Nexus Intelligence Labs’ Post

Make gen AI work: The landscape, SLMs vs. LLMs, cost and more https://zurl.co/FeTE #ai #genai #llm #slm

AI needs real-world smarts, not just data. That's why PredictHQ empowers AI with context for smarter decisions. See what our CEO, Campbell Brown has to say about moving beyond synthetic data to make demand forecasting context-aware 💡 #predictiveanalytics #predictiveai #eventeconomy

Make gen AI work: The landscape, SLMs vs. LLMs, cost and more https://zurl.co/FeTE #ai #genai #llm #slm

Not Every Hammer Needs a Nail: A Pragmatic Approach to AI in Data Strategy. This is a topic that I have been thinking about for a while. In the current age we are in, where it seems everyone is panning for that AI gold, it's easy to assume every business problem needs an AI solution. But the reality is that sometimes the simplest tools, or traditional methods, can be just as effective. In this blog I explore when to pick up the AI hammer and when to stick with simpler tools based on my recent experience. https://lnkd.in/g5bJyCDG Or don’t have time to read? We’ve got you covered, with our captivating video version of this blog on our YouTube Channel. And don't forget to subscribe! https://lnkd.in/gRgrUBki #DataTiles #Latttice #DataStrategy #AI #DataManagement #DigitalTransformation #Nocode

This is a great article on how to objectively question and approach AI projects. I love the question of “What is the cost of AI getting it wrong?” Like any project, AI projects have risk, but it’s important to weigh them up against potential benefits. Like any project, AI projects should always start with “What problem are you trying to solve?” Whether it’s an experience play - to make your teams or customers feel better about a journey or process, or a pure reduction in effort or cost based on wasted effort - understanding the problem first to define your solution should always be your first step. https://lnkd.in/gZjdT7WC

Bigger isn’t always better: How high-quality data can outsmart AI scaling laws. When it comes to AI, conventional wisdom says: “The more data, the better.” But what if you could achieve better results with less data? Turns out, the key isn’t just quantity—it’s quality. Key talking points: ⬇️ 1️⃣ Scaling laws explained: AI models improve predictably as dataset size increases, but the returns diminish fast. Example: - Reducing error from 20% → 10% takes 4x more data. - Cutting it further to 5%? 16x more data. 2️⃣ The quality-over-quantity breakthrough: Studies like Beyond Neural Scaling Laws show that a 10,000-sample curated dataset often outperforms 100,000 random samples. Why? Precision beats volume. 3️⃣ How it works: Techniques like deduplication, noise reduction, and selecting domain-relevant data improve the "signal-to-noise ratio." This means models learn more from less data, driving faster, more efficient improvements. 4️⃣ Real-world impact: With high-quality data, AI can achieve the same performance with just 20-30% of the raw data, reducing costs, improving reliability, and making models more robust. This shift from "data quantity" to "data quality" isn’t just theoretical—it’s a game-changer for building cost-effective, production-grade AI systems. The future of AI is smarter, not just bigger. P.S. If you found this perspective valuable, consider resharing for others. ... 📷 @AlphaSignal #AIInnovation #DataQuality #ScalingLaws #MachineLearning

Generative AI is transforming the way the government operates, but its success hinges on the quality of the data fueling it. Do you have a solid data foundation? Check out this article to learn more about the crucial role of data in GenAI: https://ow.ly/jstY50QBpM5 #GovernmentTransformations #NACO#NASTD #NASCIO

Just like Chat GPT can ‘make stuff up’ - with a particular favourite of mine being fake legal precedents - other applications of AI need accurate inputs. That is why I love Rampersand portfolio company PredictHQ and the work Campbell Brown and his team have been doing for years, to get insane proprietary data feeds on real events around the world, on which they’ve built their incredible prediction engine. Unique inputs + AI magic = unique and accurate predictions

For years, we’ve been at the forefront of enabling our customers to make their AI models context-aware. Adapting operations, forecasts, and planning to what is ACTUALLY predicted to impact their demand. The rise of large and small scale LLMs underscores the critical role of context in enhancing accurate and timely decision-making. Here are my thoughts on empowering smarter AI, moving beyond synthetic data and what can be done today 👉 https://lnkd.in/gySKim_d

The results are in! 85% of organizations will be using generative AI tools by year's end. Learn what industries are leading the pack and discover what it takes to stay agile, adaptive, and powerful in the shifting landscape of AI. Don't miss this incisive analysis of 500 Director to C-level data leaders' plans for the Gen AI future. 📙 #generativeAI #datastrategy #genAIresearch https://lnkd.in/gPjNFXAA