{"id":13027,"date":"2024-06-23T10:10:33","date_gmt":"2024-06-23T10:10:33","guid":{"rendered":"https:\/\/dailyai.com\/?p=13027"},"modified":"2024-06-25T11:36:18","modified_gmt":"2024-06-25T11:36:18","slug":"university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur","status":"publish","type":"post","link":"https:\/\/dailyai.com\/sv\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/","title":{"rendered":"Studie fr\u00e5n University of Oxford identifierar n\u00e4r AI-hallucinationer \u00e4r mer sannolika att intr\u00e4ffa"},"content":{"rendered":"

I en studie fr\u00e5n University of Oxford har man utvecklat ett s\u00e4tt att testa n\u00e4r spr\u00e5kmodeller \u00e4r \"os\u00e4kra\" p\u00e5 sin output och riskerar att hallucinera.\u00a0<\/b><\/p>\n

AI-\"hallucinationer\" avser ett fenomen d\u00e4r stora spr\u00e5kmodeller (LLM) genererar flytande och trov\u00e4rdiga svar som inte \u00e4r sanningsenliga eller konsekventa.\u00a0<\/span><\/p>\n

Hallucinationer \u00e4r sv\u00e5ra - om inte om\u00f6jliga - att skilja fr\u00e5n AI-modeller. AI-utvecklare som OpenAI, Google och Anthropic har alla medgett att hallucinationer sannolikt kommer att f\u00f6rbli en biprodukt av att interagera med AI.\u00a0<\/span><\/p>\n

Som Dr. Sebastian Farquhar, en av studiens f\u00f6rfattare, <\/span>f\u00f6rklarar i ett blogginl\u00e4gg<\/span><\/a>\"LLM:er \u00e4r mycket skickliga p\u00e5 att s\u00e4ga samma sak p\u00e5 m\u00e5nga olika s\u00e4tt, vilket kan g\u00f6ra det sv\u00e5rt att avg\u00f6ra n\u00e4r de \u00e4r s\u00e4kra p\u00e5 ett svar och n\u00e4r de bokstavligen bara hittar p\u00e5 n\u00e5got.\"\u00a0<\/span><\/p>\n

Cambridge Dictionary har till och med lagt till en <\/span>AI-relaterad definition av ordet<\/span><\/a> \u00e5r 2023 och uts\u00e5g det till \"\u00c5rets ord\".\u00a0<\/span><\/p>\n

Detta University of Oxford <\/span> studie<\/span><\/a>, publicerad i Nature,<\/span> f\u00f6rs\u00f6ker svara p\u00e5 hur vi kan uppt\u00e4cka n\u00e4r det \u00e4r mest sannolikt att dessa hallucinationer intr\u00e4ffar.<\/span><\/p>\n

Den introducerar ett begrepp som kallas \"semantisk entropi\", som m\u00e4ter os\u00e4kerheten i en LLM:s resultat p\u00e5 betydelseniv\u00e5 snarare \u00e4n bara de specifika ord eller fraser som anv\u00e4nds.\u00a0<\/span><\/p>\n

Genom att ber\u00e4kna den semantiska entropin i en LLM:s svar kan forskarna uppskatta modellens f\u00f6rtroende f\u00f6r sina utdata och identifiera tillf\u00e4llen d\u00e5 den sannolikt hallucinerar.<\/span><\/p>\n

Semantisk entropi i LLM f\u00f6rklaras<\/h2>\n

Semantisk entropi, enligt studiens definition, m\u00e4ter os\u00e4kerheten eller inkonsekvensen i inneb\u00f6rden av en LLM:s svar. <\/span>Det hj\u00e4lper till att uppt\u00e4cka n\u00e4r en LLM kanske hallucinerar eller genererar otillf\u00f6rlitlig information.<\/span><\/p>\n

I enklare termer m\u00e4ter semantisk entropi hur \"f\u00f6rvirrad\" en LLM:s utdata \u00e4r.\u00a0<\/span><\/p>\n

LLM kommer sannolikt att ge tillf\u00f6rlitlig information om betydelsen av dess resultat \u00e4r n\u00e4ra relaterad och konsekvent. <\/span>Men om betydelserna \u00e4r utspridda och inkonsekventa \u00e4r det en varningssignal om att LLM kan hallucinera eller generera felaktig information.<\/span><\/p>\n

