Scale AI

Between major elements of the EU AI Act going into effect this week and the many conversations I’m having with governments right now around AI sovereignty, it feels like the right time to share Scale AI's first Sovereign AI Framework. This framework breaks down the big strategic questions and lays the foundation for a practical decision-making guide for anyone developing a sovereign approach to AI. You can read the full write-up: https://lnkd.in/eZxmGxwe In the paper, we outline 5 steps to force thoughtful answers to tough tradeoffs, their implications, and fail-safe options: 1. Define the problem AI is actually solving. Is this really an AI question, or a bigger one about policy, cost, or who's accountable when something breaks? 2. Map the dependencies to successful deployments, not just the assets. A foreign dependency you can swap out easily is likely safer than a domestic one you can't replace, even if the domestic one feels more "owned." 3. Name the tradeoffs you're accepting, and rank them honestly. Safety and security come first. 4. Build the fallback before you need it. When you wait for the crisis, most of the good options are gone. 5. Revisit all of this on a regular cadence. A frontier model that’s a chokepoint today could be a commodity six months later. Proud of the team for all of the hard work on this. Please have a read: https://lnkd.in/em6zr-Kn

Excited to share that Francis deSouza is joining us as CEO, stepping in to lead the next chapter for Scale AI. The opportunity ahead is bigger than ever: partnering with frontier labs, enterprises, and governments to build the reliable AI systems the world's most important decisions depend on. Welcome to Scale, Francis. https://lnkd.in/eH7Ny5HN

Today Scale AI announced that I'm joining as CEO on August 10. AI will transform how the world's most important companies and institutions operate. But making these systems reliable enough to trust with critical work remains one of the industry's biggest challenges. Company after company gets an AI model to an impressive demo but stalls long before it reaches real decisions in production. Closing that gap is the work, and it's the problem Scale exists to solve. Scale has a rare vantage point. It works with the world's leading AI labs to push the frontier forward, while helping enterprises and governments deploy AI in the real world. Through its data, applications and evaluation work, Scale helps answer the question that increasingly matters: does the technology actually work? Leaving Google Cloud is hard. It's a remarkable team doing consequential work, and I'm grateful to Thomas Kurian and every Googler I’ve had the privilege to work with. I'm cheering for you. To the Scale board, thank you for the trust.  To Alexandr Wang, Jason Droege and the Scale team: you've created something special. I'm going to spend my first weeks listening. I’m excited to build with you.

190 days. That's the average wait for a liver transplant in the US. As Senior Manager of Enterprise Forward Deployed Engineering at Scale, Allegra Larche is working to change that. Her job is to dig into our customers’ toughest challenges, partnering with them to find the right problems to solve and build the foundation for real AI transformation to industries that can't afford to get it wrong.

Come build with us at SFHQ. As we continue to grow, we're looking for people who want to help develop reliable AI systems for the world's most important decisions. Discover open positions: https://scale.com/careers 

A new healthcare future is possible, and we’re proud to be working with Mayo Clinic to make it happen. This partnership is about more than technology; it is about creating a better and more effective healthcare experience to streamline clinical processes, enhance clinical decision support, and create a more personalized patient experience. We are already seeing meaningful impact, including patients getting 11 more minutes on average with their doctors, because doctors can now be doctors, and not just record reviewers. Learn more about how reliable AI is changing care delivery at https://lnkd.in/ek4mSSqe

For 250 years, generations of Americans have built, served, and solved some of the world’s hardest problems. This Independence Day, we recognize those who have strengthened and defended our nation. National security and those who serve depend on reliable AI. We’re committed to building AI systems to help protect the next 250.

Next week, we'll be joining the global AI research community at [ICML] Int'l Conference on Machine Learning in Seoul with 7 accepted papers. 🇰🇷✨ Our research explores the breakthroughs shaping the next generation of AI systems. Here's a preview of the work we'll be sharing ⬇️