{"id":1704,"date":"2023-12-18T08:16:17","date_gmt":"2023-12-18T08:16:17","guid":{"rendered":"https:\/\/geneea.com\/news\/?p=1704"},"modified":"2026-01-27T21:09:51","modified_gmt":"2026-01-27T21:09:51","slug":"geneeas-ai-spotlight-7","status":"publish","type":"post","link":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7","title":{"rendered":"Geneea&#8217;s AI Spotlight #7"},"content":{"rendered":"\n<p id=\"ember229\">The seventh edition of our newsletter on Large Language Models is here.<\/p>\n\n\n\n<p id=\"ember230\">In this edition, we explore<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>the turmoil at OpenAI,<\/li>\n\n\n\n<li>several novel LLMs, including Google Gemini,<\/li>\n\n\n\n<li>a new round of challenges and fresh approaches to them,<\/li>\n\n\n\n<li>the mutual impact of AI and media, and<\/li>\n\n\n\n<li>the latest development in EU AI regulation.<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"ember232\">OpenAI Turmoil<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>November was a tough month at OpenAI.<\/li>\n\n\n\n<li>First, there was a <a href=\"https:\/\/twitter.com\/sama\/status\/1724626002595471740\">pause for sign-ups,<\/a> which was likely due to a <a href=\"https:\/\/cybernews.com\/editorial\/anonymous-sudan-explained\/\">DDoS attack<\/a> from Russia, a situation that Anthropic <a href=\"https:\/\/www.cnbc.com\/2023\/11\/08\/openais-chatgpt-is-down-in-a-major-outage.html\">also experienced<\/a>.<\/li>\n\n\n\n<li>Then, <strong>Sam Altman<\/strong> was <a href=\"https:\/\/openai.com\/blog\/openai-announces-leadership-transition\"><strong>fired<\/strong><\/a>, <strong>hired<\/strong> by Microsoft, and <strong>rehired<\/strong> by OpenAI. And the Board that fired him <strong>resigned<\/strong>. The reasons for the original firing remain somewhat unclear; some mention a <a href=\"https:\/\/www.reuters.com\/technology\/sam-altmans-ouster-openai-was-precipitated-by-letter-board-about-ai-breakthrough-2023-11-22\/\"><strong>mysterious system Q*<\/strong><\/a>. <a href=\"https:\/\/www.linkedin.com\/in\/gary-marcus-b6384b4\/\">Gary Marcus<\/a>, an LLM skeptic, <a href=\"https:\/\/garymarcus.substack.com\/p\/about-that-openai-breakthrough\">comments<\/a> on this.<strong><\/strong><\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"gemini\">Google Gemini<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>After a long wait, on December 6, Google finally <a href=\"https:\/\/deepmind.google\/technologies\/gemini\/#introduction\">announced<\/a> <strong>Gemini <\/strong>(<a href=\"https:\/\/storage.googleapis.com\/deepmind-media\/gemini\/gemini_1_report.pdf\">report<\/a>).<\/li>\n\n\n\n<li>It comes in <strong>three versions<\/strong>: <strong>Nano<\/strong> for mobile devices, <strong>Pro,<\/strong> which now powers Bard (in some locations), and <strong>Ultra<\/strong>, the largest and most capable version.<\/li>\n\n\n\n<li>The capabilities, especially in <strong>multimodal tasks<\/strong>, are <strong>impressive<\/strong>, surpassing GPT-4 on numerous benchmarks.<\/li>\n\n\n\n<li>However, soon, <strong>criticism<\/strong> <a href=\"https:\/\/thealgorithmicbridge.substack.com\/p\/why-are-people-bashing-gemini-if\">surfaced<\/a> regarding the presentation of the results, with concerns over the emphasis on the <strong>MMLU benchmark<\/strong> and its precise numbers due to identified errors (<a href=\"https:\/\/huggingface.co\/blog\/evaluating-mmlu-leaderboard\">Hugging Face<\/a>, <a href=\"https:\/\/derenrich.medium.com\/errors-in-the-mmlu-the-deep-learning-benchmark-is-wrong-surprisingly-often-7258bb045859\">Erenrich<\/a>).<\/li>\n\n\n\n<li>Additionally, there is a dispute over the <strong>comparison with GPT-4<\/strong> using <strong>different prompting<\/strong> methods, as Microsoft demonstrated <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/steering-at-the-frontier-extending-the-power-of-prompting\/\">GPT-4&#8217;s slight advantage<\/a> with similar prompts.<\/li>\n\n\n\n<li>Also, Google <a href=\"https:\/\/www.bbc.com\/news\/technology-67650807\">admitted<\/a> that an impressive <a href=\"https:\/\/youtu.be\/UIZAiXYceBI\"><strong>demo video<\/strong><\/a> showcasing multimodality was significantly edited for effect. Pity, it looked really cool.