Raphaël MANSUY

🚨 Anthropic has unveiled Claude 3.5 Sonnet, its latest advanced large-language model. A massive upgrade that you shouldn't miss 🔥 𝗣𝗹𝘂𝘀: It boasts 2x the speed of its predecessor at a fraction of the cost🤯 𝗪𝗵𝗮𝘁 𝘆𝗼𝘂 𝘀𝗵𝗼𝘂𝗹𝗱 𝗸𝗻𝗼𝘄: • Anthropic introduces 'Artifacts'’ enabling real-time interaction with Claude in a side panel workspace. • With this upgrade, Sonnet exceeds competitor models in key metrics, delivering twice the speed of Claude 3 Opus at just one-fifth of the cost. • It can be accessed for free on Claude and the Claude iOS app This upgrade surpasses competitors like GPT-4o across benchmarks for reasoning, code, math, and knowledge abilities and has also passed the ‘AI Mirror Test.’ 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀💥 This presents a major leap in AI capabilities, surpassing competitors like GPT-4o, Gemini Pro, and Llama. 𝗜𝘀 𝗦𝗽𝗲𝗲𝗱 𝘁𝗵𝗲 𝗢𝗻𝗹𝘆 𝗥𝗮𝗰𝗲? AI firms mimic Apple's marketing but face a tough reality. Frequent updates like Sonnet and Haiku can confuse users. This rapid release cycle feels rushed and may shift focus away from developing user-friendly AI solutions. What do you think? 🔔 Follow me for more updates.

4

2 Comments

Anthropic became the first AI frontier model company to release a public beta of “computer use” that allows developers to direct Claude (via an API) to use computers like humans by viewing a screen, moving a cursor, clicking buttons, filling out forms and typing text. Google, OpenAI, Microsoft and others are known to be working on—and likely nearly beta releases of—similar technology. This could have profound implications to knowledge workers and consumers, but it is very, very early. Historical context is important to understand so you don’t get misled by the hype. Leading AI research labs have been pursuing AI agents that could see, understand and take action on our computers for years. Andrej Karpathy and OpenAI published “World of Bits: An Open-Domain Platform for Web-Based Agents” back in 2017, but they ran into technical roadblocks. The invention and advancements of large language models (LLMs) changed everything though, and rekindled interest in the idea. While I think these sort of AI agents with computer use will eventually be disruptive, there are enormous obstacles ahead to making the technology reliable, safe and precise. Plus, users (and corporations) will be largely resistant to them due to privacy and security concerns (e.g. having to give them access to your credit card information and login credentials). So, this is definitely tech you should follow, but don’t expect it to change your personal or professional lives anytime soon. Mike and I discuss computer use, and a ton of AI model releases and updates, in the opening of Episode 121. 00:05:04 — AI Model Releases and Rumors: New Claude Model + Computer Use, Claude Analysis Tool, OpenAI Doubles Down on AI for Code, Perplexity Pro Reasoning Update, Runway Act-One, Eleven Labs Voice Design, Stable Diffusion 3.5, The Rumors  00:27:07 — The Dark Side of AI Companions + Kid Safe GPT for Parents 00:39:29 — Ex-OpenAI Researcher Sounds Alarm on AGI Preparedness 00:47:57 — AI + National Security 00:53:14 — Microsoft vs. Salesforce Over Agents 00:57:08 — Disney AI Initiative  01:00:17 — Apple Intelligence Photos 01:03:03 — Google Open Sourcing SynthID 01:06:32 — OpenAI + Fair Use 01:10:43 — Using Gemini to Prep for Public Speaking As always, you can listen on your favorite podcast network, or watch it on YouTube. https://lnkd.in/ggFS9D_D

90

18 Comments

If you want to understand why LLMs are not the solution for AGI (Artificial General Intelligence), I highly recommend listening to Lex Fridman podcast with Yann LeCun : Meta AI, Open Source, Limits of LLMs, AGI & the Future of AI T͟L͟D͟R͟: 1. LLMs do not understand the physical world 2. LLMs have not the ability to remember 3. LLMs have no persistent memory 4. LLMs are not able to reason and to plan

