Modelling the world at scale.

This is Humanis' north star.

From physics to biology, telecommunications to energy grids, robotics motion to fluid dynamics, we are building foundational technology to simulate and neuro-symbolically model these systems at scale.

World-models can enable deploying advanced AI in a safe way, by acting as a sandbox and a tripwire. Scaling world-modelling technology to human-level AI may well be the highest assurance AGI deployment path available[1].

Beyond the mission, we love the aesthetics of this research direction: modelling the world in an understandable way is arguably one of the most beautiful scientific endeavours one could pursue.

Building such models in a neuro-symbolic way unlocks the modularity sufficient to obtain guarantees regarding AI policies, ranging from statistical bounds to formal verification.

A renaissance of such methods is coming, driven by massively scaling both frontier AI labour and the computing power spent on such algorithms. If you're one of the few who believes in both scaling and symbolic methods, we intend to be the best place to pursue your life's work.

We are based in London and are assembling a talent-dense team of researchers and engineers willing to bend the arc of AI development.

You can reach us at contact@humanis.ai

You can express interest in working with us here.