Hello, hi! Itâs Friday, a great day to ship to production. Today, Iâm talking about Zuckâs glowup and Metaâs new...
[Product Hunt]( [Read in browser](
[This newsletter is brought to you by]( TOP NEWS Hello, hi! Itâs Friday, a great day to ship to production. Today, Iâm talking about Zuckâs glowup and Metaâs new AI models. But first⦠The headlines: ð [The UK is weighing up whether AI can help protect younger internet users](. ð° [Mistral, the French AI company, wants to raise $500m at a $5B valuation.]( ð [Apple has removed Threads and WhatsApp from the Chinese app store.]( PRODUCT HIGHLIGHT
[Meta just announced its largest and most impressive AI models yet]( Meta isnât going to just sit back and watch OpenAI run away with the AI trophy. Quite the opposite, actually, based on yesterdayâs announcement. ICYMI: Mark Zuckerberg, or should I say â[Zuck](,â took to [Instagram Live]( to announce [Llama 3](, the companyâs largest suite of AI models to date. So whatâs the big deal? Well, a few things. First is the size of the models. In total, there are three of them. The two smallest, Llama 8B and 70B, contain 8 billion and 70 billion parameters (settings that guide AI's learning process), respectively. A third and much larger model, Llama 400B, contains 400 billion parameters. Yes, you read that right. Second is the amount of data these models are trained on. [According to the company announcement](, Llama 3 models were trained on over 15 trillion tokens (specific units like words that models are trained on) while the previous generation, Llama 2, was trained on 2 trillion tokens â a respectable amount, for sure, but eclipsed by todayâs model. Third is how it compares to other models on the market. Meta claims Llama 8B and 70B are âamong the best-performing models for their respective parameter count.â Thatâs a big claim to make, so how does it actually stack up? [Meta points to Llama 3âs score]( across different AI benchmarks like MMLU and ARC, which test models based on a number of questions ranging from biology to mathematics. According to those scores, Llama 8B beats the likes of [Mistral 7B]( and [Gemini 7B](, while Llama 70B beats Gemini 1.5 Pro and comes up just short of Anthropics most powerful model, [Claude 3 Opus.]( Like previous models, Llama 3 is entirely [open-source](, so developers can contribute to its direction and improvements. Who knows, maybe we will crowdsource AGI? If you want to test out Llama 3, head over to [Metaâs standalone AI website]( where you can chat directly with the currently available models (8B and 70B) as long as youâre in a country where itâs currently available. Alongside that, you can try out Metaâs AI assistant [directly on Instagram now.Â]( [Check out Llama 3]( Sponsored By AI builders: There's a new security standard in town and you'll need to find out what it is, if you need it, and how it works. ISO 42001 was introduced by the International Standards Organization so that companies can demonstrate their security practices around AI in a verifiable way. Join Vanta and A-LIGN in a live webinar to learn about... - The components of ISO 42001. Understand it and its relevance.
- Whether you need it. Kearn who benefits from it and how ISO 42001 ensures operational excellence in AI.
- Best practices and strategies for successfully integrating ISO 420001.
[Register Here]( Friday, 19 April 2024
Yesterday's Top Products
[Collabwriting for Teams â The easiest way to capture, store, and find knowledge]( [â²Â 582+](
[Pullpo.io â Complete code reviews in hours, not days, right from Slack]( [â²Â 432+](
[Tabnine â The private, personalized, & protected AI coding assistant.]( [â²Â 333+](
[Latitude â Open-source framework for embedded analytics]( [â²Â 409+](
[Cascadeur 2.0 â You can clean up and edit assets and mocap with AI]( [â²Â 356+](
[Tabula â Startup analytics, accelerated by AI]( [â²Â 277+](
[Brave Search â Brave Searchâs AI-powered answer engine]( [â²Â 234+](
[Sweego â Deliver your messages in the blink of an eye]( [â²Â 202+](
[Delta â The first Gameboy emulator approved in the App Store]( [â²Â 194+](
[OptimateAI â AI sales software, get paid & boost sales]( [â²Â 175+](
[AlphaCorp AI â Group chat with AIs]( [â²Â 166+](
[Discover many more]( CAT NIPS - [Nothing]( launched a new set of colorful, AI-enabled earphones.
- [Ugly Duckling]( is an analytics tool for being more consistent on X.
- [HuggingFace]( launched a chatbot app for iOS. SHOUTOUTS Shout out are where makers call out the products that help make their own possible. That doesnât mean theyâre all developer tools! Take this story: [Ugly Duckling]( is a web app for X analytics by #buildinpublic maker Norbert Jurga. It first launched a year ago, but when Elon cracked down on Twitter scraping, things were looking ugly for that duck. Now, Jurga has relaunched as a web app that uses the Xâs API, and he shouted out the [Morning Maker Show]( for helping him overcome roadblocks at his low points. The Morning Maker Show, hosted by Sandra Djajic and Dan Mindru, is a live, weekday audio show focused on topics special to independent makers and those #buildinginpublic. [Check it out](. MAKER'S CORNER - [Pietra]( launched a new tool for generating product images with AI.
- [Bridge]( is a platform that allows users to give feedback more easily.Â
- [Cascadeur]( helps you clean up and edit 3D assets with AI. [Tell us how we're doing](
[We're working on improving Product Hunt every day and we want to hear from you. What do you love? How could we do better? All feedback is helpful.](
[Speak your mind](
We send this email daily. Feel free to [update your preferences]( or [unsubscribe]( from these emails at any time. [Advertise](, [Twitter](, [Facebook](, [Stories](