{"id":8353,"date":"2023-12-16T04:46:40","date_gmt":"2023-12-16T04:46:40","guid":{"rendered":"https:\/\/dailyai.com\/?p=8353"},"modified":"2023-12-16T04:46:40","modified_gmt":"2023-12-16T04:46:40","slug":"microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt","status":"publish","type":"post","link":"https:\/\/dailyai.com\/pt\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/","title":{"rendered":"A Microsoft protege a Gemini com o GPT-4 impulsionado pelo Medprompt"},"content":{"rendered":"<p><strong>No in\u00edcio deste m\u00eas, a Google anunciou orgulhosamente que o seu modelo Gemini mais potente venceu o GPT-4 nos testes de refer\u00eancia MMLU (Massive Multitask Language Understanding). Com a nova t\u00e9cnica de solicita\u00e7\u00e3o da Microsoft, o GPT-4 recuperou o primeiro lugar, embora por uma fra\u00e7\u00e3o de um por cento.<\/strong><\/p>\n<p>Para al\u00e9m do drama em torno do seu v\u00eddeo de marketing, o Gemini da Google \u00e9 um grande neg\u00f3cio para a empresa e os seus resultados de refer\u00eancia MMLU s\u00e3o impressionantes. Mas a Microsoft, o maior investidor da OpenAI, n\u00e3o esperou muito tempo para criticar os esfor\u00e7os da Google.<\/p>\n<p>A manchete \u00e9 que a Microsoft conseguiu que o GPT-4 superasse os resultados do MMLU do Gemini Ultra. A realidade \u00e9 que superou o resultado do Gemini de 90,04% por apenas 0,06%.<\/p>\n<p>A hist\u00f3ria dos bastidores que tornaram isto poss\u00edvel \u00e9 mais empolgante do que a competi\u00e7\u00e3o incremental que vemos nestas tabelas de classifica\u00e7\u00e3o. As novas t\u00e9cnicas de est\u00edmulo da Microsoft podem melhorar o desempenho de modelos de IA mais antigos.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p dir=\"ltr\" lang=\"en\">Lembra-se de como o Gemini Ultra, ainda n\u00e3o lan\u00e7ado, da Google acabou de bater o GPT-4 para se tornar a IA de topo?<\/p>\n<p>Bem, a Microsoft acabou de demonstrar que, com a devida solicita\u00e7\u00e3o, o GPT-4 realmente vence o Gemini nos benchmarks.<\/p>\n<p>H\u00e1 muito espa\u00e7o para ganhos, mesmo com modelos mais antigos. <a href=\"https:\/\/t.co\/YQ5zJI6Gad\">https:\/\/t.co\/YQ5zJI6Gad<\/a> <a href=\"https:\/\/t.co\/X3HFmXa30X\">pic.twitter.com\/X3HFmXa30X<\/a><\/p>\n<p>- Ethan Mollick (@emollick) <a href=\"https:\/\/twitter.com\/emollick\/status\/1734637276330651953?ref_src=twsrc%5Etfw\">12 de dezembro de 2023<\/a><\/p><\/blockquote>\n<p><script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<h2>Medprompt<\/h2>\n<p>Quando se ouve falar em \"orientar\" um modelo, quer-se apenas dizer que, com uma orienta\u00e7\u00e3o cuidadosa, \u00e9 poss\u00edvel guiar um modelo para obter um resultado que esteja melhor alinhado com o que se pretendia.<\/p>\n<p>A Microsoft desenvolveu uma combina\u00e7\u00e3o de t\u00e9cnicas de solicita\u00e7\u00e3o que provaram ser realmente boas neste dom\u00ednio. <a href=\"https:\/\/arxiv.org\/pdf\/2311.16452.pdf\" target=\"_blank\" rel=\"noopener\">Medprompt<\/a> come\u00e7ou como um projeto para fazer com que o GPT-4 desse melhores respostas em testes de refer\u00eancia de desafios m\u00e9dicos, como o conjunto de testes MultiMedQA.<\/p>\n<figure id=\"attachment_8354\" aria-describedby=\"caption-attachment-8354\" style=\"width: 1722px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"size-full wp-image-8354\" src=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance.png\" alt=\"\" width=\"1722\" height=\"906\" srcset=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance.png 1722w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-300x158.png 300w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-1024x539.png 1024w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-768x404.png 768w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-1536x808.png 1536w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-370x195.png 370w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-800x421.png 800w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-20x11.png 20w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-740x389.png 740w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-1600x842.png 1600w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-1320x694.png 1320w, https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Medprompt-improvements-on-MedQA-performance-91x48.png 91w\" sizes=\"auto, (max-width: 1722px) 100vw, 1722px\" \/><figcaption id=\"caption-attachment-8354\" class=\"wp-caption-text\">Melhorias do Medprompt no desempenho dos testes MedQA. Microsoft<\/figcaption><\/figure>\n<p>Os investigadores da Microsoft pensaram que, se o Medprompt funcionasse bem em testes m\u00e9dicos especializados, tamb\u00e9m poderia melhorar o desempenho generalista do GPT-4. E, assim, a Microsoft e a OpenAI recuperaram o direito de se gabarem com o GPT-4 em rela\u00e7\u00e3o ao Gemini Ultra.