{"id":721,"date":"2022-12-31T13:40:01","date_gmt":"2022-12-31T13:40:01","guid":{"rendered":"https:\/\/tbekk.com\/devstream\/?p=721"},"modified":"2022-12-31T13:46:45","modified_gmt":"2022-12-31T13:46:45","slug":"an-open-source-version-of-chatgpt-is-coming","status":"publish","type":"post","link":"https:\/\/tbekk.com\/devstream\/2022\/12\/31\/an-open-source-version-of-chatgpt-is-coming\/","title":{"rendered":"An Open-Source Version of ChatGPT is Coming"},"content":{"rendered":"\n<hr class=\"wp-block-separator is-style-wide\"\/>\n\n\n\n<ul class=\"wp-block-list\"><li><em><strong>Link: <\/strong><a href=\"https:\/\/metaroids.com\/news\/an-open-source-version-of-chatgpt-is-coming\/\">an-open-source-version-of-chatgpt-is-coming\/<\/a><\/em><\/li><li><em><strong>Publication date: <\/strong>December, 31st 2022<\/em><\/li><li><em><strong>Author: <\/strong><\/em><a href=\"https:\/\/metaroids.com\/author\/falkris-the-lowlife\/\"><em>Falkris<\/em><\/a><\/li><\/ul>\n\n\n\n<hr class=\"wp-block-separator is-style-wide\"\/>\n\n\n\n<p><strong>Philip Wang, the mastermind behind the reverse-engineering of several closed-source AI systems like DALL-E 2, AlphaFold, and Imagen, has just released an implementation of&nbsp;<a href=\"https:\/\/github.com\/lucidrains\/PaLM-rlhf-pytorch\" target=\"_blank\" rel=\"noreferrer noopener\">PaLM + RLHF<\/a>, a text-generating model that works just like ChatGPT.<\/strong><\/p>\n\n\n\n<p>This system combines Google\u2019s PaLM, a large language model with 540 billion parameters (<a href=\"https:\/\/metaroids.com\/news\/googles-ai-is-allegedly-3x-more-powerful-than-chatgpt\/\" target=\"_blank\" rel=\"noreferrer noopener\">3x more parameters than ChatGPT<\/a>), and a technique called Reinforcement Learning with Human Feedback (RLHF) to allow for the creation of a chatbot that can do pretty much anything ChatGPT can, including answering general questions, drafting emails, and suggesting computer code.<\/p>\n\n\n\n<p>Note that this version isn\u2019t exactly the same as the PaLM model developed by Google a few years ago, but it has a very similar architecture and approach. Wang is well-known for being able to \u201cport\u201d several other famous architectures, hence, he might know a thing or two about emulating these models.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The Power of PaLm + RLHF<\/h2>\n\n\n\n<p>Since its release, ChatGPT, a fine-tuned version of GPT-3.5, has taken the tech world by storm due to its ability to generate human-like text with high clarity, allowing it to respond in a conversational manner. While it may be a significant advancement from earlier chatbots, many proponents in the AI field have&nbsp;<a href=\"https:\/\/metaroids.com\/news\/chatgpt-could-end-open-research-in-deep-learning-says-ex-google-employee\/\" target=\"_blank\" rel=\"noreferrer noopener\">raised concerns over the closed nature of ChatGPT<\/a>.<\/p>\n\n\n\n<p>As of today, the ChatGPT model remains proprietary, which means that the public cannot view its underlying code. Only OpenAI truly knows how it works and what data it processes. This lack of transparency can have far-reaching implications and may affect trust from users in the long term.<\/p>\n\n\n\n<p>Many developers have been eager to build an open-source alternative and now it has finally arrived. PaLM + RLHF is built exclusively for the Python language, and can be implemented for PyTorch. Developers can easily train PaLM like an autoregressive transformer and then train the reward model using human feedback.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What\u2019s The Catch?<\/h2>\n\n\n\n<p>Well, you can\u2019t use it today yet. So what would it take for it to be available to the public?<\/p>\n\n\n\n<p>To launch PaLM + RLHF, you will need to compile gigabytes of text taken from various sources such as blogs, social media posts, news articles, e-books, etc. These data are fed to the fine-tuned PaLm model, which will generate several responses. For example, queries like \u201cwhat are the basics of Economics,\u201d might yield responses like \u201cEconomics is the social science that studies\u2026\u201d<\/p>\n\n\n\n<p>Human volunteers will be employed to rank those responses from best to worst, using the rankings to create a reward model that takes the original model\u2019s responses and sorts them in order of preference, filtering for the top answers to a given prompt.