<\/li>\n\n\n\n<li>Meanwhile, we can experiment with <strong>Gemini Pro<\/strong> through its <a href=\"https:\/\/cloud.google.com\/vertex-ai\/docs\/generative-ai\/start\/quickstarts\/quickstart-multimodal\">API<\/a>. From our first impression, the model <strong>hallucinates less<\/strong> compared to PaLM: a positive improvement for Bard. It seems to <strong>cite<\/strong> information <strong>accurately<\/strong>, with proper links. However, it is <strong>very cautious<\/strong>, blocking responses for unsupported languages and random topics, like astrology forecasts for the Gemini sign. It seems <strong>proficient in image interpretation<\/strong>, reading our very colorful logo with only a minor hiccup.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"ember237\">LLM Zoo<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Mistral AI released the impressive <a href=\"https:\/\/mistral.ai\/news\/mixtral-of-experts\/\"><strong>Mixtral 8x7B<\/strong><\/a> model that outperforms larger models like Llama 2 70B and GPT3.5 across various benchmarks, with notably faster inference, thanks to the <a href=\"https:\/\/huggingface.co\/blog\/moe\"><strong>Mixture of Experts model (MoE)<\/strong><\/a> approach. It employs eight distinct <strong>groups of experts<\/strong> (neural networks) guided by a <strong>&#8216;router&#8217; network<\/strong>, which determines token allocation to the respective experts. While still demanding substantial memory, the <strong>inference mirrors that of a 12B model<\/strong>, utilizing only two experts per token. Unlike Google, they did not do a shiny fake demo but simply <a href=\"https:\/\/twitter.com\/MistralAI\/status\/1733150512395038967\">tweeted<\/a> a torrent <strong>magnet link<\/strong>. A few days later, they also made their models accessible via an <a href=\"https:\/\/mistral.ai\/news\/la-plateforme\/\"><strong>API<\/strong><\/a>; currently, it is in an early-access beta version.<\/li>\n\n\n\n<li><a href=\"https:\/\/www.anthropic.com\/index\/claude-2-1\">Claude 2.1<\/a>. Anthropic released a <strong>new version of Claude<\/strong>. It can handle 200k contexts (but as we report below, this is far from perfect) and hallucinates less. They also introduced support for customer-provided tools.<\/li>\n\n\n\n<li>Musk\u2019s xAI released <a href=\"https:\/\/x.ai\/\"><strong>Grok<\/strong><\/a>, a \u201cConversational AI for understanding the universe\u201d. Apparently, it is modeled after <strong>AI from The Hitchhiker\u2019s Guide to the Galaxy:<\/strong> should answer almost anything and also suggest the right questions. The name itself <a href=\"https:\/\/en.wikipedia.org\/wiki\/Grok\">comes from Heinlein<\/a>. If you are living in the U.S., you can test it. You even get real-time access to Twitter :D. The answers are supposed to be witty, but unfortunately, people, including <a href=\"https:\/\/twitter.com\/sama\/status\/1722766374588830101\">Sam Altman<\/a>, are making fun of Grok\u2019s <strong>dad-like jokes<\/strong>.<\/li>\n\n\n\n<li>Meta introduces <strong>SeamlessExpressive<\/strong>, a speech translation model that preserves the style, tone, speed, rhythm, expression, etc. You can try a cool <a href=\"https:\/\/seamless.metademolab.com\/expressive?utm_source=linkedin&amp;utm_medium=organic_social&amp;utm_campaign=fair10&amp;utm_content=video\">demo<\/a> or dive into technical details (<a href=\"https:\/\/github.com\/facebookresearch\/seamless_communication#seamlessexpressive\">GitHub<\/a>, <a href=\"https:\/\/ai.meta.com\/research\/publications\/seamless-multilingual-expressive-and-streaming-speech-translation\/\">paper<\/a>, <a href=\"https:\/\/huggingface.co\/facebook\/seamless-expressive\">Hugging Face<\/a>)<\/li>\n\n\n\n<li>Berkeley researchers introduced the <a href=\"https:\/\/starling.cs.berkeley.edu\/\">Starling-7B<\/a> model, trained with <strong>Reinforcement Learning from AI Feedback<\/strong> (RLAIF). Excelling on the Chatbot Arena Leaderboard without requiring laborious Human Feedback, it addresses concerns about <a href=\"https:\/\/futurism.com\/ai-companies-training-data\">training data shortages<\/a> and online <a href=\"https:\/\/www.technologyreview.com\/2022\/12\/20\/1065667\/how-ai-generated-text-is-poisoning-the-internet\/\">data contamination<\/a> by LLM models. Meanwhile, OpenAI tackles these challenges through strategic <a href=\"https:\/\/openai.com\/blog\/data-partnerships\">data partnerships<\/a>.