{"id":2201,"date":"2024-05-27T07:32:16","date_gmt":"2024-05-27T07:32:16","guid":{"rendered":"https:\/\/howtogeek.blog\/lv\/?p=2201"},"modified":"2024-05-27T07:32:16","modified_gmt":"2024-05-27T07:32:16","slug":"googles-ai-search-tells-users-to-eat-rocks-and-put-glue-on-pizza-lv","status":"publish","type":"post","link":"https:\/\/howtogeek.blog\/lv\/googles-ai-search-tells-users-to-eat-rocks-and-put-glue-on-pizza-lv\/","title":{"rendered":"Google AI mekl\u0113\u0161ana liek lietot\u0101jiem \u0113st akme\u0146us un uzkl\u0101t picu l\u012bmi"},"content":{"rendered":"

Tehnolo\u0123iju uz\u0146\u0113mumi vair\u0101k nek\u0101 jebkad agr\u0101k piev\u0113r\u0161 uzman\u012bbu m\u0101ksl\u012bgajam intelektam, k\u0101 rezult\u0101t\u0101 AI mode\u013ci ir nep\u0101rtraukti uzlaboju\u0161ies un k\u013c\u016bst lab\u0101ki da\u017e\u0101du uzdevumu veik\u0161an\u0101. Piem\u0113ram, OpenAI nesen demonstr\u0113ja savu jaun\u0101ko vado\u0161o modeli GPT-4o, kas pat sp\u0113j noteikt un simul\u0113t cilv\u0113ka emocijas. Jaunais multimod\u0101lais modelis var apkopot datus un sniegt inform\u0101ciju, izmantojot vizu\u0101l\u0101s un audio iesp\u0113jas papildus teksta iesp\u0113j\u0101m.<\/p>\n

Tom\u0113r ne visur t\u0101 ir laba zi\u0146a, jo \u0161ie mode\u013ci joproj\u0101m var k\u013c\u016bd\u012bties un pied\u0101v\u0101t nepareizu inform\u0101ciju un ieteikumus. Jaun\u0101kais piem\u0113rs tam ir Google AI p\u0101rskati, ko mekl\u0113\u0161anas gigants atkl\u0101ja \u0161\u012b m\u0113ne\u0161a s\u0101kum\u0101. Tas bija paredz\u0113ts, lai sniegtu AI \u0123ener\u0113tus lietot\u0101ju mekl\u0113t\u0101s inform\u0101cijas kopsavilkumus. Faktiski \u0161\u012b funkcija izr\u0101d\u0101s neuzticama, jo AI lietot\u0101jiem ir pied\u0101v\u0101jis faktiski nepareizas atbildes un d\u012bvainus ieteikumus, kuriem nav j\u0113gas.<\/p>\n

AI p\u0101rskati pied\u0101v\u0101 d\u012bvainus ieteikumus\u00a0\u2014 internets ir sadal\u012bts<\/h2>\n

Google AI p\u0101rskati ir paredz\u0113ti, lai samazin\u0101tu p\u016bles, kas saist\u012btas ar inform\u0101cijas mekl\u0113\u0161anu, nodro\u0161inot AI \u0123ener\u0113tus inform\u0101cijas kopsavilkumus, kas sav\u0101kti no da\u017e\u0101d\u0101m lap\u0101m. Probl\u0113ma ir t\u0101, ka AI pa\u0161laik nevar noteikt, kur\u0161 avots sniedz ticamu un prec\u012bzu inform\u0101ciju, kas noz\u012bm\u0113, ka tas var izveidot kopsavilkumus, izmantojot nepatiesus datus.<\/p>\n

Turkl\u0101t \u0161\u0137iet, ka tas nav lab\u0101kais, lai noteiktu lietot\u0101ja nol\u016bku. Piem\u0113ram, tas ierosin\u0101ja, ka \u0123eologi iesaka ap\u0113st vienu mazu akmeni dien\u0101 uz lietot\u0101ja vaic\u0101jumu, ac\u012bmredzot savu atbildi pamatojot ar inform\u0101ciju no humora vietnes The Onion.<\/p>\n

Turpin\u0101jum\u0101 tika nor\u0101d\u012bts, ka akme\u0146i ir b\u016btisks miner\u0101lvielu un vitam\u012bnu avots, kas nepiecie\u0161ami lab\u0101kai gremo\u0161anas vesel\u012bbai, un pat ieteica sl\u0113pt akme\u0146us un o\u013cus p\u0101rtikas produktos, piem\u0113ram, sald\u0113jum\u0101. T\u0101pat uz citu jaut\u0101jumu par to, k\u0101 pan\u0101kt, lai siers lab\u0101k piel\u012bp pie picas, Google AI ieteica uzkl\u0101t l\u012bmi, lai palielin\u0101tu lip\u012bgumu.<\/p>\n

