Eduardo Ordax’s Post

Morning Santiago 🇨🇱 Buenos días Santiago 🇨🇱 Thrilled to spend a week with my Amazon Web Services (AWS) colleagues in Chile where I’ll be visiting customers and giving the GenAI Keynote this Thursday. Let’s rock 🤘🏼 Amazing views of “Los Andes” (Not AI generated of course) Always safe flying with Iberia #ai #travelling

𝐖𝐡𝐞𝐧 𝐭𝐨 𝐔𝐬𝐞 𝐋𝐥𝐚𝐦𝐚𝟑 𝟕𝟎𝐁 𝐯𝐬 𝟖𝐁: 𝐀 𝐐𝐮𝐢𝐜𝐤 𝐆𝐮𝐢𝐝𝐞 𝐭𝐨 𝐒𝐦𝐚𝐥𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥 𝐚𝐧𝐝 𝐋𝐚𝐫𝐠𝐞 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 🦙 While we’re still waiting for the arrival of Llama3 400B, open source community agrees 70B and 8B versions are powerful models, but each has its ideal use cases. Below some tips 👇 𝐋𝐥𝐚𝐦𝐚𝟑 𝟕𝟎𝐁: 𝐖𝐡𝐞𝐧 𝐭𝐨 𝐔𝐬𝐞 𝐈𝐭 🔸Complex Tasks: If your application involves understanding nuanced language, generating detailed content, or performing complex analysis, the 70B model’s larger capacity can handle it better. 🧠 🔸High Accuracy: For tasks where accuracy and precision are paramount, like medical research, legal analysis, or high-stakes decision-making, the 70B model provides a deeper understanding and more reliable outputs. 🎯 🔸Large-Scale Deployments: When you have the computational resources to support it and need to serve large user bases with diverse queries, the 70B model can manage high volumes while maintaining quality. 🌐 🔸Advanced AI Research: For cutting-edge AI research and development, where pushing the boundaries of what’s possible is the goal, the 70B model offers more capabilities. 🔬 𝐋𝐥𝐚𝐦𝐚𝟑 𝟖𝐁: 𝐖𝐡𝐞𝐧 𝐭𝐨 𝐔𝐬𝐞 𝐈𝐭 🔸Speed and Efficiency: If you need quick responses with lower computational costs, the 8B model is more efficient and faster. Ideal for real-time applications like chatbots or interactive systems. 🚀 🔸Resource Constraints: When working with limited hardware or budget, the 8B model is more cost-effective and requires fewer resources. Perfect for startups and smaller projects. 💡 🔸Less Complex Tasks: For straightforward tasks such as simple text generation, summarization, or moderate-level question answering, the 8B model is sufficient and performs well. 📝 🔸Testing and Prototyping: When developing new applications and needing to iterate quickly, the 8B model allows for faster experimentation and prototyping without heavy computational demands. ⚙️ Choose wisely based on your specific needs, use cases and resources to get the best out of these powerful LLMs. Wondering how wooly 400B will be…. #AI #Llama3 #ML #GenAI #LLM

Another week has gone and it brought lot of exciting stuff! As usual, enjoy your Sunday morning with a Nespresso (tea is also valid) and some dosis of Artificial Intelligence. Exciting news this week and just getting ready to fly to Chile 🇨🇱 in just a few hours! 𝗦𝘂𝗻𝗱𝗮𝘆: LLMs Hallucinations: Feature or Bug?; CNN by Yann LeCun; 𝗠𝗼𝗻𝗱𝗮𝘆: Unlocking the secrets behind your product design; 𝗧𝘂𝗲𝘀𝗱𝗮𝘆: GenAI from Idea to Production; Why can't AI nail human hands?; 𝗪𝗲𝗱𝗻𝗲𝘀𝗱𝗮𝘆: GenAI Unicorns; MOSHI by Kyutai; 𝗧𝗵𝘂𝗿𝘀𝗱𝗮𝘆: LMOS by Andrej Karpathy; 𝗙𝗿𝗶𝗱𝗮𝘆: The AI Bubble; Amazon 30th Anniversary; 𝗦𝗮𝘁𝘂𝗿𝗱𝗮𝘆: Why you need AI Engineers; Follow me for more exciting insights into the world of generative AI 😎 #ai #genai #newsletter

