Causal Labs: Towards Causal Intelligence via AI Physics Models for Weather

A key bottleneck on the path to superintelligence is understanding causal relationships. We plan to fix that at Causal Labs by building a new kind of AI that understands cause and effect in the physical world, starting with predicting & controlling the weather.

Today, we’re excited to share that we’ve raised $6 million in seed funding, led by Kindred Ventures, with participation from Refactor, BoxGroup, Factorial, Otherwise, Karman Ventures, and an amazing group of angel investors and advisors. This funding will allow us to expand our team, make critical progress on our first-generation model, and continue pilot programs across key industries.

Our Mission: A general model for causal intelligence

We spent our formative days building and deploying safety-critical models for robotics and self-driving cars at companies like Cruise, Waymo, and Google Brain. We realized that one of the largest hurdles to full-scale deployment of autonomous vehicles was the inability of models to generalize beyond their training data—they struggled with out-of-distribution, unexpected, real-world scenarios. This limitation stems from a lack of causal reasoning: today’s AI doesn’t reason from first principles, nor does it understand how actions influence the world.

We’re on a mission to solve that problem. That’s why we’ve named it Causal Labs.

Up first: Predicting and controlling the weather

We believe that solving weather—particularly the breakthroughs required to model and shape chaotic yet physics-driven systems captured by widely available large-scale, multi-sensor datasets—will unlock a new class of physics-based AI models capable of Causal Intelligence.

Our approach draws from our experience in the self-driving domain. Just as large-scale, multi-sensor datasets enabled vehicles to perceive, plan, and act autonomously in the real world, we are applying similar techniques to create models capable of real-time, high-resolution weather forecasts, and interpreting them to help individuals, businesses, and governments make optimal decisions.

This is important not just as a pathway to useful superintelligence, but for the profound societal value it delivers today. Weather underpins the daily life of every individual and business, and as extreme events become more frequent and volatile, humanity needs better methods to predict, adapt, and ultimately take control of our environment. As our models become more advanced over time, we aim to one day enable humanity to—safely and responsibly—shape the weather to effectively fight wildfires, alleviate droughts, and decrease the intensity of hurricanes.

Our operating principles

We’re not a research lab. First, we are strong believers that solving difficult problems requires learning from intentional, real-world feedback loops. Second, safety is the centerpiece of how we’ll build here at Causal Labs, underpinning our model, our business, and the use cases we address.

We’ve spent years leading the development of safety standards in robotics and plan to similarly lead the way of ensuring our physics models are safe and steerable. We’ll follow in the footsteps of some frontier labs performing state-of-the-art safety research on LLMs while keeping in mind that it will be even more critical as our models operate in the real world where actions may not be reversible.

Join us

We’re excited to reveal our first generation model soon as we use this seed funding to build up our team, pilot programs across critical sectors, and make foundational progress on our model.

If the idea of building physics models to tackle unsolved problems that most think are impossible is more exciting than it is scary, come join us!

-Dar & Kelsie, on behalf of the Causal Labs team