Charles H. Martin, PhD’s Post

I have been building production machine learning models for clients on Wall Street, Main Street, and in Silicon Valley for over 20 years. I believe that the next generation of AI models will use large scale instruction fine tuning. And to achieve this, you need to know if the fine-tuning is working. To solve this problem, I invented weightwatcher. If you want to get into the space and understand how to fine-tune models at scale, please check out these examples of what you should expect from a good model

Running the SHAP package on default settings? Then your explanations are probably wrong! I'll be giving a free preview of my summer Maven course on explainable #AI (#XAI), covering the topic I get asked about the most: Shapley values. This free course preview will be held Monday, April 8th, 2024 at 10:30 AM EDT (30 minutes) via Zoom. In this preview, I'll be giving pointers on the good, bad, and ugly of using Shapley values and the SHAP package to explain #machinelearning models. I'll be covering: - The Good: How SHAP is ideally supposed to work. - The Bad: Silly results due to correlation and high variance in feature attribution values, and waiting forever for those silly results! - The Ugly: What it takes to improve SHAP, including interventional feature perturbation, monotonicity and interaction constraints, large reference datasets, minimizing correlation, feature grouping, Owen values, and more! Learn more about the free course preview here: https://bit.ly/3TIQnba.

Ready to Unbox the Future? Google Dev Groups Build with AI - Calling All Machine Learning Mavens and Neural Network Ninjas! Get ready to #transmute your ideas into groundbreaking innovations! We're brewing up a #PowerfulPotion of knowledge at the next Google Developer Groups event in the #BuildwithAI series, and guess who's stirring the cauldron? (That's right, yours truly!) @AnuradhaTibile Mark your calendars for #May10th because we're about to unleash a potent blend of cutting-edge AI techniques and practical implementation strategies. This isn't your average developer gathering – we're diving deep into the heart of AI and crafting spells (a.k.a. code) that will #revolutionize the future. Stay tuned for more & enjoy this sneak peek. #BuildWithAI #GoogleDevGroups #AI #MachineLearning #GirlsInITC P.S. Spread the word! Let's make this event a gathering of the brightest minds in the AI realm.

🚀 Demystifying OpenAI’s New Model: #o1 🤖 The buzz around #OpenAI’s o1 model is impossible to ignore—but let’s take a moment to break it down, beyond the hype. Many people see these advanced AI models as magical, mysterious black boxes. But here’s the thing: AI, especially models like o1, aren’t magic—they’re systems designed with layers of logic and reasoning, much like how our own brains approach complex problems. Imagine you’re in a brainstorming session. 🧠 You start with basic ideas, then gradually add layers of depth, considering different angles, combining thoughts, and eventually reaching a refined conclusion. That’s how models like o1 work too. Just as a deep neural network has layers—each contributing something new and unique—the o1 model processes information in layers of reasoning. It starts with simple, foundational understanding and builds toward more complex insights. It’s like how a human would reason through a problem, one step at a time. Rather than being a mystical force, the o1 model is a tool that mimics this human-like reasoning, breaking problems down into understandable pieces and working through them logically. It’s closer to having a collaborative brainstorming partner than anything magical. Let’s embrace the real power of these models: making reasoning more accessible and scalable—not mysterious. #AI #OpenAI #DeepLearning #NeuralNetworks #AIReasoning #Innovation #Brainstorming #FutureTech #LogicNotMagic #LetsBreakItDownLogically

Really looking forward to this! Check it out and sign up if you're interested in explainable AI (and if you're in or around AI and ML you should be)!

🏆 Great blogpost from HF explaining Vision-Language Models! Very much recommended, even if we are working with these models on a day to day basis - its alway good to take a step back and see the field as a whole In a nutshell; these models combine vision and language to for complex tasks like image captioning and visual question answering. Since 2021, interest in these models has surged, with examples like OpenAI’s CLIP. They excel in zero-shot generalization, making them very versatile Vision-language models use strategies like contrastive learning and multi-modal fusing to align images and text. They are pre-trained on large datasets and fine-tuned for specific tasks. Thanks to HF, they can be used really easily with the Transformers library. Kudos! 🔗 Read the blogpost: https://lnkd.in/evC-QWkb ⤵ Helpful? Follow me and join ⚡️ AI Pulse (https://lnkd.in/eWudwDsd) for daily, curated, bite-sized updates on AI—focused on what truly matters to keep you ahead of the curve 🔥

🎥 First Step Towards Machine Learning 🚀 Thrilled to share my journey into the world of machine learning with this hands-on project! 🌟 I explored how we can train models to recognize patterns and make predictions, all without diving into complex code. This project is just the beginning, and I’m excited to explore more advanced concepts in ML and AI. Your feedback is invaluable, so let me know what you think! #MachineLearning #DeepLearning #AI #LearningJourney #Projects

Workers AI now supports fine-tuned models using LoRAs. But what is a LoRA and how does it work? In this post, we dive into fine-tuning, LoRAs and even some math to share the details of how it all works under the hood. https://cfl.re/4ajt39H #DeveloperWeek

Just when you thought AI Advancement would be slowing down... OpenAI is releasing a new series of models: o1 The benchmarks are absolutely mind-blowing. I literally talked yesterday about how AI hasn't surpassed humans yet in competition level maths, and now this. Just amazing. What's particularly interesting is the new approach they've taken. Instead of just relying on a larger training dataset, the model "thinks" before it answers, refining its responses autonomously in the background. Full article: https://lnkd.in/eWKiry2f

I had the opportunity to build this feature with Cloudflare's incredible Workers AI team. It is a bit surreal to be able to ship something like this on a platform with over 2 million developers. Thankfully I get to work with some of the smartest people I've ever met. But what even is this feature? You can now bring your own LLM fine-tunes to Cloudflare trained on your own data. This means you can customize the models we host and adapt them to your usecase. It also means that you can achieve beyond GPT4 performance on specific tasks. We're just getting started with fine-tuning and many more task types are coming soon. Read more below about how we built this feature!