\"\"
Avots: X<\/figcaption><\/figure>\n

Uz jaut\u0101jumu par ASV prezidentu musulma\u0146u skaitu, t\u0101 nor\u0101d\u012bja, ka Baraks Obama bija vien\u012bgais musulma\u0146u prezidents, kas faktiski nav pareizi, jo vi\u0146\u0161 ir kristietis. Un uz jaut\u0101jumu par nierakme\u0146u izdal\u012b\u0161anos AI atbild\u0113ja, ka ik p\u0113c 24 stund\u0101m ieteicams izdzert divus litrus ur\u012bna.<\/p>\n

Google m\u0101ksl\u012bgais intelekts ar\u012b pazi\u0146oja, ka suns ir sp\u0113l\u0113jis NBA un pat sp\u0113jis izveidot jaunu m\u0113r\u012b\u0161anas veidu, ko sauc par \u201ckilotom\u0101tu\u201d, kad vi\u0146am jaut\u0101ja par \u0161\u012bm t\u0113m\u0101m. Ir vair\u0101ki citi piem\u0113ri, kas ilustr\u0113, k\u0101 AI p\u0101rskati var sniegt un ir sniegu\u0161i nepareizu inform\u0101ciju.<\/p>\n

\"\"
Avots: Sherr\/CNET<\/figcaption><\/figure>\n

Kam\u0113r internet\u0101 notiek jautras atbildes, \u0161\u012bs atbildes var b\u016bt ar\u012b b\u012bstamas. Vien\u0101 mirkl\u012b Google AI p\u0101rskati sniedza lietot\u0101jam jaut\u0101jumu par depresiju, nolecot no Zelta v\u0101rtu tilta k\u0101 vienu no iesp\u0113jamiem ieteikumiem.<\/p>\n

\"\"
Avots: Reddit<\/figcaption><\/figure>\n

K\u0101da ir Google atbilde?<\/h2>\n

Lai risin\u0101tu sare\u017e\u0123\u012bto situ\u0101ciju, ko rad\u012bju\u0161i AI p\u0101rskati, Google saka, ka t\u0101 \u0101tri r\u012bkojas, lai labotu k\u013c\u016bdas faktiskaj\u0101 inform\u0101cij\u0101. T\u0101 ar\u012b izmantos \u0161os gad\u012bjumus, lai uzlabotu savu AI, lai samazin\u0101tu \u0161\u0101du incidentu iesp\u0113jam\u012bbu.<\/p>\n

Tom\u0113r uz\u0146\u0113mums apgalvo, ka AI liel\u0101koties darbojas t\u0101, k\u0101 vajadz\u0113tu, un \u0161\u012bs nepareiz\u0101s atbildes rad\u0101s politikas p\u0101rk\u0101pumu un \u013coti neparastu lietot\u0101ju vaic\u0101jumu d\u0113\u013c. Tie neatspogu\u013co pieredzi, kas liel\u0101kajai da\u013cai cilv\u0113ku ir bijusi ar AI p\u0101rskatiem. Google ar\u012b pazi\u0146oja, ka daudzi piem\u0113ri, kuros par\u0101d\u012btas nepareizas vai d\u012bvainas atbildes, tika izstr\u0101d\u0101ti, un tas nevar\u0113ja reproduc\u0113t l\u012bdz\u012bgus rezult\u0101tus, p\u0101rbaudot AI iek\u0161\u0113ji.<\/p>\n

M\u0101ksl\u012bg\u0101 intelekta ierobe\u017eojumi<\/h2>\n

Kop\u0161 OpenAI ChatGPT palai\u0161anas pirms diviem gadiem AI un ar to saist\u012bt\u0101s tehnolo\u0123ijas ir g\u0101ju\u0161as garu ce\u013cu. Vi\u0146i ir sp\u0113ju\u0161i lab\u0101k noteikt, ko lietot\u0101ji mekl\u0113, un sniegt atbilsto\u0161\u0101kas atbildes. Rezult\u0101t\u0101 arvien vair\u0101k pat\u0113r\u0113t\u0101jiem paredz\u0113tu produktu tagad tiek integr\u0113ti tehnolo\u0123ij\u0101.<\/p>\n