<\/p>\n<h2>Como \u00e9 que o Medprompt funciona?<\/h2>\n<p>O Medprompt \u00e9 uma combina\u00e7\u00e3o de t\u00e9cnicas inteligentes de solicita\u00e7\u00e3o, todas reunidas numa s\u00f3. Baseia-se em tr\u00eas t\u00e9cnicas principais.<\/p>\n<h4>Aprendizagem din\u00e2mica de poucos disparos (DFSL)<\/h4>\n<p>A \"aprendizagem com poucas tentativas\" refere-se ao facto de dar ao GPT-4 alguns exemplos antes de lhe pedir para resolver um problema semelhante. Quando se v\u00ea uma refer\u00eancia como \"5-shot\", significa que foram dados 5 exemplos ao modelo. \"Zero-shot\" significa que teve de responder sem quaisquer exemplos.<\/p>\n<p>O documento do Medprompt explica que \"por uma quest\u00e3o de simplicidade e efici\u00eancia, os exemplos de poucos disparos aplicados na solicita\u00e7\u00e3o de uma determinada tarefa s\u00e3o normalmente fixos; mant\u00eam-se inalterados em todos os exemplos de teste\".<\/p>\n<p>O resultado \u00e9 que os exemplos que s\u00e3o apresentados aos modelos s\u00e3o muitas vezes apenas relevantes ou representativos em termos gerais.<\/p>\n<p>Se o seu conjunto de treino for suficientemente grande, pode fazer com que o modelo analise todos os exemplos e escolha aqueles que s\u00e3o semanticamente semelhantes ao problema que tem de resolver. O resultado \u00e9 que os exemplos de aprendizagem de poucos disparos est\u00e3o mais especificamente alinhados com um determinado problema.<\/p>\n<h4>Cadeia de pensamento auto-gerada (CoT)<\/h4>\n<p>A sugest\u00e3o da Cadeia de Pensamento (CoT) \u00e9 uma \u00f3ptima forma de orientar um LLM. Quando se diz \"pense bem\" ou \"resolva o problema passo a passo\", os resultados s\u00e3o muito melhores.<\/p>\n<p>\u00c9 poss\u00edvel ser muito mais espec\u00edfico na forma como se orienta a cadeia de pensamento que o modelo deve seguir, mas isso implica uma engenharia manual r\u00e1pida.<\/p>\n<p>Os investigadores descobriram que \"podiam simplesmente pedir ao GPT-4 para gerar cadeias de pensamento para os exemplos de treino\". A abordagem deles basicamente diz ao GPT-4: 'Aqui est\u00e1 uma pergunta, as op\u00e7\u00f5es de resposta e a resposta correcta. Que CoT dever\u00edamos incluir num prompt que chegasse a esta resposta?<\/p>\n<h4>Sele\u00e7\u00e3o de conjuntos aleat\u00f3rios<\/h4>\n<p>A maioria dos testes de refer\u00eancia do MMLU s\u00e3o perguntas de escolha m\u00faltipla. Quando um modelo de IA responde a estas perguntas, pode ser v\u00edtima de uma tend\u00eancia posicional. Por outras palavras, pode favorecer a op\u00e7\u00e3o B ao longo do tempo, apesar de nem sempre ser a resposta correcta.<\/p>\n<p>O Agrupamento Baralhado de Op\u00e7\u00f5es baralha as posi\u00e7\u00f5es das op\u00e7\u00f5es de resposta e faz com que o GPT-4 responda novamente \u00e0 pergunta. Faz isto v\u00e1rias vezes e depois a resposta mais consistentemente escolhida \u00e9 selecionada como a resposta final.<\/p>\n<p>A combina\u00e7\u00e3o destas tr\u00eas t\u00e9cnicas de prompt foi o que deu \u00e0 Microsoft a oportunidade de lan\u00e7ar um pouco de sombra sobre os resultados do Gemini. Ser\u00e1 interessante ver quais os resultados que o Gemini Ultra alcan\u00e7aria se utilizasse uma abordagem semelhante.<\/p>\n<p>O Medprompt \u00e9 empolgante porque mostra que os modelos mais antigos podem ter um desempenho ainda melhor do que pens\u00e1vamos, se os avisarmos de formas inteligentes. No entanto, o poder de processamento adicional necess\u00e1rio para estes passos extra pode n\u00e3o tornar esta abordagem vi\u00e1vel na maioria dos cen\u00e1rios.