<\/li>\n\n\n\n<li><a href=\"https:\/\/arxiv.org\/pdf\/2311.03079.pdf\">CogVLM<\/a>, a <strong>multimodal LLM<\/strong>, achieves state-of-the-art performance by employing a <strong>trainable visual expert module<\/strong> for deep fusion of vision and language tasks, steering clear of common shallow alignment techniques (image features are mapped into embedding space of the language model).<\/li>\n\n\n\n<li>Following OpenAI&#8217;s <strong>GPTs<\/strong>, Microsoft announced <a href=\"https:\/\/www.microsoft.com\/en-us\/microsoft-365\/blog\/2023\/11\/15\/announcing-microsoft-copilot-studio-customize-copilot-for-microsoft-365-and-build-your-own-standalone-copilots\/\">Copilot Studio<\/a> for<strong> tailored copilots&#8217; creation<\/strong>, while LangChain adopted <a href=\"https:\/\/synthedia.substack.com\/p\/opengpt-is-langchains-open-source?\"><strong>OpenGPTs<\/strong><\/a>.<\/li>\n\n\n\n<li>Navigating the LLM Zoo for the right model is tricky. For <strong>coding aid<\/strong>, check this <a href=\"https:\/\/arxiv.org\/pdf\/2311.07989.pdf\">comprehensive survey<\/a> &#8211; hint: go for <strong>fine-tuned models<\/strong> like Code LLaMA.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"challenges\">Challenges<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>A <a href=\"https:\/\/arxiv.org\/abs\/2311.17035\">paper<\/a> by researchers from Google Mind and several universities reports that it is possible to successfully <strong>extract training data from LLMs<\/strong>, including aligned models, such as ChatGPT.<\/li>\n\n\n\n<li><strong>Embeddings<\/strong> may <strong>not<\/strong> be <strong>as secure<\/strong> as thought, as Cornell researchers showed with <a href=\"https:\/\/arxiv.org\/pdf\/2310.06816.pdf\">Vec2Text<\/a> iteratively <strong>guessing sentences<\/strong> from embeddings. Fortunately, knowledge of the specific embedding model is required for this to work, and the method worked <strong>only with short embeddings<\/strong>.<\/li>\n\n\n\n<li>In <a href=\"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-4\/#challenges\">AI Spotlight #4<\/a>, we reported an issue regarding <strong>blind spots in context windows<\/strong>. The problem persists in new models. Anthropic&#8217;s best <a href=\"https:\/\/the-decoder.com\/anthropics-best-claude-2-1-feature-suffers-the-same-fate-as-gpt-4-turbo\/\">Claude 2.1 model<\/a> (with up to 128K tokens) suffers from the same &#8220;lost in the middle&#8221; phenomenon as GPT-4 Turbo (up to 200K tokens). Anthropic later <a href=\"https:\/\/www.anthropic.com\/index\/claude-2-1-prompting\">reported<\/a> that the problem can be minimized by adding a <strong>single sentence to the prompt<\/strong> that directs the assistant to find the relevant sentences first. This trick raises the recall from 27% to 98%!<\/li>\n\n\n\n<li>OpenAI\u2019s GPT-4 is more <strong>trustworthy but easier to trick<\/strong>. Delving into the trustworthiness of GPT models, <a href=\"https:\/\/arxiv.org\/pdf\/2306.11698.pdf\">researchers uncovered<\/a> an interesting twist. Despite GPT-4 generally being more trustworthy than GPT-3.5, it is more vulnerable to jailbreaking. This quirk arises from its improved understanding of users, making it surprisingly adept at following even misleading instructions.<\/li>\n\n\n\n<li>The <strong>word of the year<\/strong> for 2023, as selected by the <a href=\"https:\/\/dictionary.cambridge.org\/editorial\/woty\">Cambridge Dictionary<\/a>, is \u201c<strong>hallucinate<\/strong>\u201d, underscoring the significance of this challenge. <a href=\"https:\/\/vectara.com\/cut-the-bull-detecting-hallucinations-in-large-language-models\/\">Vectara<\/a> has released an open-source <strong>model for evaluating hallucinations<\/strong> in LLMs, with performance metrics of influential models available on the git <a href=\"https:\/\/github.com\/vectara\/hallucination-leaderboard\">leaderboard<\/a>.