{"id":261856,"date":"2024-10-07T01:17:04","date_gmt":"2024-10-07T01:17:04","guid":{"rendered":"https:\/\/michigandigitalnews.com\/index.php\/2024\/10\/07\/gemini-nano-is-now-available-on-android-via-experimental-access\/"},"modified":"2025-06-25T17:10:57","modified_gmt":"2025-06-25T17:10:57","slug":"gemini-nano-is-now-available-on-android-via-experimental-access","status":"publish","type":"post","link":"https:\/\/michigandigitalnews.com\/index.php\/2024\/10\/07\/gemini-nano-is-now-available-on-android-via-experimental-access\/","title":{"rendered":"Gemini Nano is now available on Android via experimental access"},"content":{"rendered":"<p> [ad_1]<br \/>\n<\/p>\n<div>\n<meta content=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEgxutgVcuY1fxk1tNiS-78vKbC_swceezRTpeoj7hNrR1_-xUa5OblQmOVV0RT3wDXMIiLsp5w28krwGBAii1QreRbxn40uVmp1xXghJ625u-qY1ec1a_SJxo7sB6kpPT2PuxUUEezTTijF9-he4gKBBgiE_cbVKXcnJ6CGle-WMRN5GSLzJVHIYi32fjA\/s1600\/Gemini-Nano-on-Android-Social.png\" name=\"twitter:image\"\/><br \/>\n<img decoding=\"async\" src=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEgxutgVcuY1fxk1tNiS-78vKbC_swceezRTpeoj7hNrR1_-xUa5OblQmOVV0RT3wDXMIiLsp5w28krwGBAii1QreRbxn40uVmp1xXghJ625u-qY1ec1a_SJxo7sB6kpPT2PuxUUEezTTijF9-he4gKBBgiE_cbVKXcnJ6CGle-WMRN5GSLzJVHIYi32fjA\/s1600\/Gemini-Nano-on-Android-Social.png\" style=\"display: none;\"\/><\/p>\n<p><em>Posted by Taj Darra \u2013 Product Manager<\/em><\/p>\n<p><a href=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEjW_v6PdEkcKdIYGcJlt3Pj-xuMxwd8uDJA7LrrsxKLk5Bf-b2_cMs-H6pIp_rhA1ESVeIyVqfHJr0-fG8rrmQg51p5Ep6qmmsrYnjJ05fHqllzf7GB1AudjmtZ1xo8Tde49lIn174-5laJq0yW8mJR2eFoQrRjI4arvggUe7buzn7cF672Wo-mujhDFuk\/s1600\/AI-on-Android-Google-AI-Edge%20%286%29.png\"><img decoding=\"async\" border=\"0\" data-original-height=\"800\" data-original-width=\"100%\" src=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEjW_v6PdEkcKdIYGcJlt3Pj-xuMxwd8uDJA7LrrsxKLk5Bf-b2_cMs-H6pIp_rhA1ESVeIyVqfHJr0-fG8rrmQg51p5Ep6qmmsrYnjJ05fHqllzf7GB1AudjmtZ1xo8Tde49lIn174-5laJq0yW8mJR2eFoQrRjI4arvggUe7buzn7cF672Wo-mujhDFuk\/s1600\/AI-on-Android-Google-AI-Edge%20%286%29.png\"\/><\/a><\/p>\n<p><a href=\"https:\/\/deepmind.google\/technologies\/gemini\/\" target=\"_blank\" rel=\"noopener\">Gemini<\/a>, introduced last year, is Google\u2019s most capable family of models yet; designed for flexibility, it can run on everything from data centers to mobile devices. Since announcing <a href=\"https:\/\/developer.android.com\/ai\/aicore\" target=\"_blank\" rel=\"noopener\">Gemini Nano<\/a>, our most efficient model built for on-device tasks, we&#8217;ve been working with a limited set of partners to support a range of use cases for their apps.<\/p>\n<p><b>Today, we\u2019re opening up access to experiment with Gemini Nano<\/b> to all Android developers with the AI Edge SDK via AICore. Developers will initially have access to experiment with text-to-text prompts on Pixel 9 series devices. Support for more devices and modalities will be added in the future. Check out our <a href=\"https:\/\/developer.android.com\/ai\/gemini-nano\/experimental\" target=\"_blank\" rel=\"noopener\">documentation<\/a> and <a href=\"https:\/\/youtu.be\/EpKghZYqVW4\" target=\"_blank\" rel=\"noopener\">video<\/a> to get started. Note that experimental access is for development purposes, and is not for production usage at this time.<\/p>\n<p><\/p>\n<p><iframe title=\"A walkthrough for Android\u2019s on-device GenAI solutions | Spotlight Week\" width=\"1200\" height=\"675\" src=\"https:\/\/www.youtube.com\/embed\/EpKghZYqVW4?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe><\/p>\n<h3>Fast, private and cost-effective on-device AI<\/h3>\n<p>On-device generative AI processes prompts directly on your device without server calls. It offers many benefits: sensitive user data is processed locally on the device, full functionality without internet connectivity, and no additional monetary cost for each inference.