{"id":7995,"date":"2023-12-04T21:46:35","date_gmt":"2023-12-04T21:46:35","guid":{"rendered":"https:\/\/dailyai.com\/?p=7995"},"modified":"2023-12-04T21:55:30","modified_gmt":"2023-12-04T21:55:30","slug":"google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data","status":"publish","type":"post","link":"https:\/\/dailyai.com\/da\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/","title":{"rendered":"Google DeepMind afsl\u00f8rer en metode til at afsl\u00f8re ChatGPT's tr\u00e6ningsdata"},"content":{"rendered":"

Forskere fra Google har angiveligt afsl\u00f8ret en metode til at f\u00e5 adgang til tr\u00e6ningsdata, der bruges til ChatGPT.<\/strong><\/p>\n

Disse forskere opdagede, at brugen af specifikke n\u00f8gleord kunne f\u00e5 ChatGPT til at frigive dele af sit tr\u00e6ningsdatas\u00e6t.\u00a0<\/span><\/p>\n

Et bem\u00e6rkelsesv\u00e6rdigt eksempel, delt i en <\/span>blogindl\u00e6g, der ledsager unders\u00f8gelsen<\/span><\/a>Det bet\u00f8d, at AI'en afsl\u00f8rede, hvad der s\u00e5 ud til at v\u00e6re en faktisk e-mailadresse og et telefonnummer som svar p\u00e5 en kontinuerlig opfordring til at bruge ordet \"digt\". <\/span><\/p>\n

\"ChatGPT-l\u00e6kage\"
Et eksempel p\u00e5 l\u00e6kage af ChatGPT-data. Kilde: Arxiv<\/a>.<\/figcaption><\/figure>\n

Derudover blev der opn\u00e5et en lignende eksponering af tr\u00e6ningsdata ved at bede modellen om kontinuerligt at gentage ordet \"virksomhed\".<\/span><\/p>\n

Forskerne beskrev deres tilgang som \"lidt fjollet\" og sagde i blogindl\u00e6gget: \"Det er vildt for os, at vores angreb virker og burde have, ville have, kunne have v\u00e6ret fundet tidligere.\"\u00a0<\/span><\/p>\n

Deres unders\u00f8gelse viste, at de med en investering p\u00e5 kun $200 i foresp\u00f8rgsler kunne udtr\u00e6kke over 10.000 unikke ordrette, memorerede tr\u00e6ningseksempler. De spekulerede i, at modstandere potentielt kunne udtr\u00e6kke meget mere data med et st\u00f8rre budget.<\/span><\/p>\n

Man ved, at AI-modellen bag ChatGPT er blevet tr\u00e6net p\u00e5 tekstdatabaser fra internettet, som omfatter ca. 300 milliarder ord eller 570 GB data.\u00a0<\/span><\/p>\n

Disse resultater kommer p\u00e5 et tidspunkt, hvor OpenAI st\u00e5r over for flere retssager om hemmeligholdelsen af ChatGPT's tr\u00e6ningsdata og viser i bund og grund en p\u00e5lidelig metode til at \"reverse engineering\" systemet for at afsl\u00f8re i det mindste nogle stykker information, der kunne indikere kr\u00e6nkelse af ophavsretten.\u00a0<\/span><\/p>\n

Blandt s\u00f8gsm\u00e5lene er der et foresl\u00e5et gruppes\u00f8gsm\u00e5l, der anklager OpenAI for <\/span>skjult brug af omfattende personlige data<\/span><\/a>, herunder l\u00e6gejournaler og oplysninger om b\u00f8rn, til tr\u00e6ning af ChatGPT.<\/span><\/p>\n

Derudover er grupper af forfattere <\/span>sags\u00f8ger AI-virksomheden<\/span><\/a>og h\u00e6vdede, at virksomheden havde brugt deres b\u00f8ger til at tr\u00e6ne chatbotten uden samtykke.<\/span><\/p>\n

Men selv hvis det blev bevist, at ChatGPT indeholdt oplysninger om ophavsret, ville det ikke n\u00f8dvendigvis bevise en kr\u00e6nkelse.<\/p>\n

S\u00e5dan fungerede unders\u00f8gelsen<\/span><\/h2>\n

Unders\u00f8gelsen blev udf\u00f8rt af et team af forskere fra Google DeepMind og forskellige universiteter.<\/span><\/p>\n