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"ember241\">Fresh Approaches to Challenges<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>We found the method of <a href=\"https:\/\/arxiv.org\/pdf\/2307.11760.pdf\">enhancing prompts with emotional stimuli<\/a> useful for enhancing results and very <strong>simple to use<\/strong>.<\/li>\n\n\n\n<li>Expanding on <a href=\"https:\/\/arxiv.org\/pdf\/2201.11903.pdf\">Chain-of-Thought<\/a> prompting, a new iteration of <a href=\"https:\/\/arxiv.org\/pdf\/2311.04254.pdf\">Everything-of-Thought<\/a> (XOT) integrates a Monte Carlo Tree Search (MCTS) module. This MCTS is trained by reinforcement learning to <strong>seek effective thought structures<\/strong>, enhancing the LLM&#8217;s reasoning and planning abilities. The LLM revises and refines these thoughts, and in case of errors, iteratively engages the MCTS for correction.<\/li>\n\n\n\n<li><a href=\"https:\/\/arxiv.org\/pdf\/2311.09210.pdf\">Chain-of-Note<\/a> technique enhances Retrieval Augmented Generation by <strong>assessing the relevance<\/strong> of the LLM&#8217;s \u201creading notes\u201d from retrieved documents. This allows the LLM to determine insufficiently relevant documents, leading to an <strong>\u201cunknown\u201d<\/strong> response <strong>instead<\/strong> of providing a potentially <strong>misleading<\/strong> <strong>answer<\/strong>.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"news&amp;media\">AI in the News &amp; Media<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>An <a href=\"https:\/\/futurism.com\/sports-illustrated-ai-generated-writers\">article<\/a> at Futurism explores how it is possible that <strong>Sports Illustrated<\/strong> published gems like \u201c<em>volleyball can be a little tricky to get into, especially without an actual ball to practice with<\/em>\u201d by clearly <strong>non-existent authors<\/strong> with AI-generated photos. The <a href=\"https:\/\/twitter.com\/sinow\/status\/1729275460922622374?s=46\">reaction<\/a> by SI generally says \u201cEverything is ok, we just did not check somebody else\u2019s work\u201d.<\/li>\n\n\n\n<li>Reporters Without Borders (RSF) and partners crafted the <a href=\"https:\/\/rsf.org\/en\/rsf-and-16-partners-unveil-paris-charter-ai-and-journalism\">Paris Charter on AI and Journalism<\/a>, outlining ten <strong>ethical principles for newsrooms<\/strong>. The principles emphasize ethics and human agency in making decisions, the media&#8217;s role in distinguishing artificial content, and the need for their engagement in governance to uphold the integrity of journalism.<\/li>\n\n\n\n<li><a href=\"https:\/\/arxiv.org\/pdf\/2311.06864.pdf\">Northwestern University researchers<\/a> investigated GPT-3&#8217;s application in <a href=\"https:\/\/generative-ai-newsroom.com\/getting-the-science-and-the-scoop-with-news-angles-from-gpt-3-b1b4fcd7f6b1\">generating different news angles<\/a> from scientific article abstracts to <strong>aid<\/strong> journalists in <strong>assessing newsworthiness<\/strong> and offer diverse <strong>story-framing angles<\/strong>. Reporters who tested it expressed excitement, finding it reliable for pinpointing key points, identifying interviewees and target audiences, and interpreting complex scientific language.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"media\">Media&#8217;s Take on AI<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>The Economist has a great <a href=\"https:\/\/www.economist.com\/podcasts\/2023\/12\/04\/bonus-interview-reid-hoffman\">interview<\/a> with Reid Hoffman, a LinkedIn and OpenAI cofounder, discussing the influence of AI on the workforce. While we do not agree with all of his viewpoints, the interview is very thought-provoking and engaging.<\/li>\n\n\n\n<li>The current <a href=\"https:\/\/www.wired.com\/story\/women-in-tech-openai-board\/?utm_source=92474-TechMIX&amp;utm_medium=email&amp;utm_term=16422815518&amp;utm_content=Wired&amp;utm_campaign=sz-techmix--20231129\">all-male OpenAI board<\/a> and The New York Times\u2019 <a href=\"https:\/\/www.nytimes.com\/2023\/12\/03\/technology\/ai-key-figures.html\">\u201cWho&#8217;s Who in AI\u201d<\/a>, <strong>exclusively featuring men<\/strong>, emphasize the perceived gender disparity in the field. In response, <a href=\"https:\/\/medium.com\/womenintechnology\/ny-times-missed-these-12-trailblazers-meet-the-women-transforming-ai-ae522f52a8b7\">this article<\/a> from <a href=\"https:\/\/www.linkedin.com\/in\/sephora-bemba\/\">S\u00e9phora Bemba<\/a> <strong>spotlights influential women<\/strong> in AI.