<\/p>\n\n\n\n<p>However, the process of aligning this model with what users want to accomplish with ChatGPT is both costly and time-consuming, as PaLM has a massive 540 billion parameters. Note that the&nbsp;<a href=\"https:\/\/arxiv.org\/pdf\/2004.08900.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">cost<\/a>&nbsp;of developing a text-generating model with only 1.5 billion parameters can reach up to $1.6 million.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">When Can I Use an Open-Source ChatGPT?<\/h2>\n\n\n\n<p>At the moment, it is unclear how many organizations have the technicals and financial infrastructure to run the implementation built by Wang without degrading their user experience. So far, we have three known players working on this open-source ChatGPT alternative:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/github.com\/CarperAI\/trlx\" target=\"_blank\" rel=\"noreferrer noopener\">CarperAI<\/a>&nbsp;(in partnership with Hugging Face, Scale AI, and EleutherAI)<\/li><li><a href=\"https:\/\/github.com\/LAION-AI\/Open-Assistant\" target=\"_blank\" rel=\"noreferrer noopener\">LAION<\/a>&nbsp;\u2013 the non-profit that supplied the dataset used to train Stable Diffusion<\/li><li><a href=\"https:\/\/youtu.be\/sswA4j_IUxg\" target=\"_blank\" rel=\"noreferrer noopener\">Yannic Kilcher<\/a><\/li><\/ul>\n\n\n\n<p>None of them have given any hard dates for release as of yet. But to give you a rough timeline, it took three months to train&nbsp;<a href=\"https:\/\/huggingface.co\/bigscience\/bloom\" target=\"_blank\" rel=\"noreferrer noopener\">Bloom<\/a>, an open-source model that boasts 176 billion parameters. With that in mind, it might take more than 6-8 months for us to see a worthwhile release.<\/p>\n\n\n\n<p>For now, all we can do is wait. In the meantime, it might be a great idea to continue using ChatGPT whilst it\u2019s free. It\u2019s also worth noting that by the time an open clone exists, OpenAI might be far ahead in development. There are many rumors of GPT-4 being ultra-powerful and is set for release in 2023.<\/p>\n\n\n\n<hr class=\"wp-block-separator is-style-default\"\/>\n\n\n\n<p><strong>Join our newsletter as we build a community of metaverse and web3 pioneers.<\/strong><\/p>\n\n\n\n<p>The next 3-5 years is when new industry titans will emerge, and we want you to be one of them.<\/p>\n\n\n\n<p><strong>Benefits include<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Receive updates on the most significant trends<\/li><li>Receive crucial insights that will help you stay ahead in web3<\/li><li>Access whitelist spots of the most hyped NFT drops<\/li><li>The chance to be part of our OG community, which will have exclusive membership perks<\/li><\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Link: an-open-source-version-of-chatgpt-is-coming\/ Publication date: December, 31st 2022 Author: Falkris Philip Wang, the mastermind behind the reverse-engineering of several closed-source AI systems like DALL-E 2, AlphaFold, and Imagen, has just released&#8230; <a class=\"read-more-link\" href=\"https:\/\/tbekk.com\/devstream\/2022\/12\/31\/an-open-source-version-of-chatgpt-is-coming\/\">Read more &raquo;<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[66,9],"tags":[174,162,179,165,166,175,176,178,177,163,173],"class_list":["post-721","post","type-post","status-publish","format-standard","hentry","category-ai","category-news","tag-chat-bots","tag-chatgpt","tag-google","tag-large-language-models","tag-llms","tag-palm","tag-pathways-language-model","tag-reinforcement-learning-with-human-feedback","tag-rlhf","tag-text-ai","tag-text-generation"],"_links":{"self":[{"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/posts\/721","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/comments?post=721"}],"version-history":[{"count":1,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/posts\/721\/revisions"}],"predecessor-version":[{"id":722,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/posts\/721\/revisions\/722"}],"wp:attachment":[{"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/media?parent=721"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/categories?post=721"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tbekk.com\/devstream\/wp-json\/wp\/v2\/tags?post=721"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}