S\u00e5 h\u00e4r fungerar det:<\/span><\/p>\n

    \n
  1. Forskarna har aktivt uppmanat LLM att generera flera m\u00f6jliga svar p\u00e5 samma fr\u00e5ga. Detta uppn\u00e5s genom att mata fr\u00e5gan till LLM flera g\u00e5nger, varje g\u00e5ng med ett annat slumpm\u00e4ssigt fr\u00f6 eller en liten variation i inmatningen.<\/span><\/li>\n
  2. Semantisk entropi granskar svaren och grupperar dem med samma underliggande inneb\u00f6rd, \u00e4ven om de anv\u00e4nder olika ord eller formuleringar.<\/span><\/li>\n
  3. Om LLM \u00e4r s\u00e4ker p\u00e5 svaret b\u00f6r dess svar ha liknande inneb\u00f6rd, vilket resulterar i en l\u00e5g semantisk entropipo\u00e4ng. Detta tyder p\u00e5 att LLM:n f\u00f6rst\u00e5r informationen p\u00e5 ett tydligt och konsekvent s\u00e4tt.<\/span><\/li>\n
  4. Men om LLM \u00e4r os\u00e4ker eller f\u00f6rvirrad kommer svaren att ha en st\u00f6rre variation av betydelser, varav vissa kan vara inkonsekventa eller orelaterade till fr\u00e5gan. Detta resulterar i en h\u00f6g semantisk entropipo\u00e4ng, vilket indikerar att LLM kan hallucinera eller generera otillf\u00f6rlitlig information.<\/span><\/li>\n<\/ol>\n

    F\u00f6r att utv\u00e4rdera dess effektivitet till\u00e4mpade forskarna semantisk entropi p\u00e5 en rad olika fr\u00e5gor och svarsuppgifter. Detta omfattade riktm\u00e4rken som<\/span>\u00a0triviafr\u00e5gor, l\u00e4sf\u00f6rst\u00e5else, ordproblem och biografier.\u00a0<\/span><\/p>\n

    \u00d6verlag \u00f6vertr\u00e4ffade semantisk entropi befintliga metoder n\u00e4r det g\u00e4llde att uppt\u00e4cka n\u00e4r en LLM sannolikt skulle generera ett felaktigt eller inkonsekvent svar.<\/span><\/p>\n

    \"Hallucinationer\"
    H\u00f6g genomsnittlig semantisk entropi tyder p\u00e5 konfabulation (i huvudsak hallucinerade fakta som anges som verkliga), medan l\u00e5g entropi, trots varierande ordalydelse, tyder p\u00e5 en sannolikt sann faktoid. K\u00e4llan \u00e4r: Natur<\/a> (\u00f6ppen tillg\u00e5ng)<\/figcaption><\/figure>\n

    I diagrammet ovan kan du se hur vissa fr\u00e5gor f\u00e5r LLM att generera ett konfabulerat (felaktigt, hallucinatoriskt) svar. Den ger t.ex. en f\u00f6delsedag och f\u00f6delsem\u00e5nad f\u00f6r fr\u00e5gorna l\u00e4ngst ned i diagrammet n\u00e4r den information som kr\u00e4vs f\u00f6r att besvara dem inte fanns med i den ursprungliga informationen.<\/p>\n

    Konsekvenser av att uppt\u00e4cka hallucinationer<\/h2>\n

    Detta arbete kan bidra till att f\u00f6rklara hallucinationer och g\u00f6ra LLM:er mer tillf\u00f6rlitliga och trov\u00e4rdiga.\u00a0<\/span><\/p>\n

    Genom att tillhandah\u00e5lla ett s\u00e4tt att uppt\u00e4cka n\u00e4r en LLM \u00e4r os\u00e4ker eller ben\u00e4gen att hallucinera banar semantisk entropi v\u00e4g f\u00f6r att anv\u00e4nda dessa AI-verktyg i dom\u00e4ner med h\u00f6ga insatser d\u00e4r faktakorrekthet \u00e4r avg\u00f6rande, som sjukv\u00e5rd, juridik och ekonomi. <\/span><\/p>\n

    Felaktiga resultat kan f\u00e5 potentiellt katastrofala f\u00f6ljder n\u00e4r de p\u00e5verkar situationer med h\u00f6ga insatser, vilket visats av vissa misslyckad prediktiv polisverksamhet<\/a> och sjukv\u00e5rdssystem<\/a>.\u00a0<\/span><\/p>\n

    Men det \u00e4r ocks\u00e5 viktigt att komma ih\u00e5g att hallucinationer bara \u00e4r en typ av fel som LLM:er kan g\u00f6ra.\u00a0<\/span><\/p>\n

    Som Dr. Farquhar f\u00f6rklarar: \"Om en LLM g\u00f6r konsekventa misstag kommer den h\u00e4r nya metoden inte att f\u00e5nga upp det. De farligaste AI-misslyckandena intr\u00e4ffar n\u00e4r ett system g\u00f6r n\u00e5got d\u00e5ligt, men \u00e4r sj\u00e4lvs\u00e4kert och systematiskt. Det finns fortfarande mycket arbete kvar att g\u00f6ra.\"<\/span><\/p>\n