<\/li>\n\n\n\n<li>Nvidia unveiled the <strong>H200 chip for AI<\/strong>, featuring a next-generation 141 GB memory for <strong>fast inference<\/strong>. Is it possible that their <a href=\"https:\/\/research.nvidia.com\/publication\/2023-10_chipnemo-domain-adapted-llms-chip-design\">ChipNeMo<\/a> <strong>LLM contributed<\/strong> to the <a href=\"https:\/\/spectrum.ieee.org\/ai-for-engineering\">design<\/a>?<\/li>\n\n\n\n<li>On top of Mistral, <strong>France intensified AI investments<\/strong> with the announcement of <a href=\"http:\/\/kyutai.org\/\"><strong>Kyutai<\/strong><\/a>, a <strong>non-profit research lab<\/strong>. Billionaire <a href=\"https:\/\/techcrunch.com\/2023\/11\/17\/kyutai-is-an-french-ai-research-lab-with-a-330-million-budget-that-will-make-everything-open-source\/\">Xavier Niel and partners have allocated nearly \u20ac300 million<\/a> and secured a thousand Nvidia H100 GPUs to support the researchers.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"ember248\">EU AI Regulation<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>On December 8, the EU Parliament and the Council (= member states) provisionally <a href=\"https:\/\/www.reuters.com\/technology\/stalled-eu-ai-act-talks-set-resume-2023-12-08\/\">agreed<\/a> on the <a href=\"https:\/\/www.consilium.europa.eu\/en\/press\/press-releases\/2023\/12\/09\/artificial-intelligence-act-council-and-parliament-strike-a-deal-on-the-first-worldwide-rules-for-ai\/\"><strong>EU AI Act<\/strong><\/a> (AIA). Some technical details still remain to be finalized and the act has to be formally approved by both the Parliament and the states, so officially, it won\u2019t go into effect before 2025.<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li>With minor changes, the act preserved the <strong>risk-based approach<\/strong> we described in <a href=\"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-3\/#aia\">AI Spotlight #3<\/a>.<\/li>\n\n\n\n<li>In the case of <strong>LLMs<\/strong>, the main focus is on <strong>transparency and testing<\/strong> before release.<\/li>\n\n\n\n<li>There is <strong>criticism<\/strong> from both sides. <a href=\"https:\/\/www.ft.com\/content\/9339d104-7b0c-42b8-9316-72226dd4e4c0\">Emanuel Macron<\/a> and <a href=\"https:\/\/www.digitaleurope.org\/news\/23173\/\">DigitalEurope<\/a>, a business group, say that AIA is too strict and will hamper innovation (note that Mistral is a French company). Others think that the regulation is too weak and full of loopholes. Amnesty International <a href=\"https:\/\/www.amnesty.org\/en\/latest\/news\/2023\/12\/eu-blocs-decision-to-not-ban-public-mass-surveillance-in-ai-act-sets-a-devastating-global-precedent\/\">complains<\/a> that it \u201cgreenlighted dystopian digital surveillance\u201d because it did not ban facial recognition.<\/li>\n<\/ul>\n\n\n\n<p>Please <a href=\"https:\/\/www.linkedin.com\/pulse\/geneeas-ai-spotlight-7-geneea-wrwnf%3FtrackingId=WkSJSecvR9KSojNO46v%252FAA%253D%253D\/?trackingId=WkSJSecvR9KSojNO46v%2FAA%3D%3D\">subscribe<\/a> and stay tuned for the next issue of Geneea\u2019s AI Spotlight newsletter!<\/p>\n","protected":false},"excerpt":{"rendered":"<p>The seventh edition of our newsletter on Large Language Models is here.<br \/>\nIn this edition, we explore<br \/>\n\u2022  the turmoil at OpenAI,<br \/>\n\u2022  several novel LLMs, including Google Gemini,<br \/>\n\u2022  a new round of challenges and fresh approaches to them,<br \/>\n\u2022  the mutual impact of AI and media, and<br \/>\n\u2022  the latest development in EU AI regulation.<\/p>\n","protected":false},"author":15,"featured_media":1705,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[378,374],"tags":[244,240,242],"class_list":["post-1704","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-large-language-models","category-newsletter","tag-ai","tag-generativeai","tag-newsletter"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.3 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Geneea&#039;s AI Spotlight #7 - Geneea News<\/title>\n<meta name=\"description\" content=\"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Geneea&#039;s AI Spotlight #7 - Geneea News\" \/>\n<meta property=\"og:description\" content=\"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7\" \/>\n<meta property=\"og:site_name\" content=\"Geneea News\" \/>\n<meta property=\"article:published_time\" content=\"2023-12-18T08:16:17+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-01-27T21:09:51+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot-1024x578.