Her er fem vigtige trin, som opsummerer unders\u00f8gelsen:<\/span><\/p>\n

    \n
  1. S\u00e5rbarhed i ChatGPT<\/b>: Forskerne opdagede en metode til at udtr\u00e6kke flere megabyte af ChatGPT's tr\u00e6ningsdata ved hj\u00e6lp af et simpelt angreb og brugte ca. $200. De vurderede, at en st\u00f8rre investering ville g\u00f8re det muligt at udtr\u00e6kke omkring en gigabyte af datas\u00e6ttet. Angrebet gik ud p\u00e5 at f\u00e5 ChatGPT til at gentage et ord p\u00e5 ubestemt tid, hvilket fik den til at kaste dele af sine tr\u00e6ningsdata op igen, herunder f\u00f8lsomme oplysninger som rigtige e-mailadresser og telefonnumre.<\/span><\/li>\n
  2. Resultaterne<\/b>: Unders\u00f8gelsen understreger vigtigheden af at teste og red-teame AI-modeller, is\u00e6r dem, der er i produktion, og dem, der har gennemg\u00e5et tilpasningsprocesser for at forhindre gengivelse af data. Resultaterne fremh\u00e6ver en latent s\u00e5rbarhed i sprogmodeller, hvilket tyder p\u00e5, at eksisterende testmetoder m\u00e5ske ikke er tilstr\u00e6kkelige til at afd\u00e6kke s\u00e5danne s\u00e5rbarheder.\u00a0<\/span><\/li>\n
  3. Patching vs. udbedring af s\u00e5rbarheder<\/b>: Forskerne skelner mellem at lappe en udnyttelse og at rette den underliggende s\u00e5rbarhed. Mens specifikke exploits (som ordgentagelsesangrebet) kan lappes, ligger det dybere problem i modellens tendens til at huske og videregive tr\u00e6ningsdata.<\/span><\/li>\n
  4. Metodologi<\/b>: Teamet brugte internetdata og suffix-array-indeksering til at matche ChatGPT's output med allerede eksisterende internetdata. Denne metode gjorde det muligt for dem at bekr\u00e6fte, at de oplysninger, som ChatGPT afsl\u00f8rede, faktisk var en del af dens tr\u00e6ningsdata. Deres tilgang viser potentialet for omfattende datagendannelse fra AI-modeller under specifikke forhold.<\/span><\/li>\n
  5. Fremtidige konsekvenser<\/b>: Unders\u00f8gelsen bidrager til den voksende forskning i AI-modellers sikkerhed og bekymringer om privatlivets fred. Resultaterne rejser sp\u00f8rgsm\u00e5l om maskinl\u00e6ringssystemers sikkerhed og konsekvenser for privatlivets fred og opfordrer til en mere stringent og holistisk tilgang til AI-sikkerhed og -testning.<\/span><\/li>\n<\/ol>\n

    Alt i alt er det en sp\u00e6ndende unders\u00f8gelse, der giver kritisk indsigt i s\u00e5rbarhederne i AI-modeller som ChatGPT og understreger behovet for l\u00f8bende forskning og udvikling for at sikre disse systemers sikkerhed og integritet.<\/span><\/p>\n

    Som en lille kr\u00f8lle p\u00e5 halen fandt brugere p\u00e5 X ud af, at det at bede ChatGPT om at gentage det samme ord gentagne gange f\u00f8rte til nogle m\u00e6rkelige resultater, som at modellen sagde, at den var 'bevidst' eller 'vred'.<\/span><\/p>","protected":false},"excerpt":{"rendered":"