<\/p>\n<p>Since on-device generative AI models run on devices with less computational power than cloud servers, they are significantly smaller and less generalized than their cloud-based equivalents. As a result, the model works best for tasks where the requests can be clearly specified rather than open-ended use cases such as chatbots. Here are some use cases you can try:<\/p>\n<ul>\n<ul>\n<li><b>Rephrasing<\/b> &#8211; Rephrasing and rewriting text to change the tone to be more casual or formal.<\/li>\n<\/ul>\n<ul>\n<li><b>Smart reply<\/b> &#8211; Given several chat messages in a thread, suggest the next likely response.<\/li>\n<\/ul>\n<ul>\n<li><b>Proofreading<\/b> &#8211; Removing spelling or grammatical errors from text.<\/li>\n<\/ul>\n<ul>\n<li><b>Summarization<\/b> &#8211; Generating a summary of a long document, either as a paragraph or as bullet points.<\/li>\n<\/ul>\n<\/ul>\n<p>Check out our <a href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/prompting-strategies\" target=\"_blank\" rel=\"noopener\">prompting strategies<\/a> to achieve best results when experimenting with the above use-cases. If you want to test your own use case, you can download our <a href=\"https:\/\/github.com\/android\/ai-samples\" target=\"_blank\" rel=\"noopener\">sample app<\/a> for an easy way to start experimenting with Gemini Nano.<\/p>\n<p><\/p>\n<h3>Gemini Nano performance and usage<\/h3>\n<p>Compared to its predecessor, the model being made available to developers today (referred to in the <a href=\"https:\/\/arxiv.org\/pdf\/2312.11805\" target=\"_blank\" rel=\"noopener\">academic paper<\/a> as \u201cNano 2\u201d) delivers a substantial improvement in quality. At nearly twice the size of the predecessor (\u201cNano 1\u201d), it excels in both academic benchmarks and real-world applications, offering capabilities that rival much larger models.<\/p>\n<div align=\"center\">\n<table style=\"border-collapse: collapse; border: none; width: 100%;\">\n<tbody>\n<tr>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\"><\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\"><b>MMLU (5-shot)*<\/b><\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\"><b>MATH (4-shot)*<\/b><\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\"><b>Paraphrasing**<\/b><\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\"><b>Smart Reply**<\/b><\/span><\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n                    <b>Nano 1<br \/>\n                <\/b><\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">46%<\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">14%<\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">44%<\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">44%<\/span><\/p>\n<\/td>\n<\/tr>\n<tr>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\"><b>Nano 2<br \/>\n                    <\/b><\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\"><span style=\"font-size: 11pt;\">56%<\/span><\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">23%<\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">90%<\/span><\/p>\n<\/td>\n<td style=\"border: 1pt solid rgb(0, 0, 0);\">\n<p><span face=\"Arial,sans-serif\" style=\"font-size: 11pt;\">82%<\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><i>** Percentage of good answers measured on public datasets via an autorater powered by Gemini 1.5 Pro.<\/i><\/p>\n<p>Gemini Nano is already in use by Google apps. Pixel Screenshots, Talkback, Recorder and many more have leveraged Gemini Nano\u2019s text and image understanding to deliver new experiences:<\/p>\n<ul>\n<ul>\n<li><a href=\"https:\/\/android-developers.googleblog.com\/2024\/09\/talkback-uses-gemini-nano-to-increase-low-vision-accessibility.html\" target=\"_blank\" rel=\"noopener\">Talkback<\/a> &#8211; Android\u2019s accessibility app leverages Gemini Nano\u2019s multimodal capabilities to improve image descriptions for blind and low vision users.