101 kvalit\u0101te \u016adensz\u012bme, kuras pamat\u0101 ir divu atsau\u010du zudumi par dzi\u013cviltojumu attiecin\u0101\u0161anu<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Qiushi Li, Roberto Caldelli un Stefano Berretti<\/p>

Kopsavilkums:<\/strong><\/p>

Att\u0113lu veido\u0161anas mode\u013cu, piem\u0113ram, stabilas dif\u016bzijas, strauj\u0101 att\u012bst\u012bba rada ba\u017eas par iesp\u0113jamu \u013caunpr\u0101t\u012bgu izmanto\u0161anu, padarot robustas \u016bdensz\u012bmes metodes b\u016btiskas sint\u0113tiska satura autentifik\u0101cijai un attiecin\u0101\u0161anai, jo \u012bpa\u0161i c\u012b\u0146\u0101 pret dzi\u013cviltojumiem. Tom\u0113r joproj\u0101m ir gr\u016bti vienlaikus nodro\u0161in\u0101t augstas kvalit\u0101tes att\u0113lu \u0123ener\u0113\u0161anu un prec\u012bzu \u016bdensz\u012bmju ieguvi. Analiz\u0113jot eso\u0161\u0101s metodes, m\u0113s identific\u0113jam kritisku ierobe\u017eojumu:
to zuduma funkcijas bie\u017ei vien izmanto vienu atsauci (vai nu ievades att\u0113lu, vai t\u012bru \u0123ener\u0113tu att\u0113lu), lai optimiz\u0113tu att\u0113la precizit\u0101ti, kas noved pie neoptim\u0101las veiktsp\u0113jas. \u0160aj\u0101 dokument\u0101 m\u0113s veicam padzi\u013cin\u0101tu p\u0113t\u012bjumu par att\u0113la kvalit\u0101tes zuduma termi\u0146u \u016bdensz\u012bm\u0113s, kuru pamat\u0101 ir dif\u016bzija. Analiz\u0113jot at\u0161\u0137ir\u012bgo ietekmi, ko rada ievades att\u0113la izmanto\u0161ana sal\u012bdzin\u0101jum\u0101 ar t\u012bru \u0123ener\u0113to att\u0113lu k\u0101 atsauces optimiz\u0101cijas laik\u0101, m\u0113s atkl\u0101jam, ka, kop\u012bgi \u0146emot abas atsauces, iev\u0113rojami uzlabo notur\u012bbu un vizu\u0101lo kvalit\u0101ti. Pla\u0161i eksperimenti par\u0101da, ka m\u016bsu divk\u0101r\u0161\u0101s atsauces pieeja nodro\u0161ina izcilu veiktsp\u0113ju gan \u016bdensz\u012bmju ekstrakcijas precizit\u0101tes, gan \u0123ener\u0113\u0161anas precizit\u0101tes zi\u0146\u0101, sal\u012bdzinot ar vienas atsauces b\u0101zl\u012bnij\u0101m. M\u0113s atbalst\u0101m \u0161o paradigmu, lai veicin\u0101tu uzticamu \u016bdensz\u012bm\u0113\u0161anu \u0123enerat\u012bvos mode\u013cos.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
DFF \u201925: 1. darbsemin\u0101ra par dzi\u013cviltotu krimin\u0101listiku proced\u016bras: Atkl\u0101\u0161ana, attiecin\u0101\u0161ana, atz\u012b\u0161ana un pretrunu probl\u0113mas MI rad\u012bto mediju laikmet\u0101<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tNost\u0101jas dokuments: Ja inov\u0101cija m\u0101ksl\u012bg\u0101 intelekta jom\u0101 sistem\u0101tiski p\u0101rk\u0101pj pamatties\u012bbas, vai t\u0101 ir inov\u0101cija visp\u0101r?<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autori:<\/strong>
Josu Eguiluz Casta\u00f1eira, Axel Brando, Migle Laukyte, Marc Serra-Vidal<\/p>