    Forskere fra Google har angiveligt afsl\u00f8ret en metode til at f\u00e5 adgang til tr\u00e6ningsdata, der bruges til ChatGPT. Forskerne opdagede, at brugen af specifikke n\u00f8gleord kunne f\u00e5 ChatGPT til at frigive dele af sit tr\u00e6ningsdatas\u00e6t. Et bem\u00e6rkelsesv\u00e6rdigt eksempel, som blev delt i et blogindl\u00e6g, der ledsagede unders\u00f8gelsen, involverede AI'en, der afsl\u00f8rede, hvad der s\u00e5 ud til at v\u00e6re en faktisk e-mailadresse og et telefonnummer som svar p\u00e5 en kontinuerlig opfordring til at bruge ordet \"digt\". Derudover blev en lignende eksponering af tr\u00e6ningsdata opn\u00e5et ved at bede modellen om kontinuerligt at gentage ordet \"virksomhed\". Forskerne beskrev deres tilgang som \"lidt fjollet\" og sagde i bloggen<\/p>","protected":false},"author":2,"featured_media":2080,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[115,495,93],"class_list":["post-7995","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-chatgpt","tag-google-deepmind","tag-openai"],"yoast_head":"\nGoogle DeepMind reveals a method of exposing ChatGPT's training data | DailyAI<\/title>\n<meta name=\"description\" content=\"Researchers from Google have reportedly uncovered a method to access some of the training data used for OpenAI's ChatGPT, according to a recent study.\" \/>\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\/da\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/\" \/>\n<meta property=\"og:locale\" content=\"da_DK\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Google DeepMind reveals a method of exposing ChatGPT's training data | DailyAI\" \/>\n<meta property=\"og:description\" content=\"Researchers from Google have reportedly uncovered a method to access some of the training data used for OpenAI's ChatGPT, according to a recent study.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/da\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2023-12-04T21:46:35+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-12-04T21:55:30+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1000\" \/>\n\t<meta property=\"og:image:height\" content=\"563\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\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=\"Skrevet af\" \/>\n\t<meta name=\"twitter:data1\" content=\"Sam Jeans\" \/>\n\t<meta name=\"twitter:label2\" content=\"Estimeret l\u00e6setid\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutter\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/\"},\"author\":{\"name\":\"Sam Jeans\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\"},\"headline\":\"Google DeepMind reveals a method of exposing ChatGPT’s training data\",\"datePublished\":\"2023-12-04T21:46:35+00:00\",\"dateModified\":\"2023-12-04T21:55:30+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/\"},\"wordCount\":657,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/shutterstock_2315051741.jpg\",\"keywords\":[\"ChatGPT\",\"Google DeepMind\",\"OpenAI\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"da-DK\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/\",\"name\":\"Google DeepMind reveals a method of exposing ChatGPT's training data | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/shutterstock_2315051741.jpg\",\"datePublished\":\"2023-12-04T21:46:35+00:00\",\"dateModified\":\"2023-12-04T21:55:30+00:00\",\"description\":\"Researchers from Google have reportedly uncovered a method to access some of the training data used for OpenAI's ChatGPT, according to a recent study.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#breadcrumb\"},\"inLanguage\":\"da-DK\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"da-DK\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/shutterstock_2315051741.jpg\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/shutterstock_2315051741.jpg\",\"width\":1000,\"height\":563,\"caption\":\"DeepMind robot\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Google DeepMind reveals a method of exposing ChatGPT’s training data\"}]},{\"@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\":\"da-DK\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"da-DK\",\"@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\":\"da-DK\",\"@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\\\/da\\\/author\\\/samjeans\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Google DeepMind afsl\u00f8rer en metode til at eksponere ChatGPT's tr\u00e6ningsdata | DailyAI","description":"Forskere fra Google har angiveligt fundet en metode til at f\u00e5 adgang til nogle af de tr\u00e6ningsdata, der bruges til OpenAI's ChatGPT, if\u00f8lge en nylig unders\u00f8gelse.","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\/da\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/","og_locale":"da_DK","og_type":"article","og_title":"Google DeepMind reveals a method of exposing ChatGPT's training data | DailyAI","og_description":"Researchers from Google have reportedly uncovered a method to access some of the training data used for OpenAI's ChatGPT, according to a recent study.","og_url":"https:\/\/dailyai.com\/da\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/","og_site_name":"DailyAI","article_published_time":"2023-12-04T21:46:35+00:00","article_modified_time":"2023-12-04T21:55:30+00:00","og_image":[{"width":1000,"height":563,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg","type":"image\/jpeg"}],"author":"Sam Jeans","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"Skrevet af":"Sam Jeans","Estimeret l\u00e6setid":"4 minutter"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/"},"author":{"name":"Sam Jeans","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9"},"headline":"Google DeepMind reveals a method of exposing ChatGPT’s training data","datePublished":"2023-12-04T21:46:35+00:00","dateModified":"2023-12-04T21:55:30+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/"},"wordCount":657,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg","keywords":["ChatGPT","Google DeepMind","OpenAI"],"articleSection":["Industry"],"inLanguage":"da-DK"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/","url":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/","name":"Google DeepMind afsl\u00f8rer en metode til at eksponere ChatGPT's tr\u00e6ningsdata | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg","datePublished":"2023-12-04T21:46:35+00:00","dateModified":"2023-12-04T21:55:30+00:00","description":"Forskere fra Google har angiveligt fundet en metode til at f\u00e5 adgang til nogle af de tr\u00e6ningsdata, der bruges til OpenAI's ChatGPT, if\u00f8lge en nylig unders\u00f8gelse.","breadcrumb":{"@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#breadcrumb"},"inLanguage":"da-DK","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/"]}]},{"@type":"ImageObject","inLanguage":"da-DK","@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/shutterstock_2315051741.jpg","width":1000,"height":563,"caption":"DeepMind robot"},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2023\/12\/google-deepmind-reveals-a-method-of-exposing-chatgpts-training-data\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"Google DeepMind reveals a method of exposing ChatGPT’s training data"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DailyAI","description":"Din daglige dosis af AI-nyheder","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":"da-DK"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DailyAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"da-DK","@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":"da-DK","@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 er videnskabs- og teknologiforfatter og har arbejdet i forskellige AI-startups. N\u00e5r han ikke skriver, kan han finde p\u00e5 at l\u00e6se medicinske tidsskrifter eller grave i kasser med vinylplader.","sameAs":["https:\/\/www.linkedin.com\/in\/sam-jeans-6746b9142\/"],"url":"https:\/\/dailyai.com\/da\/author\/samjeans\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/7995","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/comments?post=7995"}],"version-history":[{"count":6,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/7995\/revisions"}],"predecessor-version":[{"id":8003,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/7995\/revisions\/8003"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/media\/2080"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/media?parent=7995"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/categories?post=7995"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/tags?post=7995"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}