<\/li>\n<\/ul>\n<p><image><\/p>\n<div style=\"text-align: center;\"><img decoding=\"async\" alt=\"moving image of Talkback app UI highlighting improved image descriptions with multimodality model for users with low vision\" border=\"0\" id=\"imgCaption\" src=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEjX0yAAaKl4Z5FecqP-h6FiDfS2aqeSUsNTrQHaRUG0FOPkyIM5l4Z_V5RUgsXKvfXN1i9yLvKnE-7T1Yh3iawnG552ATln225N2SeeRSaQDbtboH_MxPXTqNlETbR59CyMjaZSTupKdajd3-Oj3_x984jmhZYNz54rfKhGnpnFAOVLzE0KYwW6ow67nv8\/s16000\/Nano-Talkback.gif\"\/><\/div>\n<p><\/image><\/p>\n<ul>\n<li><a href=\"https:\/\/android-developers.googleblog.com\/2024\/08\/recorder-app-on-pixel-sees-boost-in-engagement-with-gemini-nano.html\" target=\"_blank\" rel=\"noopener\">Pixel Recorder<\/a> &#8211; Gemini Nano with Multimodality model enables support for longer recordings and higher quality summaries.<\/li>\n<\/ul>\n<\/ul>\n<p><image><\/p>\n<div style=\"text-align: center;\"><img decoding=\"async\" alt=\"moving image of Talkback app UI highlighting improved image descriptions with multimodality model for users with low vision\" border=\"0\" id=\"imgCaption\" src=\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEjIHHaBqn4hL9C54peGuUeQFTOmwQUS0UsUXIVCcEf9Oe0eWqFVlp-ssiOnX7mzV8PbhkNhxmuqu84dx5j0XlFRHBwVjfYw0TP_Nf_InOQmuX_LWBJIj1G8JTINEauzYoFphdSC5bXlaH1n8pfKpB0gku1t_v1TvYytk2vs4MPPQMmozIQ0vGLo2Rq4ELc\/s16000\/Nano-Recorder.gif\"\/><\/div>\n<p><\/image><\/p>\n<h3>Seamless model integration with AI Edge SDK using AICore <\/h3>\n<p>Integrating generative AI models directly into mobile apps is challenging due to the significant computational resources and storage space they require. To address this challenge, we developed <a href=\"https:\/\/developer.android.com\/ai\/aicore\" target=\"_blank\" rel=\"noopener\">AICore<\/a>, a new system service in Android. AICore allows you to benefit from AI running directly on the device without needing to distribute runtimes, models and other components yourself.<\/p>\n<p>To run inference with Gemini Nano in AICore, you use the AI Edge SDK. The AI Edge SDK enables developers to customize prompts and inference parameters to their specific needs, enabling greater control over each inference.<\/p>\n<p>To experiment with the AI Edge SDK, add the following to your apps\u2019 dependency:<\/p>\n<div style=\"background: rgb(248, 248, 248); border: 0px; overflow: auto; width: auto;\">\n<pre style=\"line-height: 125%; margin: 0px;\">implementation(<span style=\"color: #ba2121;\">\"com.google.ai.edge.aicore:aicore:0.0.1-exp01\"<\/span>)\n<\/pre>\n<\/div>\n<p>The AI Edge SDK allows you to customize <a href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/models\/generative-models#model-parameters\" target=\"_blank\" rel=\"noopener\">inference parameters<\/a>. Some of the more commonly-used parameters include:<\/p>\n<ul>\n<ul>\n<li><b>Temperature<\/b>, which controls randomness. Higher values increase diversity and creativity of output.<\/li>\n<\/ul>\n<ul>\n<li><b>Top K<\/b>, which specifies how many tokens from the highest-ranking ones are to be considered.<\/li>\n<\/ul>\n<ul>\n<li><b>Candidate count<\/b>, which describes the maximum number of responses to return.<\/li>\n<\/ul>\n<ul>\n<li><b>Max output tokens<\/b>, which is the length of the desired response.<\/li>\n<\/ul>\n<\/ul>\n<p>When you are ready to run the inference with your model, the AI Edge SDK offers an easy way to pass in multiple strings as input to accommodate long inference data.<\/p>\n<p>Here\u2019s an example:<\/p>\n<div style=\"background: #f8f8f8; overflow:auto;width:auto;border:0;\">\n<pre style=\"margin: 0; line-height: 125%\">scope.launch {\n    <span style=\"color: #408080; font-style: italic\">\/\/ Single string input prompt<\/span>\n    <span style=\"color: #008000; font-weight: bold\">val<\/span> input = <span style=\"color: #BA2121\">\"I want you to act as an English proofreader. I will <\/span>\n    provide you texts, and I would like you to review them <span style=\"color: #008000; font-weight: bold\">for<\/span> any \n    spelling, grammar, or punctuation errors. Once you have finished \n    reviewing the text, provide me with any necessary corrections or \n    suggestions <span style=\"color: #008000; font-weight: bold\">for<\/span> improving the text: \n    These arent the droids your looking <span style=\"color: #008000; font-weight: bold\">for<\/span>.