Foto: Veras Moln\u0101ras darbs no Oskara Teikma\u0146a person\u012bg\u0101 arh\u012bva<\/p>\n
<\/p>\n
2023. gada 14. decembr\u012b Latvijas Nacion\u0101laj\u0101 bibliot\u0113k\u0101 notika semin\u0101rs \u201cM\u0101ksl\u012bg\u0101 intelekta r\u012bku pielietojumi da\u017e\u0101d\u0101s zin\u0101tnes nozar\u0113s<\/strong>\u201d, kur\u0101 Latvijas Jauno zin\u0101tnieku apvien\u012bbas (LJZA) biedri dal\u012bj\u0101s ar savu pieredzi par m\u0101ksl\u012bg\u0101 intelekta r\u012bku izmanto\u0161anu zin\u0101tniskaj\u0101 darb\u0101. Semin\u0101rs notika cikla \u201cZin\u0101tne sabiedr\u012bbai<\/strong>\u201d ietvaros, kas tiek organiz\u0113ts sadarb\u012bb\u0101 ar Latvijas Nacion\u0101lo bibliot\u0113ku<\/strong>.<\/p>\n Apm\u0113ram pirms gada pasauli p\u0101r\u0161alca diskusijas par ChatGPT un t\u0101 sniegtaj\u0101m iesp\u0113j\u0101m. T\u0101 lietot\u0101ju skaits da\u017eu m\u0113ne\u0161u laik\u0101 p\u0101rsniedza 100 miljonu cilv\u0113ku. M\u0101ksl\u012bg\u0101 intelekta tehnolo\u0123ijas strauji att\u012bst\u0101s, un t\u0101s sp\u0113j rad\u012bt ne tikai augstas kvalit\u0101tes tekstu da\u017e\u0101d\u0101s valod\u0101s, bet ar\u012b bildes, video materi\u0101lus un cita veida saturu. \u0160aj\u0101 semin\u0101r\u0101 piedal\u012bj\u0101s p\u0113tnieki no da\u017e\u0101d\u0101m zin\u0101tnes nozar\u0113m: M.Sc.<\/i> Oskars Teikmanis<\/strong> (in\u017eenierzin\u0101tnes, Elektronikas un datorzin\u0101t\u0146u instit\u016bts), Mg.biol.<\/i> Eva Em\u012blija \u010cesle<\/strong> (molekul\u0101r\u0101 biolo\u0123ija, Latvijas Biomedic\u012bnas studiju un p\u0113t\u012bjumu centrs) un Mg.sc.soc.<\/i> Ance Krist\u0101la<\/strong> (soci\u0101l\u0101s zin\u0101tnes, Latvijas Kult\u016bras akad\u0113mija).<\/p>\n <\/p>\n Pirmais st\u0101sts: M\u0101ksl\u012bgais intelekts k\u0101 koda \u0123enerators un iedvesmas avots (Oskars Teikmanis)<\/strong><\/p>\n LJZA valdes loceklis Oskars Teikmanis<\/strong> sav\u0101 prezent\u0101cij\u0101 st\u0101st\u012bja par konkr\u0113ta m\u0101ksl\u012bg\u0101 intelekta veida – lielo valodas mode\u013cu – pielietojumu tehnisku projektu un publik\u0101ciju izstr\u0101d\u0101\u0161an\u0101, k\u0101 ar\u012b koda \u0123ener\u0113\u0161an\u0101.\u00a0 Ar lielajiem valodas mode\u013ciem ir iesp\u0113jams efekt\u012bvi kod\u0113t milz\u012bgu inform\u0101cijas daudzumu, kuram var piek\u013c\u016bt ar cilv\u0113ku veidot\u0101m uzvedn\u0113m (no ang\u013cu val. prompts<\/i>). \u0160\u0101du mode\u013cu piem\u0113ri ir LaMDA<\/strong> (Google<\/i>), LLaMA<\/strong> (Meta<\/i>), StableLM<\/strong> (Stability AI<\/i>) un GPT<\/strong> (OpenAI<\/i>). Katrs no \u0161iem mode\u013ciem veido pamatu da\u017e\u0101diem MI r\u012bkiem. Piem\u0113ram, Bard<\/i><\/a> izmanto LaMDA, savuk\u0101rt ChatGPT<\/i><\/a> – GPT arhitekt\u016bru.