{"id":92812,"date":"2024-09-03T16:26:51","date_gmt":"2024-09-03T20:26:51","guid":{"rendered":"https:\/\/danielschristian.com\/learning-ecosystems\/?p=92812"},"modified":"2024-09-03T16:45:56","modified_gmt":"2024-09-03T20:45:56","slug":"when-a-i-s-output-is-a-threat-to-a-i-itself-bhatia-xai-introduces-most-powerful-ai-training-system-in-the-world-other-items-re-ai","status":"publish","type":"post","link":"http:\/\/danielschristian.com\/learning-ecosystems\/2024\/09\/03\/when-a-i-s-output-is-a-threat-to-a-i-itself-bhatia-xai-introduces-most-powerful-ai-training-system-in-the-world-other-items-re-ai\/","title":{"rendered":"&#8220;When A.I.\u2019s Output Is a Threat to A.I. Itself&#8221; [Bhatia], xAI introduces most powerful AI training system in the world +other items re: AI"},"content":{"rendered":"<p><a href=\"https:\/\/www.nytimes.com\/interactive\/2024\/08\/26\/upshot\/ai-synthetic-data.html\" target=\"_blank\" rel=\"noopener\"><strong>When A.I.\u2019s Output Is a Threat to A.I. Itself <\/strong><\/a>&#8212; from nytimes.com by Aatish Bhatia<br \/>\n<em>As A.I.-generated data becomes harder to detect, it\u2019s increasingly likely to be ingested by future A.I., leading to worse results.<\/em><\/p>\n<p style=\"padding-left: 40px;\">All this A.I.-generated information can make it harder for us to know what\u2019s real. And it also poses a problem for A.I. companies. As they trawl the web for new data to train their next models on \u2014 an\u00a0increasingly challenging task\u00a0\u2014 they\u2019re likely to ingest some of their own A.I.-generated content, creating an unintentional feedback loop in which what was once the output from one A.I. becomes the input for another.<\/p>\n<p style=\"padding-left: 40px;\">In the long run, this cycle may pose a threat to A.I. itself.\u00a0Research\u00a0has\u00a0shown\u00a0that\u00a0when generative A.I. is trained on a lot of its own output, it can get a lot worse.<\/p>\n<hr \/>\n<blockquote class=\"twitter-tweet\">\n<p dir=\"ltr\" lang=\"en\">This weekend, the <a href=\"https:\/\/twitter.com\/xai?ref_src=twsrc%5Etfw\">@xAI<\/a> team brought our Colossus 100k H100 training cluster online. From start to finish, it was done in 122 days.<\/p>\n<p>Colossus is the most powerful AI training system in the world. Moreover, it will double in size to 200k (50k H200s) in a few months.<\/p>\n<p>Excellent\u2026<\/p>\n<p>\u2014 Elon Musk (@elonmusk) <a href=\"https:\/\/twitter.com\/elonmusk\/status\/1830650370336473253?ref_src=twsrc%5Etfw\">September 2, 2024<\/a><\/p><\/blockquote>\n<p><script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p><em><strong>Per The Rundown AI:<\/strong><\/em><\/p>\n<p><strong>The Rundown:<\/strong> Elon Musk\u2019s xAI just launched \u201cColossus\u201c, the world\u2019s most powerful AI cluster powered by a whopping 100,000 Nvidia H100 GPUs, which was built in just 122 days and is planned to double in size soon.<br \/>\n&#8230;<br \/>\n<strong>Why it matters:<\/strong> xAI\u2019s Grok 2 recently caught up to OpenAI\u2019s GPT-4 in record time, and was trained on only around 15,000 GPUs. With now more than six times that amount in production, the xAI team and future versions of Grok are going to put a significant amount of pressure on OpenAI, Google, and others to deliver.<\/p>\n<hr \/>\n<p><a href=\"https:\/\/www.theverge.com\/2024\/8\/27\/24229843\/google-meets-automatic-ai-note-taking-launch\" target=\"_blank\" rel=\"noopener\"><strong>Google Meet\u2019s automatic AI note-taking is here<\/strong><\/a> &#8212; from theverge.com by Joanna Nelius<br \/>\n<em>Starting [on 8\/28\/24], some Google Workspace customers can have Google Meet be their personal note-taker.<\/em><\/p>\n<p style=\"padding-left: 40px;\">Google Meet\u2019s newest AI-powered feature, \u201ctake notes for me,\u201d\u00a0has\u00a0started rolling out today\u00a0to Google Workspace customers with the Gemini Enterprise, Gemini Education Premium, or AI Meetings &amp; Messaging add-ons. It\u2019s similar to Meet\u2019s transcription tool, only instead of automatically transcribing what everyone says, it summarizes what everyone talked about. Google\u00a0first announced this feature\u00a0at its 2023 Cloud Next conference.