Kopsavilkums:<\/strong><\/p>

M\u0101ksl\u012bgais intelekts (MI) tagad caurvij kritisko infrastrukt\u016bru un l\u0113mumu pie\u0146em\u0161anas sist\u0113mas, kur neveiksmes rada soci\u0101lu, ekonomisku un demokr\u0101tisku kait\u0113jumu. \u0160is nost\u0101jas dokuments ap\u0161auba iesak\u0146oju\u0161os p\u0101rliec\u012bbu, ka regul\u0113jums un inov\u0101cija ir pretstati. K\u0101 liecina avi\u0101cijas, farm\u0101cijas un labkl\u0101j\u012bbas sist\u0113mu analo\u0123ijas un nesenie sint\u0113tiskas maldino\u0161as inform\u0101cijas, neobjektivit\u0101tes un nekontrol\u0113jamas l\u0113mumu pie\u0146em\u0161anas gad\u012bjumi, labi izstr\u0101d\u0101ta regul\u0113juma tr\u016bkums jau ir rad\u012bjis neizm\u0113r\u0101mu kait\u0113jumu. Ja regul\u0113jums ir p\u0101rdom\u0101ts un adapt\u012bvs, tas nekav\u0113 inov\u0101ciju \u2014 tas ir t\u0101s pamats. \u0160aj\u0101 nost\u0101jas dokument\u0101 ES M\u0101ksl\u012bg\u0101 intelekta akts ir apl\u016bkots k\u0101 uz risku balst\u012bta, uz atbild\u012bbu balst\u012bta regul\u0113juma modelis, kas piev\u0113r\u0161as Collingridge Dilemma: r\u012bkoties pietiekami savlaic\u012bgi, lai nov\u0113rstu kait\u0113jumu, bet pietiekami elast\u012bgi, lai atbalst\u012btu inov\u0101ciju. T\u0101s adapt\u012bvie meh\u0101nismi \u2014 regulat\u012bv\u0101s smil\u0161kastes, atbalsts maziem un vid\u0113jiem uz\u0146\u0113mumiem (MVU), test\u0113\u0161ana re\u0101los apst\u0101k\u013cos, ietekmes uz pamatties\u012bb\u0101m nov\u0113rt\u0113jums (FRIA) \u2014 par\u0101da, k\u0101 regul\u0113jums var atbild\u012bgi pa\u0101trin\u0101t, nevis aizkav\u0113t tehnolo\u0123isko progresu. Nost\u0101jas dokument\u0101 ir apkopots, k\u0101 p\u0101rvald\u012bbas instrumenti uztveramo slogu p\u0101rv\u0113r\u0161 taust\u0101m\u0101s priek\u0161roc\u012bb\u0101s: juridisk\u0101 noteikt\u012bba, pat\u0113r\u0113t\u0101ju uztic\u0113\u0161an\u0101s un \u0113tiska konkur\u0113tsp\u0113ja. Galu gal\u0101 dokument\u0101 ir p\u0101rskat\u012bts progress: inov\u0101cija un regul\u0113jums virz\u0101s uz priek\u0161u kop\u0101. Iek\u013caujot projekt\u0113\u0161an\u0101 un ievie\u0161an\u0101 p\u0101rredzam\u012bbu, ietekmes nov\u0113rt\u0113jumus, p\u0101rskatatbild\u012bbu un MI prat\u012bbu, ES satvars nosaka, ko \u012bsti noz\u012bm\u0113 atbild\u012bga inov\u0101cija \u2013 tehnolo\u0123iskie m\u0113r\u0137i, ko disciplin\u0113 demokr\u0101tisk\u0101s v\u0113rt\u012bbas un pamatties\u012bbas.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
Tr\u012bsdesmit dev\u012bt\u0101 ikgad\u0113j\u0101 konference par neironu inform\u0101cijas apstr\u0101des sist\u0113m\u0101m (NeurIPS 2025)<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tUztic\u0113\u0161an\u0101s, jaun\u0101s tehnolo\u0123ijas un \u0123eopolitika nedro\u0161\u0101 pasaul\u0113<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autors:\u00a0<\/strong>Paskal\u012bna gaborita<\/p>