    Oxford-teamets semantiska entropimetod utg\u00f6r dock ett stort steg fram\u00e5t n\u00e4r det g\u00e4ller v\u00e5r f\u00f6rm\u00e5ga att f\u00f6rst\u00e5 och mildra begr\u00e4nsningarna i AI-spr\u00e5kmodeller.\u00a0<\/span><\/p>\n

    Genom att tillhandah\u00e5lla ett objektivt s\u00e4tt att uppt\u00e4cka dem kommer vi n\u00e4rmare en framtid d\u00e4r vi kan utnyttja AI:s potential och samtidigt se till att den f\u00f6rblir ett p\u00e5litligt och trov\u00e4rdigt verktyg i m\u00e4nsklighetens tj\u00e4nst.<\/p>","protected":false},"excerpt":{"rendered":"

    En studie fr\u00e5n University of Oxford har utvecklat ett s\u00e4tt att testa n\u00e4r spr\u00e5kmodeller \u00e4r \"os\u00e4kra\" p\u00e5 sin output och riskerar att hallucinera. AI-\"hallucinationer\" avser ett fenomen d\u00e4r stora spr\u00e5kmodeller (LLM) genererar flytande och rimliga svar som inte \u00e4r sanningsenliga eller konsekventa. Hallucinationer \u00e4r sv\u00e5ra - om inte om\u00f6jliga - att skilja fr\u00e5n AI-modeller. AI-utvecklare som OpenAI, Google och Anthropic har alla medgett att hallucinationer sannolikt kommer att f\u00f6rbli en biprodukt av att interagera med AI. Som Dr. Sebastian Farquhar, en av studiens f\u00f6rfattare, f\u00f6rklarar i ett blogginl\u00e4gg: \"LLM:er \u00e4r mycket kapabla att s\u00e4ga samma sak<\/p>","protected":false},"author":2,"featured_media":13029,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[88],"tags":[480,105],"class_list":["post-13027","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ethics","tag-hallucinations","tag-machine-learning"],"yoast_head":"\nUniversity of Oxford study identifies when AI hallucinations are more likely to occur | DailyAI<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/dailyai.com\/sv\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/\" \/>\n<meta property=\"og:locale\" content=\"sv_SE\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"University of Oxford study identifies when AI hallucinations are more likely to occur | DailyAI\" \/>\n<meta property=\"og:description\" content=\"A University of Oxford study developed a means of testing when language models are \u201cunsure\u201d of their output and risk hallucinating.\u00a0 AI “hallucinations” refer to a phenomenon where large language models (LLMs) generate fluent and plausible responses that are not truthful or consistent.\u00a0 Hallucinations are tough \u2013 if not impossible \u2013 to separate from AI models. AI developers like OpenAI, Google, and Anthropic have all admitted that hallucinations will likely remain a byproduct of interacting with AI.\u00a0 As Dr. Sebastian Farquhar, one of the study’s authors, explains in a blog post, “LLMs are highly capable of saying the same thing\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/sv\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2024-06-23T10:10:33+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-06-25T11:36:18+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1792\" \/>\n\t<meta property=\"og:image:height\" content=\"1024\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Sam Jeans\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:site\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:label1\" content=\"Skriven av\" \/>\n\t<meta name=\"twitter:data1\" content=\"Sam Jeans\" \/>\n\t<meta name=\"twitter:label2\" content=\"Ber\u00e4knad l\u00e4stid\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minuter\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/\"},\"author\":{\"name\":\"Sam Jeans\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\"},\"headline\":\"University of Oxford study identifies when AI hallucinations are more likely to occur\",\"datePublished\":\"2024-06-23T10:10:33+00:00\",\"dateModified\":\"2024-06-25T11:36:18+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/\"},\"wordCount\":813,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp\",\"keywords\":[\"Hallucinations\",\"machine learning\"],\"articleSection\":[\"Ethics & Society\"],\"inLanguage\":\"sv-SE\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/\",\"name\":\"University of Oxford study identifies when AI hallucinations are more likely to occur | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp\",\"datePublished\":\"2024-06-23T10:10:33+00:00\",\"dateModified\":\"2024-06-25T11:36:18+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#breadcrumb\"},\"inLanguage\":\"sv-SE\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"sv-SE\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp\",\"width\":1792,\"height\":1024,\"caption\":\"hallucinations\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/06\\\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"University of Oxford study identifies when AI hallucinations are more likely to occur\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"name\":\"DailyAI\",\"description\":\"Your Daily Dose of AI News\",\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/dailyai.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"sv-SE\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"sv-SE\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"width\":4501,\"height\":934,\"caption\":\"DailyAI\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/DailyAIOfficial\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/dailyaiofficial\\\/\",\"https:\\\/\\\/www.youtube.com\\\/@DailyAIOfficial\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\",\"name\":\"Sam Jeans\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"sv-SE\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"caption\":\"Sam Jeans\"},\"description\":\"Sam is a science and technology writer who has worked in various AI startups. When he\u2019s not writing, he can be found reading medical journals or digging through boxes of vinyl records.