Metagrove Ventures

Check this graph out: did you know that back in the 1950s, machine learning models were trained on only ~40 parameters?! Today, AI models are closing in on 1 trillion parameters!!! 🤯 It's insane to see the exponential growth in these numbers, but also the explosion of AI models that have come since the 2000s. But there's a big caveat here...a model with more parameters doesn't necessarily mean it's smarter. There's a risk of overtraining and relying too much on memory, but that's a topic for another day. It also requires much more computing power which is becoming a contentious topic as we look for more sustainable and environmentally friendly ways to power AI chips. In any case, we're definitely seeing a 'hockey-stick' effect taking place. At this point, I think Moore's Law can be thrown out the window. Either the curve is going to get a lot steeper or something has to give and we're going to have to rethink about how to really scale these models. I'm going for the latter.

🎙️ More companies are now trying to compete against NVIDIA's dominance. This new AI company is doing just that... Peter Buckingham is a Senior VP of Software Engineering at d-Matrix - their aim is to complement AI chip giants such as Nvidia, whose chips are used to train AI systems on huge amounts of data. This was our second conversation. In this chat we: ✴ explore the competitive landscape of AI chips, focusing on the key differences between training and inference. ✴ Discover how companies like D Matrix are challenging NVIDIA's dominance in the data center inference market by offering cost-effective and performance-driven solutions. ✴ Discuss the evolution of AI hardware, the challenges of Moore's Law, and the innovations driving the next generation of AI chips, such as in-memory computing. ✴ Gain insights into the software ecosystem, the role of frameworks like PyTorch, and the future of AI infrastructure. ✴ Plus, we touch on the impacts of AGI, data concerns in AI model training, and the importance of building high-performing teams in the rapidly evolving AI landscape. Available now on: 🔥 Spotify: https://lnkd.in/ghWvTQB6 🔥 Apple: https://lnkd.in/gJxBfwhG 🔥 YouTube: https://lnkd.in/g2r52hCZ

If you follow my linkedin, you'll know I spend a lot of my time onsite with customers. This is our (not so) secret sauce, and this is why:⁣ ⁣ 𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐑𝐞𝐚𝐥 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬: Having been on the other side using various construction tech products, I remember the frustration of tools that didn’t fit our needs and lacked true customer-centricity. By being on-site now, we gain real insight into the day-to-day challenges our clients face, helping us build a product that truly meets their needs.⁣ ⁣ 𝐒𝐨𝐥𝐯𝐢𝐧𝐠 𝐭𝐡𝐞 𝐀𝐝𝐨𝐩𝐭𝐢𝐨𝐧 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞: When I was working as a civil engineer, I could feel that the people who built the tools that I used didn't have experience as a user. Even with over a decade in civil construction I'm not immune to this trap! The only way we can guarantee a "fat finger friendly" experience is to get out to job sites and see our customers using Varicon in real-life situations. This is a rich source of insight for us. Our philosophy is that if a user is confused the problem isn't them - it's us!⁣ ⁣ 𝐈𝐧𝐬𝐭𝐢𝐥𝐥𝐢𝐧𝐠 𝐄𝐦𝐩𝐚𝐭𝐡𝐲 𝐢𝐧 𝐎𝐮𝐫 𝐓𝐞𝐚𝐦: Understanding our customer's immediate functional needs is the easy part of building a great product. The hard part is having true empathy. Our customers don't have the knowledge that we have, and as non-tech people they don't have the skills that we have. They also have different priorities - construction workers want to build things, not play around on apps.⁣ ⁣ We gain empathy by spending time with our customers' front-line employees to really understand what's important to them and how we can make their jobs easier. Our goal is to make a product that management don't need to force users to adopt.⁣ That’s why I encourage everyone at Varicon, from engineering to support, to connect directly with clients and walk in their shoes. Empathy isn’t just a value here; it’s a practice that drives every decision we make. ⁣ For us, real impact in civil construction starts on the ground. Site visits aren’t just part of our process—they’re at the heart of it, ensuring we build something I would have valued back when I was in the field.⁣ P.S. 📸 from a recent visit to the amazing team at R A Smith Contracting 🚧👷♂️👷♀️ #civilconstruction #varicon #constructiontech #constructionaustralia