Good morning. It’s Wednesday, July 24th.

Did you know: On this day in 2013, Google introduced the first generation Chromecast.

In today’s email:

• Meta’s LLaMA 3.1

• Elon’s AI Training Center

• Altman’s UBI push

• 2 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Meta Releases Largest Open-Source AI Model Set Yet

Meta has unveiled LLaMA 3.1, marking a significant leap in open-source AI capabilities. The standout feature of this release is the LLaMA 3.1 405B model, which boasts state-of-the-art performance rivaling the best closed-source models. Key advancements include an expanded context length of up to 128K, support for eight languages, and enhanced capabilities in general knowledge, steerability, and multilingual translation.

LLaMA 3.1 is designed to drive innovation across various applications, from synthetic data generation to model distillation, offering unprecedented flexibility and control to developers. This release is accompanied by new security tools, such as LLaMA Guard 3 and Prompt Guard, ensuring responsible AI development and deployment.

Meta's commitment to open-source AI is further evidenced by partnerships with over 25 companies, including AWS, NVIDIA, Databricks, Dell, and Google Cloud, providing extensive ecosystem support from day one. Developers can try LLaMA 3.1 405B on platforms like WhatsApp and meta.ai, and access the models for download on llama.meta.com and Hugging Face. Read More

Elon Musk fires up ‘the most powerful AI cluster in the world’ to create the 'world's most powerful AI' by December 

Elon Musk recently announced that xAI, his artificial intelligence company, has initiated the training of "the world's most powerful AI" using the newly operational Memphis Supercluster. This AI training cluster reportedly integrates 100,000 liquid-cooled Nvidia H100 GPUs interconnected by RDMA fabric, aiming for completion by December 2024. 

Musk has bypassed the anticipated H200 and Blackwell B100 GPUs, opting for a more immediate computational surge. The supercluster surpasses prominent computing systems like Frontier and Aurora in GPU capacity. Although Musk didn’t personally launch the supercluster, it began operations at 4:20 AM CDT. 

The setup, partly provided by Supermicro, marks a departure from Musk's Gigafactory of Compute's anticipated 2025 timeline. This development aligns with Musk's strategic shift from Tesla to xAI, showcasing a commitment to creating a highly competitive AI ecosystem with Grok 3. Read more. 

UBI Study Backed by OpenAI's Sam Altman Bolsters Support for Basic Income  

Sam Altman's OpenResearch has concluded the largest U.S. study on basic income. The three-year experiment, conducted in Illinois and Texas, provided $1,000 monthly stipends to 1,000 low-income individuals, while a control group of 2,000 received $50. Results indicate that the extra income was primarily used for essential needs like food, housing, and healthcare. 

Surprisingly, participants did not significantly reduce their work hours, instead reporting increased flexibility, autonomy, and improved financial planning, demonstrating a more future-oriented mindset. Single parents particularly benefited, using the funds to balance childcare and employment. 

The study also found disparities in spending patterns based on income levels, with lower-income participants prioritizing family support and housing. Altman has praised the research, funded in part by the National Institutes of Health, as a crucial step towards understanding universal basic income. Read more. 

Etcetera: Stories you may have missed

AI models struggle with "lost in the middle" issue when processing large image sets 

Google's new weather prediction system combines AI with traditional physics  

 Graph RAG: Access to external data becomes much more accurate with Microsoft's approach 

Study reveals rapid increase in web domains blocking AI models from training data 

Cohere raises $500M to beat back generative AI rivals  

A week of nonstop breaking political news stumps AI chatbots 

AWS launches generative AI-powered feature for Connect Contact Lens to help agents in call centers 

US Congress takes energy regulator FERC to task on AI data center electricity use 

Five thoughts from Matt Wood's keynote at AWS Summit New York 

AI model identifies certain breast tumor stages likely to progress to invasive cancer | MIT News 

Google AI slashes computer power needed for weather forecasts | New Scientist SAP: AI-Focused Restructuring to Add €200 Million to Bottom Line

Astronomers discover technique to spot AI fakes using galaxy-measurement tools | Ars Technica 

Studymap.ai allows users to create personalized learning plans using AI by converting YouTube videos and articles into structured courses.

Vozo enables users to transform videos using AI prompts, offering redubbing, voice editing, and lip-sync features for effortless storytelling.

📄The Vision of Autonomic Computing: Can LLMs Make It a Reality? 

📄 Knowledge Mechanisms in Large Language Models: A Survey and Perspective​​

📄 SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, apply here.