Rami's Readings #73 - ✨ Meta's Llama 3.1 Released! 🚀🔥
The latest on AI, LLMs, Llama 3.1, Mistral Large 2, RAG vs Long-Context, Intel CPUs, Happiness and Money, and more.
Welcome to Rami’s Readings #73 - a weekly digest of interesting articles, papers, videos, and X threads from my various sources across the Internet. Expect a list of reads covering AI, technology, business, culture, fashion, travel, and more. Learn about what I do at ramisayar.com/about.
Many announcements this week from new search engines (no comment, see #7) to multiple open source large language models breaking benchmarks! Let’s get into it!
🤖 AI Reads
Meta Announces Llama 3.1!
Notes: This is a huge release including a 405B parameter model that is effectively at the frontier of LLMs. All models support a long context length of 128k and grouped query attention for faster inference. It bests many of the commercial models on multiple benchmarks. Read the paper. Also, read Mark Zuckerberg’s letter on Open Source AI and watch his interview with Bloomberg. I am already using the Llama 3.1 8B model for a project through Ollama (easiest way for you to get started). This is a fantastic release (without any of the hiccups from Llama 3)!
Mistral Announces Mistral Large 2
Notes: Mistral announced their latest flagship model with impressive code generation and multilingual support.
MINT-1T: Scaling Open-Source Multimodal Data by 10x
Notes: Seattle Connection Alert! University of Washington.Excellent new open-source multimodal dataset with PDFs.
LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference
Notes: From Apple! Their research team is clearly focused on accelerating inference of smallish LLMs for Edge AI.
Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approach
Notes: Read this paper if you are building Generative AI applications.
We benchmark RAG and LC across various public datasets using three latest LLMs. Results reveal that when resourced sufficiently, LC consistently outperforms RAG in terms of average performance. However, RAG's significantly lower cost remains a distinct advantage. Based on this observation, we propose Self-Route, a simple yet effective method that routes queries to RAG or LC based on model self-reflection.
AI Achieves Silver-Medal Standard Solving International Mathematical Olympiad Problems
Notes: From Google DeepMind. It’s important to remember the following:
The problems were manually translated into formal mathematical language for our systems to understand.
💼 Business Reads
There Is No Fix for Intel’s Crashing 13th and 14th Gen CPUs — Any Damage Is Permanent
Notes: Beyond frustrating… I own an impacted CPU and Intel has lost much of my trust. I have a triple-fan CPU cooler just to keep the damn CPU from burning running models on my RTX 4090. Not to mention, Intel’s onboard Ethernet Controller randomly drops connections to my switch after entering power saving mode (none of the fixes online worked). I don’t think I will buy another Intel CPU again.
Warburg-Backed PDG Eyes Asia Data Center Expansion Fueled by AI
Notes: Follow the data center expansions.
Apple's AI Features Rollout Will Miss Upcoming iPhone Software Overhaul
Notes: Bloomberg is on fire with all these exclusives.
Southern Europe’s Prime Property Prices are Booming, Study Shows
Notes: Lisbon ❤️ Excited to go this fall and eat all the Pastéis de Natas.
🎨 Culture Reads
You Can Never Have Too Much Money, Happiness Researcher Finds
Notes: Hmmm…
Signing off from Redmond.
Get on the Referral Leaderboard!
If you enjoy Rami’s Readings, it would be incredible, amazing, star-worthy if you invited friends to subscribe and read with us. If you refer friends, you will receive fantastic benefits reflecting my gratitude for your contributions.
How to participate:
1. Share Rami’s Readings. You'll get credit for new subscribers when you use the referral link below or the “Share” button on any post. Send the link in a text, email, or share it on social media with friends.
2. Earn benefits. When more friends use your referral link to subscribe, you’ll receive the following.
Get 1 Hour Virtual Coffee Chat for 5 referrals - We can chat about anything you fancy, including LLMs, AI, Tech, Business, Economics, etc.
Get an In-Person Coffee Chat for 20 referrals - I will ✈️ fly out to meet you anywhere in the continental United States or Canada for an in-person chat at your preferred coffee shop.
Get 8 Hours of LLM and AI Mentorship for 50 referrals - I will mentor you and help you grow into a LLM and AI expert. I would be happy to fly out and meet with you for 8 hours, or if you prefer virtual sessions - glad to do that as well. I guarantee that you will be running Llama 2 on your machine by the end of these 8 hours.
To learn more about how Substack operates referrals, check out Substack’s FAQ.