27

5 Comments

🌟 𝐄𝐱𝐩𝐥𝐨𝐫𝐢𝐧𝐠 𝐭𝐡𝐞 𝐅𝐮𝐬𝐢𝐨𝐧 𝐨𝐟 𝐆𝐫𝐚𝐩𝐡 𝐓𝐡𝐞𝐨𝐫𝐲 𝐚𝐧𝐝 𝐃𝐞𝐞𝐩 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 🌟 During my first year of my master’s program (M1), I dove into several fascinating topics, with a strong focus on graph theory. This exploration led me to a powerful intersection: 𝐆𝐫𝐚𝐩𝐡 𝐍𝐞𝐮𝐫𝐚𝐥 𝐍𝐞𝐭𝐰𝐨𝐫𝐤𝐬 (𝐆𝐍𝐍𝐬). Inspired by this, I wrote a Medium article titled "𝗚𝗿𝗮𝗽𝗵 𝗡𝗲𝘂𝗿𝗮𝗹 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀: 𝗧𝗵𝗲 𝗕𝗿𝗮𝗶𝗻𝗰𝗵𝗶𝗹𝗱 𝗼𝗳 𝗚𝗿𝗮𝗽𝗵 𝗧𝗵𝗲𝗼𝗿𝘆 𝗮𝗻𝗱 𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴",  where I explain how GNNs work and their underlying principles. 📌 You can find the article here: 𝘩𝘵𝘵𝘱𝘴://𝘮𝘦𝘥𝘪𝘶𝘮.𝘤𝘰𝘮/@𝘢𝘣𝘥𝘦𝘭𝘩𝘦𝘥𝘪.𝘺𝘰𝘶𝘴𝘴𝘦𝘧8/𝘨𝘳𝘢𝘱𝘩-𝘯𝘦𝘶𝘳𝘢𝘭-𝘯𝘦𝘵𝘸𝘰𝘳𝘬𝘴-𝘵𝘩𝘦-𝘣𝘳𝘢𝘪𝘯𝘤𝘩𝘪𝘭𝘥-𝘰𝘧-𝘨𝘳𝘢𝘱𝘩-𝘵𝘩𝘦𝘰𝘳𝘺-𝘢𝘯𝘥-𝘥𝘦𝘦𝘱-𝘭𝘦𝘢𝘳𝘯𝘪𝘯𝘨-3𝘤83𝘥6𝘢233𝘥7 Feel free to give it a read and share your thoughts ! #𝙂𝙧𝙖𝙥𝙝𝙏𝙝𝙚𝙤𝙧𝙮 #𝘿𝙚𝙚𝙥𝙇𝙚𝙖𝙧𝙣𝙞𝙣𝙜 #𝙂𝙧𝙖𝙥𝙝𝙉𝙚𝙪𝙧𝙖𝙡𝙉𝙚𝙩𝙬𝙤𝙧𝙠𝙨 #𝘼𝙄 #𝙈𝙖𝙘𝙝𝙞𝙣𝙚𝙇𝙚𝙖𝙧𝙣𝙞𝙣𝙜

23

How fast is fast when it comes to commercial open source software traction? Here are some stats to serve as benchmarks that Matthieu Lavergne put together among many other great data and insights in Serena's first report on commercial open source software. It's available for free here: https://lnkd.in/d5uBvjAU

18

3 Comments

Sometimes I'm asked whether agents will shape the future... And I give an example, #ChatGPT is already an agent-driven workflow. Ask him ! ChatGPT from #OpenAI integrates tools (like Python Environment), other AI models (like DALL-E), memory, and multi-step tasks, likely using GPT to manage different system messages that assume various "personas.", in other words multi-simple-agent flows. It's important to recognize that agents and agent-driven workflows are already a part of our daily lives. Therefore, I recommend that enterprises explore and stay informed about the research trends concerning agent-driven workflows and multi-agent systems. To build a community for sharing information about agent-driven workflows and multi-agent systems, myself and other enthusiasts are launching AIgentic! 👥 Spread the Word! Help us reach more innovators like you. Share our journey, and let's unlock Agent's full potential together! 🌟 Elevate Your AI Agents Insights with AIgentic 🌟 #ArtificialIntelligence #Agents #GenerativeAI

11

2 Comments

Presenting Orpomis Prime 7B model series ! These models are a result of trying different merge strategies of the following models: - mistral-orpo-beta from KAIST AI - Hermes-2-Pro-Mistral-7B from NousResearch In my experiments, dare ties merge leads to a significantly better-performing model on the openllm benchmarks In general, the model performance is quite solid in text generation as well as math reasoning. Full result comparison: https://lnkd.in/gb92DXwG Dare ties Model: https://lnkd.in/gMS-H4Sq Model stock Model: https://lnkd.in/giwEJ7Gs