<\/p>","protected":false},"excerpt":{"rendered":"<p>No in\u00edcio deste m\u00eas, a Google anunciou orgulhosamente que o seu modelo Gemini mais potente venceu o GPT-4 nos testes de refer\u00eancia MMLU (Massive Multitask Language Understanding). Com a nova t\u00e9cnica de solicita\u00e7\u00e3o da Microsoft, o GPT-4 recuperou o primeiro lugar, embora por uma fra\u00e7\u00e3o de um por cento. Para al\u00e9m do drama em torno do seu v\u00eddeo de marketing, o Gemini da Google \u00e9 um grande neg\u00f3cio para a empresa e os seus resultados de refer\u00eancia MMLU s\u00e3o impressionantes. Mas a Microsoft, o maior investidor da OpenAI, n\u00e3o esperou muito tempo para lan\u00e7ar sombra sobre os esfor\u00e7os da Google. A manchete \u00e9 que a Microsoft conseguiu que o GPT-4 superasse os resultados do MMLU do Gemini Ultra. A realidade \u00e9 que superou a pontua\u00e7\u00e3o do Gemini de 90,04%<\/p>","protected":false},"author":6,"featured_media":8355,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[99,102,118,121],"class_list":["post-8353","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-ai-race","tag-google","tag-llms","tag-microsoft"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.4 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Microsoft shades Gemini with GPT-4 boosted by Medprompt | 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\/pt\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/\" \/>\n<meta property=\"og:locale\" content=\"pt_PT\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Microsoft shades Gemini with GPT-4 boosted by Medprompt | DailyAI\" \/>\n<meta property=\"og:description\" content=\"Earlier this month Google proudly announced that its most powerful Gemini model beat GPT-4 on the Massive Multitask Language Understanding MMLU benchmark tests. Microsoft\u2019s new prompting technique sees GPT-4 regain the top spot, albeit by a fraction of a percent. Besides the drama surrounding its marketing video, Google\u2019s Gemini is a big deal for the company and its MMLU benchmark results are impressive. But Microsoft, OpenAI\u2019s biggest investor, didn\u2019t wait long to throw shade on Google\u2019s efforts. The headline is that Microsoft got GPT-4 to beat Gemini Ultra\u2019s MMLU results. The reality is that it beat Gemini\u2019s score of 90.04%\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/pt\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2023-12-16T04:46:40+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1000\" \/>\n\t<meta property=\"og:image:height\" content=\"665\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Eugene van der Watt\" \/>\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=\"Escrito por\" \/>\n\t<meta name=\"twitter:data1\" content=\"Eugene van der Watt\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tempo estimado de leitura\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutos\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/\"},\"author\":{\"name\":\"Eugene van der Watt\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/7ce525c6d0c79838b7cc7cde96993cfa\"},\"headline\":\"Microsoft shades Gemini with GPT-4 boosted by Medprompt\",\"datePublished\":\"2023-12-16T04:46:40+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/\"},\"wordCount\":756,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/Microsoft-vs-Google.jpg\",\"keywords\":[\"AI race\",\"Google\",\"LLMS\",\"Microsoft\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"pt-PT\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/\",\"name\":\"Microsoft shades Gemini with GPT-4 boosted by Medprompt | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/Microsoft-vs-Google.jpg\",\"datePublished\":\"2023-12-16T04:46:40+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#breadcrumb\"},\"inLanguage\":\"pt-PT\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/Microsoft-vs-Google.jpg\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/Microsoft-vs-Google.jpg\",\"width\":1000,\"height\":665},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/12\\\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Microsoft shades Gemini with GPT-4 boosted by Medprompt\"}]},{\"@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\":\"pt-PT\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@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\\\/7ce525c6d0c79838b7cc7cde96993cfa\",\"name\":\"Eugene van der Watt\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"caption\":\"Eugene van der Watt\"},\"description\":\"Eugene comes from an electronic engineering background and loves all things tech. When he takes a break from consuming AI news you'll find him at the snooker table.