🚀 𝗔𝗱𝗮𝗽𝘁𝗶𝘃𝗲 𝗮𝗻𝗱 𝗦𝗲𝗹𝗳-𝗥𝗲𝗳𝗹𝗲𝗰𝘁𝗶𝘃𝗲 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔𝗚) Ever wondered how to make your LLMs smarter and more reliable? Imagine a system that not only retrieves information but also corrects itself to provide accurate responses. Welcome to the world of RAG with self-correction! 🤖📚 𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀 👇 1️⃣ 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗥𝗼𝘂𝘁𝗶𝗻𝗴 𝗡𝗼𝗱𝗲: Routes questions to either document retrieval or web search based on relevance. 2️⃣ 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗲𝗿 𝗡𝗼𝗱𝗲: Transforms questions into embeddings and retrieves relevant documents from the vector store. 3️⃣ 𝗚𝗿𝗮𝗱𝗶𝗻𝗴 𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝘀 𝗡𝗼𝗱𝗲: LLM grades the retrieved documents for relevance. 4️⃣ 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗔𝗻𝘀𝘄𝗲𝗿𝘀 𝗡𝗼𝗱𝗲: LLM generates answers if the documents are deemed sufficient. 5️⃣ 𝗦𝗲𝗹𝗳-𝗖𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻 𝗠𝗲𝗰𝗵𝗮𝗻𝗶𝘀𝗺: 🔸 𝗡𝗼 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀: Accurate answers proceed to the next check. 🔸 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀: Inaccurate answers return to the Generation Node for refinement. 6️⃣ 𝗩𝗮𝗹𝗶𝗱𝗮𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗙𝗶𝗻𝗮𝗹 𝗔𝗻𝘀𝘄𝗲𝗿: 🔸 𝗬𝗲𝘀: Accurate answers are provided to the user. 🔸 𝗡𝗼: The system performs a web search for additional information. 7️⃣ 𝗪𝗲𝗯 𝗦𝗲𝗮𝗿𝗰𝗵 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻: Conducts a web search when necessary for additional data or to correct hallucinations. This framework empowers LLMs to dynamically retrieve diverse data sources, refine responses autonomously, and ensure reliability through self-correction mechanisms. 📎 Credits to medium post below:https://lnkd.in/dSucDu9Y #rag #ai #llmops #genai

𝗧𝗵𝗲 𝗥𝗲𝗮𝘀𝗼𝗻 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 𝗼𝗻 𝗬𝗼𝘂𝗿 𝗧𝗲𝗮𝗺! 👨🏿💻 OpenAI annualized revenue: $3.3B 💰 Accenture annaualize GenAI bookings: $3.6B 💼 Yet, less than 10% of customers have moved to production. Ouch! 😬 I see many companies with an impressive list of GenAI use cases ready to be implemented. That’s fantastic! They've even realized the need for external help and hired AI consultants to get the job done. Fast forward a few months, and the CEO is scratching their head, wondering why there isn’t a single use case in production... 🤔 Me: “𝐻𝑜𝑤 𝑚𝑎𝑛𝑦 𝐴𝐼 𝐸𝑛𝑔𝑖𝑛𝑒𝑒𝑟𝑠 𝑑𝑜 𝑦𝑜𝑢 ℎ𝑎𝑣𝑒 𝑜𝑛 𝑦𝑜𝑢𝑟 𝑡𝑒𝑎𝑚?” CEO: “𝐴𝐼 𝑤ℎ𝑎𝑡????” 😲 And so the story repeats itself. Lesson learned: Start with the basics—hire AI engineers—and you'll be on the road to success! 🛤️✨ Check out the video to see what I mean! 🎥😂 #AI #TechTrends #Innovation #Leadership #AIEngineers #GenAI #SuccessTips

Today it’s an special day, it’s Amazon 10.950 Day 1 anniversary! What Day 2 means? Day 2 is stasis. Followed by irrelevance. Followed by excruciating, painful decline. Followed by death. Happy 30th anniversary 🙌 #day1 #aws