<\/p>\n<hr \/>\n<p><a href=\"https:\/\/www.bloomberg.com\/news\/features\/2024-08-27\/philippines-call-centers-navigate-ai-impact-on-jobs\" target=\"_blank\" rel=\"noopener\"><strong>The World\u2019s Call Center Capital Is Gripped by AI Fever \u2014 and Fear<\/strong><\/a>\u00a0&#8212; from bloomberg.com by Saritha Rai <em><span style=\"color: #800000;\">[behind a paywall]<\/span><\/em><br \/>\n<em>The experiences\u00a0of staff in the Philippines\u2019 outsourcing industry are a preview of the challenges and choices coming soon to white-collar workers around the globe.<\/em><\/p>\n<hr \/>\n<p><a href=\"https:\/\/www.anthropic.com\/news\/artifacts\" target=\"_blank\" rel=\"noopener\"><strong>[Claude] Artifacts are now generally available<\/strong><\/a> &#8212; from anthropic.com<\/p>\n<p style=\"padding-left: 40px;\">[On 8\/27\/24], we\u2019re making Artifacts available for all Claude.ai users across our Free, Pro, and Team plans. And now, you can create and view Artifacts on our iOS and Android apps.<\/p>\n<p style=\"padding-left: 40px;\">Artifacts turn conversations with Claude into a more creative and collaborative experience. With Artifacts, you have a dedicated window to instantly see, iterate, and build on the work you create with Claude. Since launching as a\u00a0feature preview\u00a0in June, users have created tens of millions of Artifacts.<\/p>\n<hr \/>\n<p><a href=\"https:\/\/airisk.mit.edu\/\" target=\"_blank\" rel=\"noopener\"><img decoding=\"async\" class=\"alignnone wp-image-92818 size-full\" src=\"http:\/\/danielschristian.com\/learning-ecosystems\/wp-content\/uploads\/2024\/09\/AIRiskRepository-Sept2024.jpg\" alt=\"MIT's AI Risk Repository -- a comprehensive database of risks from AI systems\" width=\"100%\" height=\"100%\" srcset=\"http:\/\/danielschristian.com\/learning-ecosystems\/wp-content\/uploads\/2024\/09\/AIRiskRepository-Sept2024.jpg 805w, http:\/\/danielschristian.com\/learning-ecosystems\/wp-content\/uploads\/2024\/09\/AIRiskRepository-Sept2024-150x74.jpg 150w, http:\/\/danielschristian.com\/learning-ecosystems\/wp-content\/uploads\/2024\/09\/AIRiskRepository-Sept2024-768x380.jpg 768w\" sizes=\"(max-width: 805px) 100vw, 805px\" \/><\/a><\/p>\n<p><a href=\"https:\/\/airisk.mit.edu\/\" target=\"_blank\" rel=\"noopener\"><strong>What are the risks from Artificial Intelligence?<\/strong><\/a><br \/>\n<em>A comprehensive living database of over 700 AI risks categorized by their cause and risk domain.<\/em><\/p>\n<p><strong>What is the AI Risk Repository?<br \/>\n<\/strong>The AI Risk Repository has three parts:<\/p>\n<div class=\"rich-text-block-2 w-richtext\">\n<ul class=\"list-2\" role=\"list\">\n<li class=\"list-item\">The\u00a0<strong>AI Risk Database<\/strong>\u00a0captures 700+ risks extracted from 43 existing frameworks, with quotes and page numbers.<\/li>\n<li class=\"list-item\">The\u00a0<strong>Causal Taxonomy of AI Risks<\/strong>\u00a0classifies how, when, and why these risks occur.<\/li>\n<li class=\"list-item\">The\u00a0<strong>Domain Taxonomy of AI<\/strong>\u00a0<strong>Risks<\/strong>\u00a0classifies these risks into seven domains (e.g., \u201cMisinformation\u201d) and 23 subdomains (e.g., \u201cFalse or misleading information\u201d).<\/li>\n<\/ul>\n<hr \/>\n<p><a href=\"https:\/\/www.newsday.com\/news\/nation\/california-ai-election-deepfakes-safety-regulations-f65845\" target=\"_blank\" rel=\"noopener\"><strong>California lawmakers approve legislation to ban deepfakes, protect workers and regulate AI<\/strong> <\/a>&#8212; from newsday.com by The Associated Press<\/p>\n<p style=\"padding-left: 40px;\">SACRAMENTO, Calif. \u2014 California lawmakers approved a host of proposals this week aiming to regulate the artificial intelligence industry, combat deepfakes and protect workers from exploitation by the rapidly evolving technology.<\/p>\n<p><em><strong>Per Oncely:<\/strong><\/em><\/p>\n<p><strong>The Details:<\/strong><\/p>\n<ul>\n<li><strong>Combatting Deepfakes:\u00a0<\/strong>New laws to restrict election-related deepfakes and deepfake pornography, especially of minors, requiring social media to remove such content promptly.