Universal Algorithmic Intelligence

This is a guest post by Iliad. Our thanks to Marcus Hutter for the invitation to post here.

TL;DR: Iliad runs funded training and research programs in mathematical AI alignment, and we think the UAI community is well prepared for them. Applications for the October and November Intensives are open (apply here). If the programs aren’t the right fit for you at the moment, we pay up to $1,500 for referring someone who is.

Why we’re posting here

The UAI community studies idealized notions of intelligence by investigating Solomonoff induction, AIXI, and algorithmic information, with the hope that insights therein have implications for real minds and machines. AI safety can benefit from that perspective. Iliad is a nonprofit doing research and fieldbuilding to develop AI safety as a theoretical and experimental science. We run taught and research programs, host the ILIAD conference series, recently launched the peer-reviewed Alignment Journal, and incubate research groups. One of them being AIXI Labs, which models AI risk factors and safety mitigations in terms of AIXI variants and develops the means to translate both to real agents.

AIXI in the Iliad curriculum

AIXI has been part of the curriculum of the Iliad Intensive from the beginning, with half a day devoted to it in the April iteration. The materials are created by David Quarel, co-author of the textbook on Universal Artificial Intelligence together with Marcus Hutter and Elliott Catt. Since then, the materials have grown larger, with a math-based worksheet that covers more than a typical student can work through in a day, and an additional module on Solomonoff induction. In the intensive itself, they are typically taught together in a one-day component.

The component on AIXI is embedded into a larger cluster in agent foundations, which studies various aspects of reinforcement learning (including goal misspecification and goal misgeneralization, and an axiomatic justification of RL), classical MIRI-style agent foundations, decision theory, world models including transducers, natural abstractions including condensation, and theoretical approaches to instrumental convergence pioneered by Alex Turner et al. Each intensive iteration then picks a subset of these modules, and teaches them alongside modules in the clusters of learning, interpretability, and safety procedures. The exact instantiation of many of these notions in the context of AIXI may not always be established, which we think can provide ample opportunity for further research that formalizes and analyzes aspects of agency and AI safety in a principled way. 

Why you, specifically

Foundational alignment research is bottlenecked not only on headcount but also on people who can do mathematics at the frontier. And many of the field’s open problems are AIT-shaped: what is an agent; when is a learned world model faithful; what does optimality mean for embedded, resource-bounded reasoners; which guarantees survive distribution shift or self-modification. If you’ve spent real time with Kolmogorov complexity, universal priors, or AIXI, you already have the native vocabulary. What’s usually missing is a map of the field, collaborators, and funding. That’s the gap Iliad exists to close: find mathematicians who can do foundational work, bring them to the frontier of foundational AI safety research fast, then help them find a footing in the field.

The programs

Iliad Intensive: a four-week, full-time taught course on foundational alignment research (deep learning theory, agent foundations, interpretability, safety guarantees) for mathematicians, physicists, and theoretical computer scientists. Runs several times a year in London (at LISA) and in Berkeley. No tuition; participants receive a $5,000 travel-and-housing allowance, plus lunch and dinner on course days. Applications are open for the October and November 2026 cohorts: apply here. No PhD required.

Iliad Fellowship: a three-month mentored research fellowship in applied mathematics for AI alignment, in person at LISA in London, with a $6,000/month travel-and-housing allowance. Fellows do research, contribute chapters to the Textbook From The Future. Our goal for our Fellows is to get a job, continue in a fellowship extension, or get funding to start a project. There are many possibilities to continue the work for 6–24 months. One application covers all upcoming iterations of both programs, so you can apply once and pick dates later. Many Fellows come through the Intensive first.

Know someone who fits?

Many of you teach or supervise exactly the people we’re looking for. Our referral program pays $500 when someone you refer completes the Intensive, or $1,500 when they start the Fellowship. Details and sign-up: https://www.iliad.ac/invite. You’re among the first communities we’re bringing this to.

Thanks

Thanks again to Marcus Hutter for the opportunity to post here. Questions are welcome via contact@iliad.ac. We’d be glad to see UAI regulars in London or Berkeley.

Aron Mill – Program Coordinator & Outreach Specialist – Iliad