9

💥 New day, new model 💥 We ran our performance benchmarks on Anthropic Claude 3.5 Sonnet v/s OpenAI's GPT-4o & the results weren't what you'd expect ⏬ Anthropic says Claude 3.5 Sonnet outperforms all competitor models on key evaluations, at twice the speed of Claude 3 Opus and one-fifth the cost. In our benchmarks 3.5 Sonnet & 4o came out neck & neck. Ultimately, you need to run evaluations on your task at hand and measure the quality yourself. 🔍 Data Extraction: Both models identified 60-80% of data correctly, but neither excelled in this task. 📊 Classification: Claude 3.5 Sonnet (72%) outperformed GPT-4o (65%) in mean accuracy. However, GPT-4o led in precision (86.21%), crucial for accurately classifying customer tickets. 🧠 Verbal Reasoning: GPT-4o outperformed Claude 3.5 Sonnet with 69% accuracy, versus 44%. Both models did well on analogy and relationship questions. 📜 Context window: Claude 3.5 Sonnet boasts a larger context window (200K) compared to GPT-4o's 128K 🌪 Latency: Claude 3.5 Sonnet is 2x faster than Claude 3 Opus, but it’s still lags behind GPT-4o when it comes to latency (50% faster Time to First Token) Price and throughput for these models was very similar Want to learn more about how these models stack up? Check out our full analysis on the Vellum blog: https://lnkd.in/dwEcQJe4 Oh and did I say, Claude 3.5 Sonnet is already in Vellum? Sign up for a demo to run your evals & start building!

68

1 Comment

GPT-4o mini is out! Cost-sensitive, multi-language, small and as fast as GPT-3.5 Turbo, it is in fact, the natural transition from this model improving it in all benchmarks (82% on MMLU vs. 69.8%). - 128k context window - 16k output tokens per request

7

2 Comments

🙏 Thank you Daniel Whitenack and Chris Benson for hosting Guillaume Lemaitre and me on your excellent Practical AI Podcast 🎙️📡 ✨ Episode #296 - scikit-learn & data science you own ✨ 👉 https://lnkd.in/ehqkd8ha 📓 Chapters: - Welcome to Practical AI Podcast - What is :probabl.? - Stewarding #OpenSource projects - What is scikit-learn? - The #DataScience landscape - Scikit’s role with general purpose models - Further development - :probabl.'s open source relationship - Fun & interesting use-cases - Getting started for new devs - Future of scikit-learn & :probabl. - Outro

38

2 Comments

VLDB 2023 best paper award. Very interesting paper on PolarDB. My takeaways : - creative use of causal graphs and game theory for root cause and anomaly detection; - in-database AI inference framework; - in-database data security that implements opteron and trusted execution environments (TEE) and Behaviour Control Lists (BCL) as opposed to ACLs; - cloud resource composition; - notable use of RDMA messaging; - single master-multiple replicas, shared-storage architecture (mysql fork).

1

📢📢📢 New Blog Post : Writing a Chrome Extension that summarizes any selected text from a web page, using the Gemini Nano model in Chrome Built-in AI. This is the 4th part in the series that explores Chrome Built-in AI. Read it here: https://bit.ly/3CqVO8g Github Repository: https://bit.ly/4ecBYuM The series so far: 🗂️ Part 1 : Getting started with Chrome Built-in AI: https://bit.ly/3YcMsUO 🗂️ Part 2 : Adding AI features to existing Web Applications : https://bit.ly/3NTXfyF 🗂️ Part 3 : Exploring the Summarization API : https://bit.ly/3Yu8TFm #Google #Chrome #GoogleChrome #GoogleAI #ChromeAI #GeminiNano #Gemini #Summarization #GenerativeAI