<\/p>\n Satura veido\u0161an\u0101 t\u0101di r\u012bki k\u0101 ChatGPT ir noder\u012bgi, mekl\u0113jot iedvesmu rakstiem, apzin\u0101ti \u0123ener\u0113jot idejas un analiz\u0113jot datus. Semin\u0101ra laik\u0101 Oskars demonstr\u0113ja piem\u0113ru ar br\u012bvi pieejamu datu kopu, un par\u0101d\u012bja, k\u0101 var izmantot ChatGPT un Google Colab<\/i><\/a>, lai analiz\u0113tu un vizualiz\u0113tu datus, izmantojot tikai vienu uzvedni. T\u0101d\u0113j\u0101di p\u0113tnieki var ietaup\u012bt savu darba laiku, kas cit\u0101di b\u016btu pat\u0113r\u0113ts, veicot datu grup\u0113\u0161anas funkcijas manu\u0101li. Savuk\u0101rt t\u0101di r\u012bki k\u0101 Consensus<\/i><\/a>, Tavily<\/i><\/a>, Perplexity<\/i><\/a> un Elicit<\/i><\/a> pied\u0101v\u0101 funkcijas zin\u0101tnisko rakstu anal\u012bzei un apkopo\u0161anai, nodro\u0161inot efekt\u012bv\u0101ku p\u0113tniec\u012bbas praksi. Clipdrop<\/i><\/a> var izmantot k\u0101 r\u012bku att\u0113lu p\u0101rveido\u0161anai, \u013caujot lietot\u0101jiem redi\u0123\u0113t, uzlabot un papildin\u0101t eso\u0161os att\u0113lus, izmantojot Stable Diffusion <\/i>mode\u013cu klasi.<\/p>\n Savas prezent\u0101cijas beig\u0101s Oskars dal\u012bj\u0101s ar st\u0101stu par Veru Moln\u0101ru<\/strong> (Vera Moln\u00e1r<\/i>), Ung\u0101rij\u0101 dzimu\u0161u m\u0101kslinieci, kura iev\u0113rojamu savas dz\u012bves da\u013cu pavad\u012bja Francij\u0101. \u0160\u012b m\u0101ksliniece str\u0101d\u0101ja ar datoru rad\u012btu m\u0101kslu ilgi pirms t\u0101 k\u013cuva par pla\u0161i izplat\u012btu par\u0101d\u012bbu. Vi\u0146a par\u0101d\u012bja, ka datorus var izmantot k\u0101 cilv\u0113ka rado\u0161uma papildin\u0101t\u0101jus, izmantojot nejau\u0161\u012bbas principu. L\u012bdz\u012bga nejau\u0161\u012bba figur\u0113 ar\u012b MI r\u012bkos, kas \u013cauj \u0123ener\u0113t negaid\u012btu saturu, kas var vest pie jaun\u0101m idej\u0101m. Vera Moln\u0101ra m\u016b\u017e\u012bb\u0101 dev\u0101s 2023.\u00a0 gada 7. decembr\u012b 99 gadu vecum\u0101.<\/p>\n <\/p>\n Otrais st\u0101sts: MI r\u012bku izmanto\u0161ana molekul\u0101r\u0101s biolo\u0123ijas eksperimentu pl\u0101no\u0161an\u0101 un dizain\u0101 (Eva Em\u012blija \u010cesle)<\/strong><\/p>\n LJZA biedre Eva Em\u012blija \u010cesle<\/strong> skaidroja, k\u0101 m\u0101ksl\u012bg\u0101 intelekta r\u012bki pal\u012bdz molekul\u0101r\u0101s biolo\u0123ijas eksperimentu pl\u0101no\u0161an\u0101 un izstr\u0101d\u0113, k\u0101 ar\u012b papla\u0161ina un uzlabo p\u0113tniec\u012bbas pieredzi struktur\u0101l\u0101s biolo\u0123ijas jom\u0101. Att\u012bstoties m\u0101ksl\u012bg\u0101 intelekta r\u012bku pielietojumam, arvien liel\u0101ku noz\u012bmi tie ieg\u016bst ar\u012b molekul\u0101r\u0101s biolo\u0123ijas p\u0113tniec\u012bb\u0101, p\u0113d\u0113jo gadu laik\u0101 b\u016btiski uzlabojot AlphaFold2<\/i><\/a> (AF2) m\u0101ksl\u012bg\u0101 intelekta sist\u0113mu. AlphaFold augst\u0101 l\u012bme\u0146a sp\u0113ja sal\u012bdzino\u0161i prec\u012bzi paredz\u0113t prote\u012bnu strukt\u016bras iev\u0113rojami main\u012bjusi biolo\u0123ijas un medic\u012bnas jomu, gadiem ilgu darbu prote\u012bnu strukt\u016bru eksperiment\u0101lai noteik\u0161anai nereti aizst\u0101jot ar AF2 mode\u013ca ieg\u016b\u0161anu nepilnas stundas laik\u0101. AF2 mode\u013ci nevar piln\u012bb\u0101 aizst\u0101t eksperiment\u0101los rezult\u0101tus, ta\u010du \u0161\u012bs sist\u0113mas pieejam\u012bba daudzos gad\u012bjumos \u013cauj vair\u0101k uzman\u012bbas piev\u0113rst prote\u012bnu funkciju p\u0113tniec\u012bbai, padarot p\u0113t\u012bjumu procesu \u0101tr\u0101ku un prec\u012bz\u0101ku.