Mistral AI’s Post

First fully AI generated commit made to Breezbook. I had an hour free at the airport. I fed my entire src tree to Claude. I said "I want to return possibleCapacity and consumedCapacity for each available slot in a customer's booking calendar - so I can do surge pricing and discounting" It showed me three changed files, one of which was the test case. I pasted them, and it all worked. The code reads well. Here is the commit:

Proud to share my second machine learning project, a linear regression model. This one was a bit simpler but definitely taught me how to use lots of valuable tools within PyTorch. Check it out on my GitHub below: https://lnkd.in/ggKd37Cd

Had so much fun presenting with Amy Hodler today for Open Data Science Conference (ODSC) East. Explainability is a topic near and dear to us, especially in the age of ever-scaling AI. Definitely check out some of the resources listed in the deck, if you'd like to learn more about any of the techniques, libraries, or methods. And whether you're new to graphs or a pro, come join GraphGeeks! GH for materials: https://lnkd.in/gffRBwQq #causalinference #pywhy #dowhy #explainability #graphs #artificialintelligence

What is a Wisdom Graph? This is something I'm learning to understand and articulate better. In short, I'm interested in the "Knowledge of Knowledge", or as my good friend Ole Olesen-Bagneux called it "Search for Search". It's about understanding how information flows within organizations to achieve goals, which might be called "meta-knowledge". I decided to share my thoughts on defining a Wisdom Graph on a walk, and I'm excited to share it with you. Our upcoming O'Reilly book, "Unifying Business, Data, and Code" shows how to create very simple Wisdom Graphs using no technology, such as using a whiteboard. However, we're taking that a step further through our company, Intelligence.AI, and we're actually building our first Wisdom Graph MVP. I can't wait to share the results with you! Cheers, Ron =D

🚀 Reflecting on 6 Weeks of AI Innovation! 🧠 The AI Makerspace Engineering bootcamp has been nothing short of transformative. It's hard to believe we're already 6 weeks in and approaching the finish line! This week's deep dive into LangGraph and LangSmith was a highlight. The session not only expanded our technical horizons but the insightful questions from fellow students in our breakout sessions truly elevated the learning experience to new heights. 🌟 A massive shoutout to 👨🏫🤖 "Dr. Greg" Loughnane and Chris Alexiuk for cultivating this incredible community of accountable and driven AI Engineers. Your guidance and expertise have been invaluable. 🛠️ Excited to share a glimpse of what I built, shipped, and shared this week. Check out my work with LangGraph and LangSmith here: https://lnkd.in/ggq39ps4 🔍 Key Takeaways: - The power of graphical thinking in AI development - Reimagining state machines for modern AI challenges - Implementing AI self-evaluation for enhanced reliability This journey is just beginning, and I'm thrilled for what's next. Stay tuned for more insights and innovations! #AIEngineering #MachineLearning #LangGraph #LangSmith #ContinuousLearning #AIInnovation

Ever stumble upon an algorithm that's so mind-blowing you can't believe it isn't all over the place? Well, get ready for a wild ride with CVM, a magical formula unveiled in early 2023 that lets you count the number of unique elements in a stream using less memory than the number of elements themselves! Picture this: You're trying to figure out how many unique words there are in Shakespeare's Hamlet. But here's the catch—you've only got a whiteboard that fits 100 words. 🤯 How do you do it? Step 1: Start jotting down every unique word you see, ignoring repeats, until your board is full with 100 unique words. Step 2: Flip a coin for each word. Heads, keep it. Tails, erase it. You'll end up with about 50 words. This is round 0. Step 3: Repeat the process, but now, for every repeat word, flip a coin to decide if it stays. Heads, it stays. Tails, it's gone. Do another round of coin flips for the words on the board to decide if they stay at the end of the round. About 50 will remain. This is round 1. Step 4: Continue this pattern, making it progressively harder to keep words as you go. By round 3, you need 3 heads in a row to keep a word. Round 4, 4 heads in a row. And so on. By the end, each word's chance of sticking around is (1/2)^k. Say you finish Hamlet with 62 words left after 6 rounds. Divide 62 by (1/2)^6 and voilà—3,968 unique words. Hamlet actually has 3,967 unique words! 🎉 Incredible, right? With just a tiny memory of 100 words, you can estimate a count 40 times larger! Imagine platforms like TikTok, Twitch, and YouTube using this to count live stream viewers. The memory savings would be astronomical! It's incredible that in the 2020s, we can still discover new ways to count things. And the most mind-boggling fact is that, it takes probably like less than 50 lines of codes to implement the whole thing! This just make my day! You can check out the paper here: https://lnkd.in/gSzBSDu5 #Algorithms #DataScience #Efficiency

I love the quantitative analysis approach in this article. Many interesting insights, not only from the results but also from the research process itself. Must read.

Learn how to evaluate and score results from GPT-like systems, to increase accuracy. Register at https://lnkd.in/gn7k-Hmq Recording and GitHub material will be available to registrants who cannot attend the live, 60-min session.

Don't miss out on this amazing piece of work by one of the most active contributors of PrivateGPT community. In this great hands-on article PrivateGPT is used to compare the impact of using different: - Models - Prompt styles - System prompts - User prompts On the task of private, local AI-powered book summarization. Tons of AI learnings and tips packed in that article. The whole research can be inspected and reproduced by checking out the related github project. Links: - Article: https://lnkd.in/eGtFpEe2 - Github project: https://lnkd.in/eEi-TwVE

Just a cool thing I'd like to present you, so you can play with over the weekend: exploring word embeddings, as a way to capture word context in a document, semantic and syntactic similarity. https://lnkd.in/ev-PUAPX #machinelearning #tensor #embeddings