<\/li>\n<li><strong>Setting Safety Guardrails:<\/strong> California is poised to set comprehensive safety standards for AI, including transparency in AI model training and pre-emptive safety protocols.<\/li>\n<li><strong>Protecting Workers:<\/strong>\u00a0Legislation to prevent the replacement of workers, like voice actors and call center employees, with AI technologies.<\/li>\n<\/ul>\n<hr \/>\n<\/div>\n<blockquote class=\"twitter-tweet\">\n<p dir=\"ltr\" lang=\"en\">Over the coming days, start creating and chatting with Gems: customizable versions of Gemini that act as topic experts. ?<\/p>\n<p>We\u2019re also launching premade Gems for different scenarios &#8211; including Learning coach to break down complex topics and Coding partner to level up your skills\u2026 <a href=\"https:\/\/t.co\/2Dk8NxtTCE\">pic.twitter.com\/2Dk8NxtTCE<\/a><\/p>\n<p>\u2014 Google DeepMind (@GoogleDeepMind) <a href=\"https:\/\/twitter.com\/GoogleDeepMind\/status\/1828855383131074997?ref_src=twsrc%5Etfw\">August 28, 2024<\/a><\/p><\/blockquote>\n<p><script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p><a href=\"https:\/\/blog.google\/products\/gemini\/google-gemini-update-august-2024\/\" target=\"_blank\" rel=\"noopener\"><strong>New in Gemini: Custom Gems and improved image generation with Imagen 3<\/strong><\/a>\u00a0&#8212; from blog.google<br \/>\n<em>The ability to create custom Gems is coming to Gemini Advanced subscribers, and updated image generation capabilities with our latest Imagen 3 model are coming to everyone.<\/em><\/p>\n<p style=\"padding-left: 40px;\">We have new features rolling out, [that started on 8\/28\/24], that we previewed at Google I\/O. Gems, a new feature that lets you customize Gemini to create your own personal AI experts on any topic you want, are now available for Gemini Advanced, Business and Enterprise users. And our new image generation model, Imagen 3, will be rolling out across Gemini, Gemini Advanced, Business and Enterprise in the coming days.<\/p>\n<hr \/>\n<p><a href=\"https:\/\/www.trendmicro.com\/en_us\/research\/24\/h\/agentic-ai-takeover.html\" target=\"_blank\" rel=\"noopener\"><strong>Cut the Chatter, Here Comes Agentic AI<\/strong><\/a>\u00a0&#8212; from trendmicro.com<\/p>\n<p style=\"padding-left: 40px;\">Major AI players caught heat in August over big bills and weak returns on AI investments, but it would be premature to think AI has failed to deliver. The real question is what\u2019s next, and if industry buzz and pop-sci pontification hold any clues, the answer isn\u2019t \u201cmore chatbots\u201d, it\u2019s agentic AI.<\/p>\n<p style=\"padding-left: 40px;\">Agentic AI transforms the user experience from application-oriented information synthesis to goal-oriented problem solving. It\u2019s what people have always thought AI would do\u2014and while it\u2019s not here yet, its horizon is getting closer every day.<\/p>\n<p style=\"padding-left: 40px;\">In this issue of\u00a0AI Pulse, we take a deep dive into agentic AI, what\u2019s required to make it a reality, and how to prevent \u2018self-thinking\u2019 AI agents from potentially going rogue.<br \/>\n&#8230;<br \/>\nCiting AWS guidance, ZDNET counts\u00a0six different potential types\u00a0of AI agents:<\/p>\n<ul>\n<li style=\"list-style-type: none;\">\n<ul>\n<li>Simple reflex agents for tasks like resetting passwords<\/li>\n<li>Model-based reflex agents for pro vs. con decision making<\/li>\n<li>Goal-\/rule-based agents that compare options and select the most efficient pathways<\/li>\n<li>Utility-based agents that compare for value<\/li>\n<li>Learning agents<\/li>\n<li>Hierarchical agents that manage and assign subtasks to other agents<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<hr \/>\n<p><a href=\"https:\/\/www.reuters.com\/technology\/artificial-intelligence\/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30\/\" target=\"_blank\" rel=\"noopener\"><strong>Ask Claude: Amazon turns to Anthropic&#8217;s AI for Alexa revamp<\/strong><\/a> &#8212; from reuters.com by Greg Bensinger<\/p>\n<p><strong>Summary:<\/strong><\/p>\n<ul>\n<li>Amazon developing new version of Alexa with generative AI<\/li>\n<li>Retailer hopes to generate revenue by charging for its use<\/li>\n<li>Concerns about