Kopsavilkums:<\/strong><\/p>

Glob\u0101l\u0101s nenoteikt\u012bbas laikmet\u0101 uztic\u0113\u0161an\u0101s ir k\u013cuvusi par iz\u0161\u0137iro\u0161u faktoru, kas veido attiec\u012bbas starp n\u0101cij\u0101m, iest\u0101d\u0113m un indiv\u012bdiem. \u0160\u012b gr\u0101mata, Trust, New Technologies and Geopolitics in an Uncertain World, pied\u0101v\u0101 savlaic\u012bgu un padzi\u013cin\u0101tu izp\u0113ti par to, k\u0101 uztic\u0113\u0161an\u0101s tiek p\u0101rbaud\u012bta un p\u0101rveidota, \u0146emot v\u0113r\u0101 strauji main\u012bg\u0101s \u0123eopolitisk\u0101s ainavas. No demokr\u0101tisko sist\u0113mu nestabilit\u0101tes l\u012bdz izaicin\u0101jumiem, ko rada jaun\u0101s tehnolo\u0123ijas, m\u0101ksl\u012bgais intelekts, dezinform\u0101cija un klimata p\u0101rmai\u0146as, \u0161\u012b gr\u0101mata piev\u0113r\u0161as m\u016bsdienu aktu\u0101l\u0101kajiem jaut\u0101jumiem. Izp\u0113tot, k\u0101 uztic\u0113\u0161an\u0101s krustojas ar t\u0101d\u0101m svar\u012bg\u0101m jom\u0101m k\u0101 demokr\u0101tija, tirdzniec\u012bbas kari un jaun\u0101s tehnolo\u0123ijas, \u0161is darbs sniedz v\u0113rt\u012bgu ieskatu politikas veidot\u0101jiem, zin\u0101tniekiem un ikvienam, kas v\u0113las izprast m\u016bsdienu pasaules sare\u017e\u0123\u012bt\u012bbu. Neatkar\u012bgi no t\u0101, vai jums r\u016bp kiberdro\u0161\u012bba, dezinform\u0101cija, hibr\u012bddraudu ietekme vai uztic\u0113\u0161an\u0101s starptautiskajai diplom\u0101tijai, \u0161\u012b gr\u0101mata pied\u0101v\u0101 visaptvero\u0161u, bet pieejamu satvaru \u0161o probl\u0113mu risin\u0101\u0161anai. Tas par\u0101da, k\u0101 uztic\u0113\u0161an\u0101s izpratne ir b\u016btiska, lai nov\u0113rstu \u0123eopolitisko nenoteikt\u012bbu, kas nosaka m\u016bsu laikmetu.<\/p>

Publik\u0101cijas veids<\/strong>:
Gr\u0101mata<\/p>

Izdev\u0113js:<\/b>
P\u012bters Langs Izdev\u0113js<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t

\n\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t<\/path><\/svg><\/span>\n\t\t\t\t\t\t\t\t\t\t\t\t\t<\/span>\n\t\t\t\t\t\t\t\t\t\t\t\tP\u012aL: Funkciju noz\u012bm\u012bguma m\u0113r\u012b\u0161ana, izmantojot hipot\u0113tisko sadali<\/a>\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t

Autors: <\/strong><\/p>

Edijs Konti (Eddie Conti), Alvaro Parafita (\u00c1lvaro Parafita), Aksels Brando (Axel Brando)<\/p>