<\/p>\n Programma veidota k\u0101 dzi\u013c\u0101s m\u0101c\u012b\u0161an\u0101s sist\u0113ma.\u00a0 K\u0101 apgalvo Google Deepmind<\/i> l\u012bdzdibin\u0101t\u0101js un vad\u012bt\u0101js Demis Hassabis<\/i><\/strong>: \u201cM\u0101ksl\u012bgais intelekts var b\u016bt viena no vissvar\u012bg\u0101kaj\u0101m un izdev\u012bg\u0101kaj\u0101m tehnolo\u0123ij\u0101m, kas jebkad izgudrotas.\u201d<\/i><\/p>\n To, neap\u0161aub\u0101mi, pier\u0101da Evas st\u0101st\u012bjums par to, ka da\u017e\u0101du molekul\u0101r\u0101s biolo\u0123ijas eksperimentu pl\u0101no\u0161ana, izmantojot m\u0101ksl\u012bgo intelektu, nodro\u0161ina m\u0113r\u0137\u0113tu eksperimentu izstr\u0101des un iepriek\u0161 eksperiment\u0101li nenoskaidrotu prote\u012bnu vizualiz\u0113\u0161anas iesp\u0113jas. T\u0101d\u0113j\u0101di AF2 \u013cauj paredz\u0113t eksperiment\u0101lo rezult\u0101tu mijiedarb\u012bbas iesp\u0113jas. Eva uzsv\u0113ra, ka molekul\u0101r\u0101s biolo\u0123ijas\u00a0 jom\u0101 m\u0101ksl\u012bg\u0101 intelekta r\u012bku ir patiesi daudz, bet to izmanto\u0161anai j\u0101pieiet kritiski, \u0146emot v\u0113r\u0101 gan izstr\u0101des procesu gan ieg\u016bto rezult\u0101tu kvalit\u0101ti. Veiksm\u012bgi MI r\u012bki nodro\u0161ina b\u016btisku cilv\u0113kstundu skaita ietaup\u012bjumu ikdienas darb\u0101 ar\u012b eksperiment\u0101laj\u0101 proces\u0101.<\/p>\n <\/p>\n Tre\u0161ais st\u0101sts: M\u0101ksl\u012bgais intelekts, rado\u0161ums un izgl\u012bt\u012bba soci\u0101laj\u0101s un humanit\u0101raj\u0101s zin\u0101tn\u0113s (Ance Krist\u0101la)<\/strong><\/p>\n Ance Krist\u0101la<\/strong> savu prezent\u0101ciju bija struktur\u0113jusi tr\u012bs da\u013c\u0101s: p\u0113tniec\u012bba, izgl\u012bt\u012bba un rado\u0161ums. Ance skaidroja, k\u0101 rado\u0161a m\u0101ksl\u012bg\u0101 intelekta izmanto\u0161ana izgl\u012bt\u012bb\u0101 var veicin\u0101t cilv\u0113ka prasmju att\u012bst\u012bbu un k\u0101 izmantot m\u0101ksl\u012bgo intelektu datu anal\u012bz\u0113 un literat\u016bras avotu mekl\u0113\u0161an\u0101. Ance nor\u0101d\u012bja, ka soci\u0101laj\u0101m zin\u0101tn\u0113m ir noz\u012bm\u012bga loma, lai p\u0113t\u012btu t\u0101dus b\u016btiskus jaut\u0101jumus k\u0101 sabiedr\u012bbas izpratne par m\u0101ksl\u012bgo intelektu un t\u0101 ien\u0101k\u0161anu ikdien\u0101, k\u0101 un ko m\u0113s dom\u0101jam par autor\u012bbu, k\u0101 ar\u012b tehnolo\u0123iju ietekmi uz sabiedr\u012bbu. Datu anal\u012bz\u0113 m\u0101ksl\u012bgais intelekts noder, piem\u0113ram, datu strukturiz\u0113\u0161an\u0101, nodro\u0161inot automatiz\u0113tu datu kateogoriz\u0113\u0161anu un tematiz\u0113\u0161anu. Soci\u0101laj\u0101s zin\u0101tn\u0113s MI tehnolo\u0123ijas sniedz ar\u012b iesp\u0113ju kvantiat\u012bvi str\u0101d\u0101t ar kvalitat\u012bvajiem datiem, veidojot datu mode\u013cus un izstr\u0101d\u0101jot datu b\u0101zes.<\/p>\n Piem\u0113ri r\u012bkiem, ko izmantot datu anal\u012bz\u0113: ATLAS.ti<\/i><\/a>, akkademia<\/i><\/a> (GitHUB), voyant-tools.org<\/i><\/a>, essentia.upf.edu<\/i><\/a>.