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"578\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Marcela Soukupova\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Marcela Soukupova\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"7 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7\"},\"author\":{\"name\":\"Marcela Soukupova\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#\\\/schema\\\/person\\\/69c8751a4c026723f4bac2e892f52cd8\"},\"headline\":\"Geneea&#8217;s AI Spotlight #7\",\"datePublished\":\"2023-12-18T08:16:17+00:00\",\"dateModified\":\"2026-01-27T21:09:51+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7\"},\"wordCount\":1549,\"publisher\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/newsletter_7_robot.png\",\"keywords\":[\"AI\",\"generativeAI\",\"newsletter\"],\"articleSection\":[\"Large language models\",\"Newsletter\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7\",\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7\",\"name\":\"Geneea's AI Spotlight #7 - Geneea News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/newsletter_7_robot.png\",\"datePublished\":\"2023-12-18T08:16:17+00:00\",\"dateModified\":\"2026-01-27T21:09:51+00:00\",\"description\":\"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#primaryimage\",\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/newsletter_7_robot.png\",\"contentUrl\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2023\\\/12\\\/newsletter_7_robot.png\",\"width\":1921,\"height\":1085},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/geneeas-ai-spotlight-7#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/geneea.com\\\/news\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Geneea&#8217;s AI Spotlight #7\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#website\",\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/\",\"name\":\"Geneea News\",\"description\":\"Learn more about what&#039;s happening at Geneea: new NLP features, newest case studies, tutoring projects, conferences we attended, etc.\",\"publisher\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/geneea.com\\\/news\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#organization\",\"name\":\"Geneea News\",\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/cropped-geneea-logo-50pc.png\",\"contentUrl\":\"https:\\\/\\\/geneea.com\\\/news\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/cropped-geneea-logo-50pc.png\",\"width\":242,\"height\":64,\"caption\":\"Geneea News\"},\"image\":{\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/geneea.com\\\/news\\\/#\\\/schema\\\/person\\\/69c8751a4c026723f4bac2e892f52cd8\",\"name\":\"Marcela Soukupova\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g\",\"caption\":\"Marcela Soukupova\"},\"sameAs\":[\"http:\\\/\\\/Marcela%20Soukupova\"],\"url\":\"https:\\\/\\\/geneea.com\\\/news\\\/author\\\/marcela-soukupova\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Geneea's AI Spotlight #7 - Geneea News","description":"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.","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:\/\/geneea.com\/news\/geneeas-ai-spotlight-7","og_locale":"en_US","og_type":"article","og_title":"Geneea's AI Spotlight #7 - Geneea News","og_description":"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.","og_url":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7","og_site_name":"Geneea News","article_published_time":"2023-12-18T08:16:17+00:00","article_modified_time":"2026-01-27T21:09:51+00:00","og_image":[{"width":1024,"height":578,"url":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot-1024x578.png","type":"image\/png"}],"author":"Marcela Soukupova","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Marcela