🎄 Day 2 of OpenAI's Christmas: Reinforcement Fine-Tuning Gets a Major Upgrade Just unwrapped OpenAI's second gift: Advanced reinforcement fine-tuning for O1 models. After spending the morning testing it, here's my deep dive into what this means for AI customization. 🎯 The Big Promise: Train models on reasoning patterns, not just data Claims you only need 12 examples to see results "Easy-to-use" interface for complex tuning 🧪 Reality Check: Having extensively tested similar tools, I can confirm these aren't empty promises. The "12 examples" claim holds up – you can achieve remarkable specialization with surprisingly small datasets. This mirrors my experience with earlier iterations of fine-tuning tech. 💰 The Cost Equation: Highly efficient for small, focused datasets Gets expensive quickly at scale Cost-benefit ratio skews unfavorable for large training sets 🔍 Key Learning From My Testing: Here's the secret sauce I've discovered: Don't try to create a "jack of all trades" model. The magic happens when you: Fine-tune for narrow, specific tasks Keep your general-purpose model separate Use them in tandem: Specialized model for specific tasks, general model for communication Think of it like having a highly specialized expert (fine-tuned model) consulting with a skilled communicator (general model) to deliver the perfect response. 🎓 Pro Tip: For those watching costs, I've found fine-tuning smaller models can often deliver better ROI for specific use cases. The key is being strategic about what you're trying to achieve. 🤔 Strategic Implementation: Rather than asking "Can we fine-tune this model?" start with "Should we fine-tune this model?" The best results I've seen come from clear, narrow use cases where the standard model consistently misses the mark. Question for my network: What specific tasks would you want to fine-tune a model for? Let's discuss use cases where this could be game-changing. Stay tuned for Day 3! 🎁 #AI #OpenAI #MachineLearning #ArtificialIntelligence #FineTuning #AIInnovation #TechNews
Scott Farrell’s Post
More Relevant Posts
-
THERE ARE THREE R'S IN STRABERRY 🍓🍓🍓 Looking through OpenAI's new o1 launch today, it got me thinking about Daniel Kahneman's book, Thinking, Fast and Slow 📚. In it, Kahneman explains that our brains operate using two systems of thought. System 1 is fast, automatic, and intuitive—it handles everyday decisions without us even realizing it 🤔. System 2, on the other hand, is slow, deliberate, and analytical 🧠. It's what we engage when we're solving complex problems or learning something new. This dual-process theory highlights how we balance efficiency and thoroughness in how we think. By the way, did you catch the typo in "STRABERRY"? If not, that's System 1 at work—glossing over details to get the gist quickly. It takes System 2 to slow down and spot the error 🔎 OpenAI's o1 feels like a significant step towards embracing that System 2 thinking — more deliberate and analytical, targeting high-complexity (or high-risk) scenarios. Meanwhile, the current GPT models are akin to System 1: fast, intuitive, but at times superficial or even inaccurate. Both systems are necessary—in our brains and in our quest towards AGI. In future, it's not hard to imagine that you won't have to pick the right model to use - based on the prompt the AI will pick the right system to use to optimize the energy and time - just like our brain does. Exciting times! 🚀🌐 P.S. – The title (except the typo) and the screenshot come from the o1 launch examples, which I thought was very witty of them given all these AI strawberry memes. #OpenAI #o1 #ThinkingFastAndSlow #AI #AGI #Innovation
-
-
6x Founder. CEO @ Mindstone, a platform teaching non-technical people how to use AI in their daily jobs, to be more productive. Ex. Co-Founder/CPTO @ SuperAwesome (Acquired by Epic Games)
Yesterday, Anthropic released its Data Analysis tool, and it's another big step in the right direction: a better, cleaner version of what OpenAI released in the past. In this video, I upload 109 typeform responses from our recent AI meetups worldwide, asking Claude to analyse the data and visualise it in a way that best communicates what's going on. It proceeds to - Read the file - Understand what it's about - Pick the most interesting data to talk about - Chooses the most compelling way to visualise it - Builds a dashboard to do so Doing this used to take hours, but now it takes seconds. Tell me again how Generative AI is just hype? :) P.S. I love how Anthropic is starting to clearly show they are the product-led version of OpenAI's research-based approach. When they release something, it's polished, works, and is beautiful, as compared to OpenAI's cutting-edge but often rough-around-the-edges approach. #PracticalAI #FutureOfWork
-
🎉 Big news: Quickbase is now connected to OpenAI! Which means your workflows just got way smarter. What does that mean for you? It means you can: ✅ Use AI to analyze data, summarize reports, or even draft content for you. ✅ Automate tasks that used to take hours. ✅ Get insights and recommendations faster than ever before. Imagine submitting a task and having AI instantly suggest improvements or pull out the key details. Game-changer, right? The best part? It’s seamless, intuitive, and designed to make your day easier. 💡 Ready to see the magic in action? Let’s talk about how you can start using this today! #WorkSmarterNotHarder #Quickbase #AI #OpenAI
-
Creating Stories out of Data | Helping Business with Best practice BI solutions | Upwork Top Rated Plus | Data Science | Data Analytics Expert | Microsoft Power BI Certified | Microsoft Data Engineer Analytics Solutions
🌟 Exciting News: The Future of AI with OpenAI O1! 🌟 🚀 OpenAI has just introduced the OpenAI O1 preview, a huge leap in building more efficient, scalable, and accessible AI systems. 🎯 With its innovative framework, O1 is designed to revolutionize how we approach AI development, making it more adaptable to complex tasks and scalable across industries. 🔑 Key Features of OpenAI O1: Faster performance, increasing productivity and efficiency 🔄 Enhanced fine-tuning capabilities for more specific, real-world applications 🎯 Advanced API integration, streamlining AI workflows 🤖 The O1 preview gives developers and organizations a sneak peek into the next generation of AI—one that's more reliable, scalable, and versatile than ever before. 🚀 Whether you're in data science, machine learning, or a business exploring AI-driven solutions, this new platform will be a game-changer for future innovations. 🌐 If you're curious to learn more, check out the full details here: OpenAI O1 Preview! #AI #ArtificialIntelligence #MachineLearning #OpenAI #TechInnovation #ScalableAI #AIIntegration
-
Continuing the magic of AI in Google Sheets!
