So, why Callosum?

Recently Dan Akarca and I launched Callosum (www.callosum.com). Here I wanted to share some longer form thoughts on what we are trying to create with Callosum:

It starts from the fact that there is no single answer to what creates intelligence, no single mystery to discover, no single 'moat' to build. Intelligence is a systems-level problem, and what makes systems powerful is not any one component but the diversity of co-optimised mechanisms working together. The real 'moat' is building the architecture to harness that diversity and realise the heterogeneity that underlies intelligence itself.

In compute terms, we are finally at a moment where genuine heterogeneity is possible. New AI accelerators, revolutionary neo-chips design, new HPC networking technologies, and new algorithms have each made remarkable strides, but the real unlock is in their orchestration. Integrating and co-optimising these into one cohesive platform is what creates something categorically more powerful than any single component alone. Callosum strives to be the Intelligent Systems Company that realises intelligence at scale through this kind of heterogeneous compute orchestration, opening up worlds of AI and computation that one could not have imagined on today's systems.

We are sharing this vision on the back of real breakthroughs. We have built new system-level algorithms showing benchmark-beating performance through heterogeneous hardware (https://www.callosum.com/blog/welcome-heterogeneous-intelligence). Additionally, we have developed a new mathematical theory demonstrating that heterogeneous scaling laws drive better performance, efficiency, and robustness, just as they do in biological, neuronal, and economic systems (https://www.callosum.com/blog/the-principle-of-maximum-heterogeneity). We have also built an alliance of cloud providers and neo-chip companies, together with the Advanced Research + Invention Agency (ARIA), to begin the first heterogeneous cluster build (https://www.callosum.com/blog/we-are-scaling-heterogeneous-compute). We will share more details in the coming weeks and months but the blog posts linked above should give you a good first overview of our vision for the future of AI and compute.

This is the vision Dan and I are so excited about — one that will lead us to both understanding and building truly intelligent systems.

We are incredibly grateful for the mentors and colleagues who shaped this vision through many discussions, spanning from our PhDs together at Cambridge, through fellowships at Imperial College and Oxford, to conversations with individuals like Matt Clifford, Ian Hogarth, James Richards, and Suraj Bramhavar who gave us the final push to go big. A special last big thank you goes to my PhD supervisor John Duncan, who really is the one who got me started on wondering about the question where intelligence comes from. Reading his book 'How intelligence happens' before starting to work together during my undergraduate and PhD has had a massive influence on my thinking ever since.

Dan and I are beyond excited for what is to come — please follow this journey of Callosum!