21

2 Comments

🤖 MetaGPT Redefines Multi-Agent Collaboration 🤖 📄 arXiv 𝗣𝗮𝗽𝗲𝗿: "MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework" by Sirui Hong et al. During my talks, people usually ask me where I expect to see a boom in the incoming months. If Sam doesn't find AGI in the short term 😂, my answer is clear: Agents, Agents everywhere! So let's take a look to this paper about Agent Collaboration. 📝 𝗔𝗯𝘀𝘁𝗿𝗮𝗰𝘁 𝗦𝘂𝗺𝗺𝗮𝗿𝘆 Take a look to MetaGPT, a game-changer in multi-agent collaboration. By leveraging Meta Programming techniques, MetaGPT optimizes collaboration efficiency through Agent Roles, Workflows, Structured Messages, and Efficient Communication Protocols. Its outstanding performance on benchmarks like HumanEval and MBPP sets a new standard, particularly excelling in software development tasks. 𝗧𝗟;𝗗𝗥 🧠 MetaGPT introduces cutting-edge Meta Programming techniques for enhanced multi-agent collaboration. 📊 Achieves unparalleled performance on industry benchmarks like HumanEval and MBPP. 👨‍💻 Excels in software development tasks with high task completion rates. 🤝 Incorporates structured messages and efficient communication protocols for streamlined collaboration. 👾 Ablation study underscores the significance of diverse agent roles in boosting performance. 𝗙𝗼𝗿 𝗺𝗼𝗿𝗲 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 👇 📜 ArXiv Paper: https://lnkd.in/dt7XuM3Z 📪 Andrew Ng post on Multi-Agent Framework: https://lnkd.in/d5pmYcpB 👩‍💻 MetaGPT Github repository: https://lnkd.in/dQa7T4wM 👏 Kudos to Val Andrei Fajardo for the amazing visualization of the pager, it's pure magic for the eyes!

53

4 Comments

🎉 𝐄𝐱𝐜𝐢𝐭𝐢𝐧𝐠 𝐍𝐞𝐰𝐬 𝐟𝐨𝐫 𝐀𝐈 𝐄𝐧𝐭𝐡𝐮𝐬𝐢𝐚𝐬𝐭𝐬! 🌐 Dive into the world of Artificial Intelligence with complete Hands-On Course on AI Ecosystem: "AI Ecosystem for the Absolute Beginners – Hands-On"! 𝐀𝐰𝐞𝐬𝐨𝐦𝐞 𝐍𝐞𝐰𝐬: We are very pleased to share that we have published our First ever, the most detailed and up-to-date AI Ecosystem course for everyone that includes Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), Generative AI (GenAI) and Prompt Engineering - which is now available at Udemy 🚀 Dive Into AI world starting with the basics of AI, the course guides you through the lifecycle of AI applications and essential concepts. It also covers the basics of Machine Learning, Deep Learning, and Generative AI, supported by practical hands-on demos using Amazon Web Services (AWS) services such as Rekognition, Forecast, Comprehend, and Bedrock. Additionally, it also explores prompt engineering, featuring a comprehensive hands-on demonstration with OpenAI ChatGPT, concluding with a capstone project focused on Generative AI. 𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫 𝐭𝐨𝐝𝐚𝐲 𝐰𝐢𝐭𝐡 𝐛𝐞𝐥𝐨𝐰 𝐜𝐨𝐮𝐩𝐨𝐧 𝐜𝐨𝐝𝐞 𝐭𝐨 𝐞𝐧𝐫𝐨𝐥𝐥 𝐟𝐨𝐫 𝐅𝐫𝐞𝐞! 𝐂𝐨𝐮𝐩𝐨𝐧 𝐕𝐚𝐢𝐥𝐝 𝐟𝐨𝐫 𝟑 𝐝𝐚𝐲𝐬, 𝐇𝐮𝐫𝐫𝐲! 𝐔𝐝𝐞𝐦𝐲 𝐂𝐨𝐮𝐩𝐨𝐧 𝐋𝐢𝐧𝐤: https://lnkd.in/giwpu89X 📚 What sets this course apart? The "AI Ecosystem for Absolute Beginners" course is designed to accommodate learners of all levels, from beginners to experienced professionals looking to explore the world of Artificial Intelligence. Acting as a foundational pillar, this course facilitates your journey into the AI ecosystem. Don’t worry if you are new to AI, we are going to get started with the absolute basics. Each lecture in this course is accompanied by real hands-on demos. 🛠️ Course Features: 🎥 In-Depth Video Tutorials: Learn complex concepts through high-quality and engaging videos ✅ Quizzes: Test your knowledge and reinforce learning at every step 📚Reference Study Material: Explore the course with curated resources for deeper understanding 🖥️Live Hands-On Demos: Witness the Live demos, setup your own learning environment and learn by doing yourself! Instructors and Credits: Yogesh Raheja Thinknyx Technologies Dheeraj Sain Deepthi Narayan Shruti Bhardwaj Naman Sharma Kumar Kanishk #thinknyx #AI #ArtificialIntelligence #MachineLearning #DeepLearning #GenAI #GenerativeAI #PromptEngineering #ChatGPT #AWS