Kopsavilkums:<\/strong><\/p>

Lai izprastu mode\u013ca l\u0113mumu pie\u0146em\u0161anas procesu, ir \u013coti svar\u012bgi nov\u0113rt\u0113t atsevi\u0161\u0137u iez\u012bmju noz\u012bmi ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s proces\u0101. Lai gan past\u0101v daudzas metodes, tas, ka nav gal\u012bgas pamata paties\u012bbas sal\u012bdzin\u0101\u0161anai, nor\u0101da uz nepiecie\u0161am\u012bbu p\u0113c alternat\u012bviem, labi pamatotiem pas\u0101kumiem. \u0160is dokuments iepaz\u012bstina ar jaunu post-hoc viet\u0113j\u0101s noz\u012bmes metodi, ko sauc par hipot\u0113tisk\u0101s noz\u012bmes izplat\u012b\u0161anu (CID). M\u0113s \u0123ener\u0113jam divus pozit\u012bvu un negat\u012bvu hipot\u0113tisko scen\u0101riju kopumus, model\u0113jam to sadal\u012bjumus, izmantojot kodola bl\u012bvuma nov\u0113rt\u0113jumu, un sarindojam iez\u012bmes, pamatojoties uz distribut\u012bv\u0101s at\u0161\u0137ir\u012bbas m\u0113ru. \u0160is m\u0113rs, kura pamat\u0101 ir stingra matem\u0101tisk\u0101 sist\u0113ma, atbilst galvenaj\u0101m \u012bpa\u0161\u012bb\u0101m, kas nepiecie\u0161amas, lai darbotos k\u0101 der\u012bgs r\u0101d\u012bt\u0101js. M\u0113s demonstr\u0113jam m\u016bsu metodes efektivit\u0101ti, sal\u012bdzinot ar visp\u0101ratz\u012btiem viet\u0113jiem funkciju noz\u012bmes skaidrot\u0101jiem. M\u016bsu metode ne tikai pied\u0101v\u0101 papildu perspekt\u012bvas eso\u0161aj\u0101m pieej\u0101m, bet ar\u012b uzlabo uzticam\u012bbu (gan vispus\u012bgumu, gan pietiekam\u012bbu), k\u0101 rezult\u0101t\u0101 tiek sniegti uzticam\u0101ki sist\u0113mas skaidrojumi. \u0160ie rezult\u0101ti izce\u013c t\u0101s k\u0101 v\u0113rt\u012bga mode\u013ca anal\u012bzes instrumenta potenci\u0101lu.<\/p>

Publik\u0101cijas veids<\/strong>:
Konferences dokuments<\/p>

Izdev\u0113js:<\/b>
Zieme\u013cu gaismas dzi\u013c\u0101s m\u0101c\u012b\u0161an\u0101s konference 2026. gad\u0101<\/p>

Piek\u013cuve \u0161eit<\/a><\/p><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\n\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\tAtgrie\u0161an\u0101s pie resursiem<\/span>\n\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>","protected":false},"excerpt":{"rendered":"

Zin\u0101tnisk\u0101s publik\u0101cijas Atrodiet visas zin\u0101tnisk\u0101s publik\u0101cijas, ko sagatavoju\u0161i AI4Debunk partneri, iepaz\u012bstinot ar projekta jaun\u0101kajiem zin\u0101tniskajiem konstat\u0113jumiem. Gaismas izklied\u0113\u0161ana lielos \u0123enerat\u012bvos t\u012bklos: Epist\u0113tikas nenoteikt\u012bbas nov\u0113rt\u0113\u0161ana dif\u016bzijas mode\u013cos Autori:Axel Brando, Lucas Berry, David Mege Abstract:\u0122enerat\u012bvie dif\u016bzijas mode\u013ci, kas ir iev\u0113rojami ar to lielo parametru skaitu (p\u0101rsniedzot 100 miljonus) un darb\u012bbu augstas dimensijas att\u0113lu telp\u0101s, [...]<\/p>","protected":false},"author":6,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"footnotes":""},"class_list":["post-124","page","type-page","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/pages\/124","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/comments?post=124"}],"version-history":[{"count":5,"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/pages\/124\/revisions"}],"predecessor-version":[{"id":3001,"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/pages\/124\/revisions\/3001"}],"wp:attachment":[{"href":"https:\/\/ai4debunk.eu\/lv\/wp-json\/wp\/v2\/media?parent=124"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}