Meta's Llama AI Gets Military Clearance
Plus: Decart and Etched created the first playable open-world game, ChatGPT gets real-time web search, GitHub opens Copilot to AI model choice, and more.
Hello Engineering Leaders and AI Enthusiasts!
This newsletter brings you the latest AI updates in just 4 minutes! Dive in for a quick summary of everything important that happened in AI over the last week.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🔒 Meta opens Llama AI for US national security
🎮 Decart and Etched created the first playable open-world game
🔍 ChatGPT gets real-time web search
🎨 Recraft’s new image model beats OpenAI and Midjourney
🤖 GitHub opens Copilot to AI model choice
📚 Knowledge Nugget: Insurance for AI: Easier Said than Done by
Let’s go!
Meta opens Llama AI for US national security
Meta is expanding access to its open-source AI model, Llama, to US government agencies and defense contractors, just weeks after reports surfaced about Chinese military researchers using older versions of the technology. The company has partnered with tech giants, including AWS, Microsoft, IBM, and Lockheed Martin, to implement Llama across various national security applications.
The initiative brings together 13 major tech companies to provide secure cloud infrastructure for sensitive government data. It is already showing practical applications: Oracle is using Llama to streamline aircraft maintenance documentation, while Scale AI is fine-tuning the model for military mission planning and threat analysis.
Why does it matter?
This move is a strategic response to the evolving geopolitical AI landscape. As Chinese military researchers reportedly exploit open-source AI models, Meta's decision could change how American tech companies balance open-source principles with national security interests.
Decart and Etched created the first playable open-world game
AI labs Decart and Etched have launched Oasis, marking a breakthrough in AI-generated interactive environments. This first-of-its-kind system generates a complete video game in real time, processing everything from physics to graphics at 20 frames per second—100x faster than current text-to-video models.
Oasis lets players move, jump, and interact with the environment without a traditional game engine. The 500M parameter model, already available for local testing, showcases the potential of Decart's inference engine and Etched’s upcoming Transformer ASIC (Sohu).
Why does it matter?
This shows that traditional game engines will become obsolete shortly. Beyond gaming, AI has great potential to create interactive environments for training, simulation, and educational purposes.
ChatGPT gets real-time web search
OpenAI has upgraded ChatGPT with integrated web search capabilities. The new feature, rolling out first to Plus and Team users, combines natural language processing with live web data, offering everything from weather updates to stock quotes with source attribution. OpenAI has also partnered with major news outlets like the Associated Press, Financial Times, and Reuters to ensure high-quality content delivery.
The update includes a Chrome extension for direct browser searches and specialized displays for different content types (weather, stocks, sports). Users can also verify information through a new "Sources" button that links directly to referenced content, addressing previous concerns about AI hallucinations and fact-checking.
Why does it matter?
This battle for AI search supremacy has far-reaching implications for digital privacy, economic power, and information control. Whoever wins this race will dominate the tech industry and have unprecedented influence over how billions of people discover, consume, and interpret information.
Recraft’s new image model beats OpenAI and Midjourney
Recraft has launched its V3 model, which has secured the #1 spot on Hugging Face's Text-to-Image Leaderboard, surpassing Midjourney and OpenAI. What sets V3 apart is its unique ability to handle long-form text generation and precise positioning control—a first in the industry.
The platform offers features like vector image generation, style consistency controls, and a complete suite of AI editing tools. Recraft's new API also enables developers to integrate these capabilities into their workflows, with special attention to brand consistency and vector art generation.
Why does it matter?
By solving the text-handling limitations, Recraft is making AI a more practical tool than just an experimental technology for brand-conscious businesses and designers, accelerating the adoption of AI in creative industries.
GitHub opens Copilot to AI model choice
GitHub announced that developers can now choose between Anthropic's Claude 3.5 Sonnet, Google's Gemini 1.5 Pro, and OpenAI's o1-preview models when using Copilot's coding assistance. The platform also reveals GitHub Spark, allowing natural language app development with multi-file editing and improved code reviews.
The rollout begins with Copilot Chat, with OpenAI's models available immediately and others following in the coming weeks. Organizations and enterprises will have complete control over which models their teams can use, while individual developers can select their preferred model based on specific project needs.
Why does it matter?
This move shows how AI coding tools are transitioning from single-vendor solutions to an open marketplace of AI capabilities. By giving developers choice and flexibility, GitHub created an "app store moment" for AI coding tools, accelerating competition in the developer tools space.
Enjoying the latest AI updates?
Refer your pals to subscribe to our newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: Insurance for AI: Easier Said than Done
In his latest analysis,
challenges the growing enthusiasm around AI insurance products. He explains that the market faces significant structural challenges despite the apparent need for insuring against AI mistakes.The current Tech E&O insurance market remains surprisingly small at under $5B globally despite software being ubiquitous. Loeber points out that insurers struggle with information asymmetry, risk concentration among a few LLM providers, and rapid AI development that complicates traditional year-long policies.
Why does it matter?
The analysis reveals a crucial blindspot in AI commercialization—the challenge of effectively insuring against AI risks. This gap could slow AI adoption in critical applications or drive innovation in AI safety and testing frameworks.
What Else Is Happening❗
🤖 Anthropic's Claude 3.5 Haiku arrives with improved capabilities but controversy, bringing a 4x price increase despite similar performance to cheaper rivals.
🗣️ Hume debuts an AI companion app that combines the EVI 2 voice model with Claude 3.5 to offer emotionally intelligent conversations and guidance.
🎥 Runway adds pro camera controls to Gen-3 Alpha Turbo, enabling precise cinematography in AI videos while maintaining 3D consistency.
📄 Anthropic launches PDF support for Claude 3.5 Sonnet, analyzing both text and visuals in documents up to 100 pages and 32MB.
🚀 OpenAI's London Dev Day debuts o1 features and voice APIs, while Altman confirms no GPT-5 in 2024 but promises major releases ahead.
🦾 Nvidia reveals HOVER, a tiny 1.5M parameter AI that masters robot movement across multiple control methods with zero real-world training.
🤖 Boston Dynamics reveals new Atlas robot capabilities, including autonomously sorting car parts and adapting in real-time through machine learning.
👃 Osmo achieves a scent teleportation breakthrough, successfully analyzing and recreating a plum's exact smell using AI and molecular mapping.
🔥 OpenAI partners with Broadcom & TSMC to develop custom chips by 2026, adds AMD alongside NVIDIA to diversify its AI computing infrastructure.
🧬 Iambic Therapeutics launches Enchant AI, achieving 74% accuracy in predicting drug trial outcomes from minimal clinical data.
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you next week! 😊