51

2 Comments

👉🏼 Performance Assessment of GPT 4.0 on the Japanese Medical Licensing Examination 🤓 Hong-Lin Wang 👇🏻 https://lnkd.in/e3b4kjU8 🔍 Focus on data insights: - 📈 GPT 4.0 achieved a correct response rate of 80.4% on the Japanese Medical Licensing Examination, indicating strong performance in comprehension and analysis. - 🔍 Analysis revealed significant differences in accuracy based on question difficulty and type, highlighting areas for targeted improvement. - 🏆 The Global Quality Scale

1

AI Hallucinations in Large Language Models : What They Are and How to Prevent Them Ever wondered what happens when AI goes a bit off-script? Imagine asking your digital assistant for a recipe and it suggests cooking instructions for a "banana lasagna" that's never existed! How to Prevent AI Hallucinations ? 1) Quality Data Inputs: Feeding AI accurate and diverse data reduces the chances of it inventing things that don't make sense. 2) Contextual Understanding: Teaching AI to grasp context helps it give responses that fit the situation, like understanding that lasagna typically involves pasta, not bananas. 3) Regular Checks and Balances: Monitoring AI's outputs and correcting mistakes promptly ensures it stays reliable and doesn't go off on culinary adventures. Following these steps helps keep AI grounded in reality, avoiding quirky scenarios like banana lasagna recipes 😀 There is very important parameter that determines how creative an LLM model can be to curate answer for your prompt called ‘temperature’ I will explain in next post.

1

3 Comments

November 6 2024 – 5 Notable Articles To Read 𝗔𝗿𝘁𝗶𝗰𝗹𝗲: Microsoft's $30B Infrastructure Fund Is Driving Vertical Integration Of AI 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: Microsoft, BlackRock, Global Infrastructure Partners (GIP), and MGX have launched the Global AI Infrastructure Investment Partnership, an ambitious new initiative seeking to raise $30B in equity capital for AI infrastructure investments. GAIIP will prioritize funding for new and expanded data centers, along with the energy infrastructure needed to power these facilities, with a primary focus on the United States. 𝗦𝗼𝘂𝗿𝗰𝗲: Forbes 𝗟𝗶𝗻𝗸: https://lnkd.in/d79SyYdD 𝗔𝗿𝘁𝗶𝗰𝗹𝗲: How AI Is Incrementally Fueling Energy Sector Innovation 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: Power generation companies have increasingly adopted AI for their next-generation systems and, in many ways, were early pioneers of AI within the energy sector. This adoption is fueled by a growing demand for enhanced throughput, operational efficiency, and sustainability. 𝗦𝗼𝘂𝗿𝗰𝗲: Forbes 𝗟𝗶𝗻𝗸: https://lnkd.in/d8Q2nJER 𝗔𝗿𝘁𝗶𝗰𝗹𝗲: Jo N.’s Invidia Capital Management seeks $850M for first fund 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: Currently raising capital for its inaugural fund, aiming to invest in upper middle-market healthcare buyouts. Invidia was launched earlier this year with a focus on healthcare investments ranging from $100M to $300M per deal. Plans to make both minority and majority investments across healthcare sectors - including pharmaceutical outsourcing and contract manufacturing, medical devices, and healthcare information technology. 𝗦𝗼𝘂𝗿𝗰𝗲: Fortune 𝗟𝗶𝗻𝗸: https://lnkd.in/dB4UMSaW 𝗔𝗿𝘁𝗶𝗰𝗹𝗲: Saudi Arabia Nears Deal to Take Over $4B Olam Unit 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: Olam has confirmed receiving an offer from Saudi Arabia's wealth fund, SALIC, for its agriculture unit, a deal that could enhance food security in the kingdom. 𝗦𝗼𝘂𝗿𝗰𝗲: Bloomberg 𝗟𝗶𝗻𝗸: https://lnkd.in/d9AJUs74 𝗔𝗿𝘁𝗶𝗰𝗹𝗲: AI demand drives growth and higher spending at Amazon’s cloud business 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: As businesses rush to develop and deploy generative AI tools, their demand for additional cloud storage capacity to hold data is continually increasing. 𝗦𝗼𝘂𝗿𝗰𝗲: Financial Times 𝗟𝗶𝗻𝗸: https://lnkd.in/dyq6kdK8 #news #business #ai #research #investments

6