<\/p>\n M\u0101ksl\u012bg\u0101 intelekta r\u012bki izgl\u012bt\u012bb\u0101 noder literat\u016bras avotu mekl\u0113\u0161an\u0101, kopsavilkumu, prezent\u0101ciju un uzskates materi\u0101lu veido\u0161an\u0101. Ance ar\u012b uzsv\u0113ra akad\u0113misko god\u012bgumu un skaidroja ikdienas izaicin\u0101jumus darb\u0101 ar studentiem, kuri nereti v\u0113las iesniegt darbus, kas veidoti ar m\u0101ksl\u012bg\u0101 intelekta r\u012bku pal\u012bdz\u012bbu. Izmantojot maksl\u012bg\u0101 intelekta r\u012bkus, lai piln\u012bb\u0101 aizst\u0101tu cilv\u0113ka ieguld\u012bto darbu, ne vienm\u0113r tiek nodro\u0161in\u0101ta nepiecie\u0161am\u0101 m\u0101c\u012b\u0161an\u0101s pieredze. Tas t\u0101d\u0113\u013c, ka netiek veicin\u0101ta, piem\u0113ram, pa\u0161a studenta kritisk\u0101s dom\u0101\u0161anas att\u012bst\u012bba un rado\u0161ums, pa\u0161am patst\u0101v\u012bgi rakstot, analiz\u0113jot inform\u0101ciju, k\u0101 ar\u012b argument\u0113jot viedokli.<\/p>\n Piem\u0113ri r\u012bkiem, kurus var izmantot izgl\u012bt\u012bbas proces\u0101: connectedpapers.com<\/i><\/a>, \u00a0humata.ai<\/i><\/a>, chatpdf.com<\/i><\/a>, consesus.app<\/i><\/a>, perplexity.ai<\/i><\/a>.<\/p>\n Sava st\u0101st\u012bjuma nosl\u0113gum\u0101 Ance min\u0113ja da\u017e\u0101dus rado\u0161us risin\u0101jumus, kas paveikti ar MI r\u012bku pal\u012bdz\u012bbu. Viens no tiem ir iev\u0113rojams sasniegums m\u016bzik\u0101. Sadarbojoties m\u016bzikas v\u0113sturniekiem, komponistiem un datorzin\u0101tniekiem, tika pabeigta Ludviga van B\u0113thovena desmit\u0101 simfonija. Kad komponists 1827. gad\u0101 nomira, vi\u0146a desmit\u0101 simfonija palika nepabeigta, atst\u0101jot entuziastiem aizraujo\u0161u m\u012bklu. Centieni pabeigt \u0161o simfoniju s\u0101k\u0101s 2019. gad\u0101. Iniciat\u012bvas vad\u012bt\u0101js Matiass Roders<\/strong> (Matthias R\u00f6der<\/i>) iesaist\u012bja zin\u0101tniekus, kuri uz\u0146\u0113m\u0101s iem\u0101c\u012bt ma\u0161\u012bnai visu Ludviga van B\u0113thovena dai\u013cradi, k\u0101 ar\u012b atdarin\u0101t vi\u0146a rado\u0161o procesu. At\u0161\u0137ir\u012bb\u0101 no iepriek\u0161\u0113jiem m\u0113\u0123in\u0101jumiem rekonstru\u0113t ska\u0146darba fragmentus, \u0161\u012b projekta m\u0113r\u0137is bija rad\u012bt visu simfoniju, saglab\u0101jot B\u0113thovena stilu. Darba gait\u0101 bija vair\u0101ki izaicin\u0101jumi, piem\u0113ram, Ludviga van B\u0113thovena ski\u010du at\u0161ifr\u0113\u0161ana un MI r\u012bka m\u0101c\u012b\u0161ana p\u0101rv\u0113rst m\u016bzikas fr\u0101zes sare\u017e\u0123\u012bt\u0101s strukt\u016br\u0101s, harmoniz\u0113t melodijas un iev\u0113rot da\u017e\u0101das m\u016bzikas formas. Sadarb\u012bbas rezult\u0101t\u0101 tapa pilns B\u0113thovena desmit\u0101s simfonijas ieraksts, kas pirmo reizi tika prezent\u0113ts\u00a0 2021. gada 9. oktobr\u012b V\u0101cij\u0101, Bonn\u0101.