🌎 Meta Pushes AI to the Masses

Plus Google’s infinite context window and Stanford’s AI report

Hello AI Enthusiast,

Meta is rolling out AI chatbots across all its platforms, making it even easier for everyone to use. But, what’s their strategy? We’ll dive deeper into this in the News Bytes section.

Speaking of chatbots, last week we asked about your experiences with them in customer service. Turns out, most of you (78%, to be exact) find these bots to be sometimes helpful and sometimes not. And surprise, surprise, NO ONE is completely satisfied with them. It looks like chatbots have some homework to do.

And while we're on the topic of making life easier for users, we often boast about our Generative AI Project Bootcamp that lets you build AI prototypes without coding, but talk is cheap, right? This time, we’re bringing in Alessandro, a past student turned AI Coach, to show you how it’s done. He'll reveal how he created a travel assistant during our program and continued developing AI prototypes.

Mark April 30th at 6 PM CEST on your calendar. Register now for our free webinar and watch Alessandro demonstrate the amazing things you can achieve with AI!

News Bytes 🗞️

  • Meta has launched the next generation of its Llama models, Meta Llama 3, which boasts 8B and 70B parameter language models enhanced with superior reasoning capabilities. Llama 3 is designed to deliver top-tier performance and stands as the best open-source model of its kind. Powering Meta AI, this new model is now expanding with features like image generation, search, and real-time assistance. Integrated on their platforms such as Facebook, Instagram, WhatsApp, and Messenger is now available in a few countries.

  • 💡 Our take: Meta isn't trying to outdo the top models; rather, their strategy is to retain its massive user base by enabling a variety of activities within its ecosystem commoditizing their complements. By making its model open-source, Meta not only disrupts other Big Tech companies that sell similar models but also enhances its brand reputation. Gianluca, our founder, has created an explanatory video about this strategy. 👇


Why we got a $1B AI present from Zuck 🎁 Strategy breakdown. #ai #meta #zuckerberg #business #strategy #startup #innovation #tech #opensource

Are you worried about the increase in data collection with these new Meta AI features?

Meta is enhancing its AI features, which probably means it'll be collecting more of your data to improve performance. We want to know how you feel about this.

Login or Subscribe to participate in polls.

  • A new paper by researchers from Google introduces Infinite attention, a modification to Transformer LLMs enabling them to process infinitely long contexts efficiently, outperforming other models while using significantly less memory. This advancement could revolutionize the customization of LLMs for specific applications.

  • 💡 Our take: Currently, the most effective method for uploading long-form documents to customize AI models is RAG, which demands some engineering effort. This new approach could be the solution for companies needing personalized knowledge. While it may not be as efficient as RAG, you can still use it to build a proof of concept and test your business ideas.

  • Stanford's HAI released their 2024 AI Index Report, underscoring the need for responsible AI. It highlights challenges in reporting standards and testing diversity, along with a rise in AI misuse and misinformation. The report calls for improved transparency, evaluation standards, and legal frameworks to better regulate AI.

  • 💡 Our take: It's beneficial to have research centers like this, encouraging more field studies that policymakers and regulators will likely cite as trusted references. These teams are essential for assessing the real-world impacts of new technologies. Their focus on a chapter about elections also emphasizes concerns over AI in election campaigns this year.

  • Brave's new "Answer with AI" feature uses AI to provide neatly formatted answers for a wide range of queries. Users can trigger AI search manually for non-informational queries.

  • 💡 Our take: AI-driven search engines are evolving to provide direct answers on the search page, bypassing traditional website clicks. This shift poses challenges for advertisers using search engines to promote their websites. The future lies in understanding how AI algorithms determine website rankings, signaling a paradigm shift in online user engagement and information retrieval.

  • Microsoft Research Asia introduced VASA-1, an AI framework that creates realistic talking faces from a static image and speech audio. This tool improves facial animation realism and expressiveness but also raises concerns about its potential misuse in creating convincing deepfakes.

  • Boston Dynamics is introducing a new all-electric Atlas robot with features like limitless 360-degree joints. This version promises to be commercially viable, stronger, and offers a broader range of motion with various gripper options for different needs. Gianluca has his candid take on this.

  • Sanctuary AI will deploy its humanoid robot at a Magna manufacturing facility in Canada, known for partnering with top automakers like Mercedes and BMW.

  • Adobe has launched the Firefly Image 3 Foundation Model, enabling fast ideation and higher-quality image creation for its creative community. Firefly integrates with popular Adobe applications like Photoshop, Illustrator, and InDesign.

  • The ongoing feud between singers Drake and Kendrick Lamar escalated with diss tracks and speculation that AI was used to imitate Tupac and Snoop Dogg's voices.

  • The BBC is investing £6 million in BBC Bitesize to enhance personalized and interactive learning experiences, aiming to engage young users and adapt to the digital age.

  • China's Baidu reports a doubling of users for its Ernie Bot chatbot since December, with the API being used 200 million times daily.

  • Amazon Music introduces Maestro, an AI playlist generator available in beta to select U.S. customers across all Amazon Music tiers. Maestro helps users create personalized playlists using prompts like emojis or activities.

Educational Pill 💊

Understanding the Context Window

A "context window" is the amount of text an AI can handle at one time. Normally, exceeding this window makes the AI forget earlier details. Google's "Infini-attention" technology expands this window significantly, allowing AI to process longer texts seamlessly, improving coherence and interaction quality.

Companies and individuals benefit from AI models that handle longer texts, allowing for seamless integration of detailed long-form documents. This enables more personalized solutions and efficient decision-making. With these AIs, developing custom applications requires minimal engineering, making sophisticated technology more accessible and practical across various industries.

LOLgorithms 😂

Anyone who says they weren't expecting that is lying.

From our community 🤝

We love the amazing work our AI Coaches are doing to help our students build AI prototypes. We make sure they receive the best possible support throughout the entire Bootcamp.

That's a wrap on our newsletter! Here’s a quick recap before you go:

Catch you next week! 👋