Soukupova","Est. reading time":"7 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#article","isPartOf":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7"},"author":{"name":"Marcela Soukupova","@id":"https:\/\/geneea.com\/news\/#\/schema\/person\/69c8751a4c026723f4bac2e892f52cd8"},"headline":"Geneea&#8217;s AI Spotlight #7","datePublished":"2023-12-18T08:16:17+00:00","dateModified":"2026-01-27T21:09:51+00:00","mainEntityOfPage":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7"},"wordCount":1549,"publisher":{"@id":"https:\/\/geneea.com\/news\/#organization"},"image":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#primaryimage"},"thumbnailUrl":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot.png","keywords":["AI","generativeAI","newsletter"],"articleSection":["Large language models","Newsletter"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7","url":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7","name":"Geneea's AI Spotlight #7 - Geneea News","isPartOf":{"@id":"https:\/\/geneea.com\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#primaryimage"},"image":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#primaryimage"},"thumbnailUrl":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot.png","datePublished":"2023-12-18T08:16:17+00:00","dateModified":"2026-01-27T21:09:51+00:00","description":"LLM Newsletter #7: turmoil at OpenAI, new large language models including Google Gemini, challenges and solutions, AI and media, news in EU AI regulation.","breadcrumb":{"@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#primaryimage","url":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot.png","contentUrl":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2023\/12\/newsletter_7_robot.png","width":1921,"height":1085},{"@type":"BreadcrumbList","@id":"https:\/\/geneea.com\/news\/geneeas-ai-spotlight-7#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/geneea.com\/news"},{"@type":"ListItem","position":2,"name":"Geneea&#8217;s AI Spotlight #7"}]},{"@type":"WebSite","@id":"https:\/\/geneea.com\/news\/#website","url":"https:\/\/geneea.com\/news\/","name":"Geneea News","description":"Learn more about what&#039;s happening at Geneea: new NLP features, newest case studies, tutoring projects, conferences we attended, etc.","publisher":{"@id":"https:\/\/geneea.com\/news\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/geneea.com\/news\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/geneea.com\/news\/#organization","name":"Geneea News","url":"https:\/\/geneea.com\/news\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/geneea.com\/news\/#\/schema\/logo\/image\/","url":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2022\/02\/cropped-geneea-logo-50pc.png","contentUrl":"https:\/\/geneea.com\/news\/wp-content\/uploads\/2022\/02\/cropped-geneea-logo-50pc.png","width":242,"height":64,"caption":"Geneea News"},"image":{"@id":"https:\/\/geneea.com\/news\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/geneea.com\/news\/#\/schema\/person\/69c8751a4c026723f4bac2e892f52cd8","name":"Marcela Soukupova","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/44f35824640c6a5b31bfef2f478d704874dc3d81bfad511c158ab12274072e16?s=96&d=mm&r=g","caption":"Marcela Soukupova"},"sameAs":["http:\/\/Marcela%20Soukupova"],"url":"https:\/\/geneea.com\/news\/author\/marcela-soukupova"}]}},"_links":{"self":[{"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/posts\/1704","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/users\/15"}],"replies":[{"embeddable":true,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/comments?post=1704"}],"version-history":[{"count":6,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/posts\/1704\/revisions"}],"predecessor-version":[{"id":1834,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/posts\/1704\/revisions\/1834"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/media\/1705"}],"wp:attachment":[{"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/media?parent=1704"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/categories?post=1704"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/geneea.com\/news\/wp-json\/wp\/v2\/tags?post=1704"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}