in-house AI prompt Amazon to turn to Anthropic&#8217;s Claude, sources say<\/li>\n<li>Amazon says it uses many different technologies to power Alexa<\/li>\n<\/ul>\n<hr \/>\n<p><a href=\"https:\/\/venturebeat.com\/ai\/alibaba-releases-new-ai-model-qwen2-vl-that-can-analyze-videos-more-than-20-minutes-long\/\" target=\"_blank\" rel=\"noopener\"><strong>Alibaba releases new AI model Qwen2-VL that can analyze videos more than 20 minutes long<\/strong><\/a> &#8212; from venturebeat.com by Carl Franzen<\/p>\n<p><a href=\"https:\/\/venturebeat.com\/ai\/alibaba-releases-new-ai-model-qwen2-vl-that-can-analyze-videos-more-than-20-minutes-long\/\" target=\"_blank\" rel=\"noopener\"><img decoding=\"async\" class=\"alignnone\" src=\"https:\/\/venturebeat.com\/wp-content\/uploads\/2024\/08\/cfr0z3n_a_sleek_and_realistic_humanoid_robot_stands_upright_wit_f5b78bbc-ccf3-4fbf-8696-c89639f7dde6.png?w=750\" alt=\"\" width=\"100%\" height=\"100%\" \/><\/a><\/p>\n<hr \/>\n<p><a href=\"https:\/\/arstechnica.com\/information-technology\/2024\/08\/hobbyists-discover-how-to-insert-custom-fonts-into-ai-generated-images\/\" target=\"_blank\" rel=\"noopener\"><strong>Hobbyists discover how to insert custom fonts into AI-generated images<\/strong><\/a> &#8212; from arstechnica.com by Benj Edwards<br \/>\n<em>Like adding custom art styles or characters, in-world typefaces come to Flux.<\/em><\/p>\n<p><a href=\"https:\/\/arstechnica.com\/information-technology\/2024\/08\/hobbyists-discover-how-to-insert-custom-fonts-into-ai-generated-images\/\" target=\"_blank\" rel=\"noopener\"><img decoding=\"async\" class=\"alignnone\" src=\"https:\/\/cdn.arstechnica.net\/wp-content\/uploads\/2024\/08\/queen_custom_font_1-800x450.jpg\" alt=\"\" width=\"100%\" height=\"100%\" \/><\/a><\/p>\n<hr \/>\n<p><a href=\"https:\/\/www.zdnet.com\/article\/200-million-people-use-chatgpt-every-week-up-from-100-million-last-fall-says-openai\/\" target=\"_blank\" rel=\"noopener\"><strong>200 million people use ChatGPT every week &#8211; up from 100 million last fall, says OpenAI<\/strong><\/a> &#8212; from zdnet.com by Sabrina Ortiz<br \/>\n<em>Nearly two years after launching, ChatGPT continues to draw new users. Here&#8217;s why.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"<p>When A.I.\u2019s Output Is a Threat to A.I. Itself &#8212; from nytimes.com by Aatish Bhatia As A.I.-generated data becomes harder to detect, it\u2019s increasingly likely to be ingested by future A.I., leading to worse results. All this A.I.-generated information can make it harder for us to know what\u2019s real. And it also poses a problem [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[556,356,817,387,208,112,298,135,210,63,264,17,285,869,309,196,454,195,321,214,367,189],"tags":[],"class_list":["post-92812","post","type-post","status-publish","format-standard","hentry","category-amazon","category-artificial-intelligence-agents-llms-and-related","category-bots","category-business","category-cloud-based-computing-apps","category-corporate-business-world","category-data-related-items","category-design","category-emerging-technologies","category-google","category-graphics","category-learning-agents","category-legislation-legislatures","category-open-ai","category-platforms","category-productivity-tips-and-tricks","category-the-downsides-of-technology","category-tools","category-united-states","category-universities","category-vendors","category-videoconferencing"],"_links":{"self":[{"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/posts\/92812","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/comments?post=92812"}],"version-history":[{"count":16,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/posts\/92812\/revisions"}],"predecessor-version":[{"id":92844,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/posts\/92812\/revisions\/92844"}],"wp:attachment":[{"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/media?parent=92812"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/categories?post=92812"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/danielschristian.com\/learning-ecosystems\/wp-json\/wp\/v2\/tags?post=92812"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}