Lai gan tas var pal\u012bdz\u0113t ietaup\u012bt laiku un p\u016bles, mekl\u0113jot inform\u0101ciju vai veidojot saturu, ir svar\u012bgi saprast, ka AI joproj\u0101m ir noteikti ierobe\u017eojumi. Pirmk\u0101rt un galvenok\u0101rt, AI mode\u013ciem joproj\u0101m ir tendence uz halucin\u0101cij\u0101m, kas noz\u012bm\u0113, ka tie var izgudrot faktus un datus, kas nav patiesi, m\u0113\u0123inot atbild\u0113t uz lietot\u0101ja vaic\u0101jumu.<\/p>\n

Turkl\u0101t, k\u0101 min\u0113ts iepriek\u0161, pat gad\u012bjumos, kad m\u0101ksl\u012bgais intelekts neizgudro savus faktus, tas var ieg\u016bt inform\u0101ciju no neuzticamas vietas. Tas atkal var ietekm\u0113t lietot\u0101ja inform\u0101ciju, ja mekl\u0113\u0161anas rezult\u0101tos k\u0101 pareiz\u0101 atbilde par\u0101d\u0101s nepatiesa inform\u0101cija. T\u0101p\u0113c gandr\u012bz katrs uz\u0146\u0113mums savos AI r\u012bkos tagad par\u0101da br\u012bdin\u0101jumu, ka AI sniegt\u0101 inform\u0101cija var neatbilst paties\u012bbai.<\/p>\n

Lai gan d\u012bvain\u0101s atbildes, ko sniedz Google AI p\u0101rskati, var b\u016bt jautri las\u012bt, tas rada nopietnu jaut\u0101jumu par AI mode\u013cu uzticam\u012bbu kopum\u0101. Tas var izrais\u012bt nopietnas k\u013c\u016bdas, ja persona pa\u013caujas uz nepareizu AI sniegto inform\u0101ciju un nevar noteikt, vai t\u0101 ir pareiza vai n\u0113.<\/p>\n

Turkl\u0101t Google tagad ne\u013cauj lietot\u0101jiem piln\u012bb\u0101 izsl\u0113gt AI p\u0101rskatus, t\u0101p\u0113c \u0161\u012b funkcija ir \u0161eit, lai paliktu, kas ir v\u0113l viena probl\u0113mas da\u013ca. Tom\u0113r varat p\u0101riet uz sava Google konta iestat\u012bjumiem un to atsp\u0113jot laboratorij\u0101s, k\u0101 m\u0113s paskaidroj\u0101m t\u0101l\u0101k eso\u0161aj\u0101 rokasgr\u0101mat\u0101. Atbil\u017eu mekl\u0113\u0161ana da\u017e\u0101d\u0101s lap\u0101s var b\u016bt l\u0113n\u0101ka, ta\u010du ir maz\u0101ka iesp\u0113ja saskarties ar izdom\u0101tiem faktiem un d\u012bvainiem ieteikumiem.<\/p>\n","protected":false},"excerpt":{"rendered":"

Tehnolo\u0123iju uz\u0146\u0113mumi vair\u0101k nek\u0101 jebkad agr\u0101k piev\u0113r\u0161 uzman\u012bbu m\u0101ksl\u012bgajam intelektam, k\u0101 rezult\u0101t\u0101 AI mode\u013ci ir nep\u0101rtraukti uzlaboju\u0161ies un k\u013c\u016bst lab\u0101ki da\u017e\u0101du uzdevumu veik\u0161an\u0101. Piem\u0113ram, OpenAI nesen demonstr\u0113ja savu jaun\u0101ko vado\u0161o modeli GPT-4o, kas pat sp\u0113j noteikt un simul\u0113t cilv\u0113ka emocijas. Jaunais multimod\u0101lais modelis var apkopot datus un sniegt inform\u0101ciju, izmantojot vizu\u0101l\u0101s un audio iesp\u0113jas papildus […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[182,21,130],"class_list":["post-2201","post","type-post","status-publish","format-standard","hentry","category-how-to","tag-artificial-intelligence","tag-google","tag-microsoft"],"acf":[],"_links":{"self":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2201","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/comments?post=2201"}],"version-history":[{"count":1,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2201\/revisions"}],"predecessor-version":[{"id":2202,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2201\/revisions\/2202"}],"wp:attachment":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/media?parent=2201"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/categories?post=2201"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/tags?post=2201"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}