<\/p>\n <\/p>\n Diskusija<\/strong><\/p>\n Semin\u0101ra nosl\u0113gum\u0101 visi tr\u012bs run\u0101t\u0101ji atbild\u0113ja uz pas\u0101kuma dal\u012bbnieku jaut\u0101jumiem. Viens no koment\u0101riem bija saist\u012bts ar m\u0101ksl\u012bg\u0101 intelekta izaicin\u0101jumu, piem\u0113ram, dzejas rad\u012b\u0161an\u0101. Ance nor\u0101d\u012bja, ka m\u0101ksl\u012bgajam intelektam var likt rakst\u012bt noteikt\u0101 stil\u0101, piem\u0113ram, k\u0101 \u0160eksp\u012bram. Tom\u0113r b\u016btu p\u0101ragri gaid\u012bt, ka m\u0101ksl\u012bgais intelekts\u00a0 sp\u0113s rad\u012bt dzeju ar emocij\u0101m.\u00a0Viedok\u013ci bija da\u017e\u0101di, tom\u0113r visi vienoj\u0101s, ka pieredzes nodo\u0161ana pasaulei caur p\u0101rdz\u012bvojumu, kuru cilv\u0113ks j\u016bt, b\u016bs nepiecie\u0161ams, radot m\u016bziku, dzeju vai k\u0101du citu m\u0101kslas darbu. Ar m\u0101ksl\u012bg\u0101 intelekta r\u012bkiem b\u016bs gr\u016bti aizst\u0101t \u0161o cilv\u0113ka v\u0113lmi pa\u0161am rad\u012bt. Semin\u0101ra dal\u012bbnieki aktualiz\u0113ja jaut\u0101jumu par skolot\u0101ja un m\u0101c\u012b\u0161an\u0101s lomu b\u0113rnu un jaunie\u0161u izgl\u012bt\u012bbas proces\u0101, jo m\u0101ksl\u012bg\u0101 intelekta r\u012bki pied\u0101v\u0101 gatavus risin\u0101jumus da\u017e\u0101du prasmju apg\u016b\u0161anai un pilnveido\u0161anai, kurus skolot\u0101jiem ir v\u0113rts izmantot m\u0101c\u012bbu proces\u0101.\u00a0Semin\u0101rs iedro\u0161in\u0101ja t\u0101 dal\u012bbniekus pam\u0113\u0123in\u0101t izmantot jaun\u0101kos pieejamos m\u0101ksl\u012bg\u0101 intelekta r\u012bkus sav\u0101 ikdien\u0101.<\/p>","protected":false},"excerpt":{"rendered":" Foto: Veras Moln\u0101ras darbs no Oskara Teikma\u0146a person\u012bg\u0101 arh\u012bva 2023. gada 14. decembr\u012b Latvijas Nacion\u0101laj\u0101 bibliot\u0113k\u0101 notika semin\u0101rs \u201cM\u0101ksl\u012bg\u0101 intelekta r\u012bku pielietojumi da\u017e\u0101d\u0101s zin\u0101tnes nozar\u0113s\u201d, kur\u0101 Latvijas Jauno zin\u0101tnieku apvien\u012bbas (LJZA) biedri dal\u012bj\u0101s ar savu pieredzi par m\u0101ksl\u012bg\u0101 intelekta r\u012bku izmanto\u0161anu zin\u0101tniskaj\u0101 darb\u0101. Semin\u0101rs notika cikla \u201cZin\u0101tne sabiedr\u012bbai\u201d ietvaros, kas tiek organiz\u0113ts sadarb\u012bb\u0101 ar […]<\/p>","protected":false},"author":10,"featured_media":7001,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[161],"tags":[],"class_list":["post-7000","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-pasakumi-aktualitates"],"acf":[],"_links":{"self":[{"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/posts\/7000","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/users\/10"}],"replies":[{"embeddable":true,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/comments?post=7000"}],"version-history":[{"count":5,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/posts\/7000\/revisions"}],"predecessor-version":[{"id":7291,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/posts\/7000\/revisions\/7291"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/media\/7001"}],"wp:attachment":[{"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/media?parent=7000"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/categories?post=7000"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jauniezinatnieki.lv\/en\/wp-json\/wp\/v2\/tags?post=7000"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}