\",\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/sam-jeans-6746b9142\\\/\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/sv\\\/author\\\/samjeans\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Studie fr\u00e5n University of Oxford identifierar n\u00e4r AI-hallucinationer \u00e4r mer sannolika att intr\u00e4ffa | DailyAI","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/dailyai.com\/sv\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/","og_locale":"sv_SE","og_type":"article","og_title":"University of Oxford study identifies when AI hallucinations are more likely to occur | DailyAI","og_description":"A University of Oxford study developed a means of testing when language models are \u201cunsure\u201d of their output and risk hallucinating.\u00a0 AI “hallucinations” refer to a phenomenon where large language models (LLMs) generate fluent and plausible responses that are not truthful or consistent.\u00a0 Hallucinations are tough \u2013 if not impossible \u2013 to separate from AI models. AI developers like OpenAI, Google, and Anthropic have all admitted that hallucinations will likely remain a byproduct of interacting with AI.\u00a0 As Dr. Sebastian Farquhar, one of the study’s authors, explains in a blog post, “LLMs are highly capable of saying the same thing","og_url":"https:\/\/dailyai.com\/sv\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/","og_site_name":"DailyAI","article_published_time":"2024-06-23T10:10:33+00:00","article_modified_time":"2024-06-25T11:36:18+00:00","og_image":[{"width":1792,"height":1024,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp","type":"image\/webp"}],"author":"Sam Jeans","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"Skriven av":"Sam Jeans","Ber\u00e4knad l\u00e4stid":"4 minuter"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/"},"author":{"name":"Sam Jeans","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9"},"headline":"University of Oxford study identifies when AI hallucinations are more likely to occur","datePublished":"2024-06-23T10:10:33+00:00","dateModified":"2024-06-25T11:36:18+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/"},"wordCount":813,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp","keywords":["Hallucinations","machine learning"],"articleSection":["Ethics & Society"],"inLanguage":"sv-SE"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/","url":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/","name":"Studie fr\u00e5n University of Oxford identifierar n\u00e4r AI-hallucinationer \u00e4r mer sannolika att intr\u00e4ffa | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp","datePublished":"2024-06-23T10:10:33+00:00","dateModified":"2024-06-25T11:36:18+00:00","breadcrumb":{"@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#breadcrumb"},"inLanguage":"sv-SE","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/"]}]},{"@type":"ImageObject","inLanguage":"sv-SE","@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/06\/DALL\u00b7E-2024-06-23-11.10.02-A-surreal-and-futuristic-depiction-of-a-face-experiencing-AI-hallucinations.-The-face-appears-to-be-merging-with-digital-elements-with-parts-of-it-di.webp","width":1792,"height":1024,"caption":"hallucinations"},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2024\/06\/university-of-oxford-study-identifies-when-ai-hallucinations-are-more-likely-to-occur\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"University of Oxford study identifies when AI hallucinations are more likely to occur"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DagligaAI","description":"Din dagliga dos av AI-nyheter","publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/dailyai.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"sv-SE"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DagligaAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"sv-SE","@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","width":4501,"height":934,"caption":"DailyAI"},"image":{"@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/DailyAIOfficial","https:\/\/www.linkedin.com\/company\/dailyaiofficial\/","https:\/\/www.youtube.com\/@DailyAIOfficial"]},{"@type":"Person","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9","name":"Sam Jeans","image":{"@type":"ImageObject","inLanguage":"sv-SE","@id":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","caption":"Sam Jeans"},"description":"Sam \u00e4r en vetenskaps- och teknikskribent som har arbetat i olika AI-startups. N\u00e4r han inte skriver l\u00e4ser han medicinska tidskrifter eller gr\u00e4ver igenom l\u00e5dor med vinylskivor.","sameAs":["https:\/\/www.linkedin.com\/in\/sam-jeans-6746b9142\/"],"url":"https:\/\/dailyai.com\/sv\/author\/samjeans\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/posts\/13027","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/comments?post=13027"}],"version-history":[{"count":10,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/posts\/13027\/revisions"}],"predecessor-version":[{"id":13087,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/posts\/13027\/revisions\/13087"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/media\/13029"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/media?parent=13027"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/categories?post=13027"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/sv\/wp-json\/wp\/v2\/tags?post=13027"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}