<span style=\"color: #BA2121\">\"<\/span>\n    <span style=\"color: #008000; font-weight: bold\">val<\/span> response = generativeModel.generateContent(input)\n    print(response.text)\n\n    <span style=\"color: #408080; font-style: italic\">\/\/ Or multiple strings as input<\/span>\n    <span style=\"color: #008000; font-weight: bold\">val<\/span> response = generativeModel.generateContent(\n        content {\n            text(<span style=\"color: #BA2121\">\"I want you to act as an English proofreader.I will <\/span>\n            provide you texts and I would like you to review them <span style=\"color: #008000; font-weight: bold\">for<\/span> \n            any spelling, grammar, or punctuation errors.<span style=\"color: #BA2121\">\")<\/span>\n            text(<span style=\"color: #BA2121\">\"Once you have finished reviewing the text, <\/span>\n            provide me with any necessary corrections or suggestions \n            <span style=\"color: #008000; font-weight: bold\">for<\/span> improving the text:<span style=\"color: #BA2121\">\")<\/span>\n            text(<span style=\"color: #BA2121\">\"These arent the droids your looking for.\"<\/span>)\n        }\n    )\n    print(response.text)\n}\n<\/pre>\n<\/div>\n<p>Our <a href=\"https:\/\/developer.android.com\/ai\/gemini-nano\/experimental\" target=\"_blank\" rel=\"noopener\">integration guide<\/a> has more information on the AI Edge SDK as well as detailed instructions to start your experimentation with Gemini Nano. To learn more about prompting, check out the <a href=\"https:\/\/ai.google.dev\/gemini-api\/docs\/prompting-strategies\" target=\"_blank\" rel=\"noopener\">Gemini prompting strategies<\/a>.<\/p>\n<p><\/p>\n<h3>Get Started<\/h3>\n<p>Learn more about Gemini Nano for app development by watching our <a href=\"https:\/\/youtu.be\/EpKghZYqVW4\" target=\"_blank\" rel=\"noopener\">video walkthrough<\/a>, and <a href=\"https:\/\/developer.android.com\/ai\/gemini-nano\/experimental\" target=\"_blank\" rel=\"noopener\">try out Gemini Nano experimental access<\/a> in your own app today.<\/p>\n<p>We are excited to see what you build and welcome your input as you evaluate this new technology for your use cases! Post your creations on social media and include the hashtag #AndroidAI to share what you build. To share your ideas and feedback for on-device GenAI and help shape our APIs, you can <a href=\"https:\/\/issuetracker.google.com\/issues\/new?component=1657650&amp;template=0&amp;pli=1\" target=\"_blank\" rel=\"noopener\">file a ticket<\/a>.<\/p>\n<p>There\u2019s a lot more that we\u2019re covering this week for you to build great AI experiences on Android so be sure to <a href=\"https:\/\/android-developers.googleblog.com\/2024\/09\/welcome-to-ai-on-android-spotlight-week.html\" target=\"_blank\" rel=\"noopener\">check out the rest of the AI on Android Spotlight Week content<\/a>! <\/p>\n<\/div>\n<p>[ad_2]<br \/>\n<br \/><a href=\"http:\/\/android-developers.googleblog.com\/2024\/10\/gemini-nano-experimental-access-available-on-android.html\">Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>[ad_1] Posted by Taj Darra \u2013 Product Manager Gemini, introduced last year, is Google\u2019s most capable family of models yet; designed for flexibility, it can<\/p>\n","protected":false},"author":1,"featured_media":261857,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_uf_show_specific_survey":0,"_uf_disable_surveys":false,"footnotes":""},"categories":[146],"tags":[],"_links":{"self":[{"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/posts\/261856"}],"collection":[{"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/comments?post=261856"}],"version-history":[{"count":0,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/posts\/261856\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/media\/261857"}],"wp:attachment":[{"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/media?parent=261856"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/categories?post=261856"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/michigandigitalnews.com\/index.php\/wp-json\/wp\/v2\/tags?post=261856"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}