\",\"sameAs\":[\"www.linkedin.com\\\/in\\\/eugene-van-der-watt-16828119\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/pt\\\/author\\\/eugene\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"A Microsoft protege a Gemini com o GPT-4 impulsionado pelo Medprompt | 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\/pt\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/","og_locale":"pt_PT","og_type":"article","og_title":"Microsoft shades Gemini with GPT-4 boosted by Medprompt | DailyAI","og_description":"Earlier this month Google proudly announced that its most powerful Gemini model beat GPT-4 on the Massive Multitask Language Understanding MMLU benchmark tests. Microsoft\u2019s new prompting technique sees GPT-4 regain the top spot, albeit by a fraction of a percent. Besides the drama surrounding its marketing video, Google\u2019s Gemini is a big deal for the company and its MMLU benchmark results are impressive. But Microsoft, OpenAI\u2019s biggest investor, didn\u2019t wait long to throw shade on Google\u2019s efforts. The headline is that Microsoft got GPT-4 to beat Gemini Ultra\u2019s MMLU results. The reality is that it beat Gemini\u2019s score of 90.04%","og_url":"https:\/\/dailyai.com\/pt\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/","og_site_name":"DailyAI","article_published_time":"2023-12-16T04:46:40+00:00","og_image":[{"width":1000,"height":665,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg","type":"image\/jpeg"}],"author":"Eugene van der Watt","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"Escrito por":"Eugene van der Watt","Tempo estimado de leitura":"4 minutos"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/"},"author":{"name":"Eugene van der Watt","@id":"https:\/\/dailyai.com\/#\/schema\/person\/7ce525c6d0c79838b7cc7cde96993cfa"},"headline":"Microsoft shades Gemini with GPT-4 boosted by Medprompt","datePublished":"2023-12-16T04:46:40+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/"},"wordCount":756,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg","keywords":["AI race","Google","LLMS","Microsoft"],"articleSection":["Industry"],"inLanguage":"pt-PT"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/","url":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/","name":"A Microsoft protege a Gemini com o GPT-4 impulsionado pelo Medprompt | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg","datePublished":"2023-12-16T04:46:40+00:00","breadcrumb":{"@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#breadcrumb"},"inLanguage":"pt-PT","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/"]}]},{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/12\/Microsoft-vs-Google.jpg","width":1000,"height":665},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2023\/12\/microsoft-shades-gemini-with-gpt-4-boosted-by-medprompt\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"Microsoft shades Gemini with GPT-4 boosted by Medprompt"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DailyAI","description":"A sua dose di\u00e1ria de not\u00edcias sobre IA","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":"pt-PT"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DailyAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"pt-PT","@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\/7ce525c6d0c79838b7cc7cde96993cfa","name":"Eugene van der Watt","image":{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","caption":"Eugene van der Watt"},"description":"Eugene vem de uma forma\u00e7\u00e3o em engenharia eletr\u00f3nica e adora tudo o que \u00e9 tecnologia. Quando faz uma pausa no consumo de not\u00edcias sobre IA, pode encontr\u00e1-lo \u00e0 mesa de snooker.","sameAs":["www.linkedin.com\/in\/eugene-van-der-watt-16828119"],"url":"https:\/\/dailyai.com\/pt\/author\/eugene\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/8353","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/comments?post=8353"}],"version-history":[{"count":4,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/8353\/revisions"}],"predecessor-version":[{"id":8359,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/posts\/8353\/revisions\/8359"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/media\/8355"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/media?parent=8353"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/categories?post=8353"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/pt\/wp-json\/wp\/v2\/tags?post=8353"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}