Todays's AI in Spreadsheets: Dalle! Image generation is still magic. Try it out in the linked GSheet. No need to install our add-on, get an openai account or even to copy the spreadsheet. It's all just magic (you can do all of those things if you want to dive deeper of course) https://lnkd.in/e9_t7baz
-
-
🚀 Ready to Unlock the Power of OpenAI? 🌟 Dive into our comprehensive guide on getting started with the OpenAI API! Whether you're a beginner or looking to sharpen your AI skills, this video has everything you need: 🔑 How to Obtain Your API Key 📊 Monitoring Your Usage 🤖 Creating & Using AI Assistants From signing up, generating your API key, to setting up AI assistants that supercharge your productivity—it's all here! Perfect for tech enthusiasts and business owners alike. 👨🏫 Learn from expert tips and practical examples that will help you integrate AI effortlessly into your projects. Don’t forget to subscribe and hit the notification bell to stay updated with the latest in AI tech. Join our community and let's innovate together! Watch Now: [https://lnkd.in/d_KGauyY)
-
🚀 Ready to Unlock the Power of OpenAI? 🌟 Dive into our comprehensive guide on getting started with the OpenAI API! Whether you're a beginner or looking to sharpen your AI skills, this video has everything you need: 🔑 How to Obtain Your API Key 📊 Monitoring Your Usage 🤖 Creating & Using AI Assistants From signing up, generating your API key, to setting up AI assistants that supercharge your productivity—it's all here! Perfect for tech enthusiasts and business owners alike. 👨🏫 Learn from expert tips and practical examples that will help you integrate AI effortlessly into your projects. Don’t forget to subscribe and hit the notification bell to stay updated with the latest in AI tech. Join our community and let's innovate together! Watch Now: [https://lnkd.in/gE9uDMwR)
-
So far this week, I've saved 44hrs 26min using the OpenAI API. Context: Working with a home services client to launch a massive direct mail campaign. To fuel this campaign, we built a database of 40k target addresses and homeowners. The problem? Homeowner names were not normalized. 🔵 "Will McCartney" 🔴 "McCartney Will" 🔴 "McCartney Will W" 🔴 "Will W McCartney" Using the BlueAcquire GPT toolkit, a simple prompt fixed this. The prompt: "Please normalize the following text into the format 'firstName_lastName'. Only return the normalized text. Do not include middle names or initials. In the event there are two names please return only one name. Keep in mind, the names might be out of order, so you may need to reorder them. Here is the text:" Normalizing this data manually would have taken 44hrs 26min. (40k records x 4 seconds per record) Using AI, we'll be ready to launch the campaign tomorrow. Comment "send" if you want the free toolkit. #gotomarket #ai #data
-
-
Today's AI deep dive: Global future scenario modeling with OpenAI's O1-preview TL;DR - check out https://lnkd.in/evgHTNVW to see 6 future scenario prompts, the AI "thought process" and the outputs as well as its evaluation of which future scenario is most likely! My own thoughts: - The guardrails that OpenAI has built in constrain the scenario modeling process, with a bias towards optimistic outcomes - Even so, using this as an interactive tool for deep diving into scenarios seems very powerful; this type of work would have previously taken consulting teams days to create - I look forward to the addition of file search to o1 - it will be very interesting to run shorter-term scenario modeling processes using annual reports and internal documents for specific companies - These scenarios make for some interesting/fun reading. I asked it to then write a story in the voice similar to Neal Stephenson and it basically created an abridged version of the Three Body Problem instead. Which goes to show that when it comes to true creativity, AI is not there. It just rehashes what it knows. That said, there's a LOT to be mined from what we already know and a lot of value in doing so for many organizations. #AIDeepDive #ScenarioModeling #OpenAIO1Preview #FutureTrends #Innovation #TechInsights #MachineLearning #AIApplications #Management #TechThoughts #o1
-
-
This is a rather clever prompting technique, essentially selecting the most relevant examples for the given prompt.
I help researchers and builders make sense of AI | ex-Stripe | aitidbits.ai | Angel Investor
Few-shot prompting is great, but dynamic few-shot prompting? Even better. Franklin Lindemberg's latest prompting technique leverages OpenAI embedding to optimize few-shot learning by using a dynamic few-shot prompt technique. Instead of bombarding the model with too many examples, this method dynamically selects only the most relevant ones, ensuring efficiency without sacrificing quality. How it works: (1) A vector store holds a library of input-output examples (2) An Embedding model transforms user input into vectors for querying the store (3) An LLM completes the tasks based on the most relevant examples. Following this approach, only the top 3 examples are pulled, leading to faster and cheaper generation thanks to fewer tokens being used. Also, selecting the most relevant examples leads to better performance. This approach refines the few-shot technique into a more scalable and cost-effective method for use cases like displaying data in tables, classifying text, or summarizing documents. Full blog and code examples post https://lnkd.in/gwGr2DTa — Join thousands of world-class researchers and engineers from Google, Stanford, OpenAI, and Meta staying ahead on AI http://aitidbits.ai
-