Research
by Parth Patel, Candice Pattisapu, Maxwell J.D. Ramstead and Guillaume Dumas
16 October 2025
by Parth Patel, Candice Pattisapu, Maxwell J.D. Ramstead and Guillaume Dumas
16 October 2025
by Maxwell J.D. Ramstead
16 October 2025
by Jeff Beck and Maxwell J.D. Ramstead
28 February 2025
by Maxwell J.D. Ramstead, Candice Pattisapu, Jason Fox and Jeff Beck
16 February 2025
by Candice Pattisapu et al.
10 September 2024
This project extends DreamerV2’s RSSM by integrating epistemic exploration and surprise-weighted replay. Both action and memory are guided by uncertainty minimisation, allowing the agent to learn efficiently from the most informative experiences. The result is a surprise-driven amortized world model that closes the epistemic loop between perception, action, and memory.
This work introduces a multi-temporal and context-conditioned world model that learns to segment continuous experience into meaningful events. Using Dynamic Markov Blanket Discovery (DMBD) to detect surprising events, the model maintains two coupled timescales: a fast sensorimotor layer for immediate transitions and a slow contextual layer that infers locally stable dynamics regimes. Context variables provided by the contextual layer modulate low-level inference and control through FiLM-style conditioning, leading to stable long-horizon dynamics and compact, event-level representations.Together, these studies outline a path toward world models that are uncertainty-driven, contextual, and temporally deep, combining the scalability of amortized world modeling with the theoretically-grounded approach of Active Inference.
In his talk, Maxwell Ramstead explored computational efficiency at the edge through the lens of thermodynamic computing, which harnesses thermal fluctuations as a native stochastic resource rather than suppressing them. He advocated for a hybrid computing stack in which digital processors orchestrate computation, quantum devices solve highly structured problems, and thermodynamic systems act as massively parallel probabilistic explorers. Together, they enable scalable, edge-deployable physical AI systems capable of safe, adaptive behavior in real-world environments.
The free energy principle (FEP), along with the associated constructs of Markov blankets and ontological potentials, have recently been presented as the core components of a generalized modeling method capable of mathematically describing arbitrary objects that persist in random dynamical systems; that is, a mathematical theory of ``every'' ``thing''. Here, we leverage the FEP to develop a mathematical physics approach to the identification of objects, object types, and the macroscopic, object-type-specific rules that govern their behavior. We take a generative modeling approach and use variational Bayesian expectation maximization to develop a dynamic Markov blanket detection algorithm that is capable of identifying and classifying macroscopic objects, given partial observation of microscopic dynamics. This unsupervised algorithm uses Bayesian attention to explicitly label observable microscopic elements according to their current role in a given system, as either the internal or boundary elements of a given macroscopic object; and it identifies macroscopic physical laws that govern how the object interacts with its environment. Because these labels are dynamic or evolve over time, the algorithm is capable of identifying complex objects that travel through fixed media or exchange matter with their environment. This approach leads directly to a flexible class of structured, unsupervised algorithms that sensibly partition complex many-particle or many-component systems into collections of interacting macroscopic subsystems, namely, ``objects'' or ``things''. We derive a few examples of this kind of macroscopic physics discovery algorithm and demonstrate its utility with simple numerical experiments, in which the algorithm correctly labels the components of Newton's cradle, a burning fuse, the Lorenz attractor, and a simulated cell.
This white paper describes some of the design principles for artificial or machine intelligence that guide efforts at Noumenal Labs. These principles are drawn from both nature and from the means by which we come to represent and understand it. The end goal of research and development in this field should be to design machine intelligences that augment our understanding of the world and enhance our ability to act in it, without replacing us. In the first two sections, we examine the core motivation for our approach: resolving the grounding problem. We argue that the solution to the grounding problem rests in the design of models grounded in the world that we inhabit — not mere word models. A machine super intelligence that is capable of significantly enhancing our understanding of the human world must represent the world as we do and be capable of generating new knowledge, building on what we already know. In other words, it must be properly grounded and explicitly designed for rational, empirical inquiry, modeled after the scientific method. A primary implication of this design principle is that agents must be capable of engaging autonomously in causal physics discovery. We discuss the pragmatic implications of this approach, and in particular, the use cases in realistic 3D world modeling and multimodal, multidimensional time series analysis.
Previous active inference accounts of emotion translate fluctuations in free energy to a sense of emotion, sometimes focusing exclusively on valence. However, in affective science, emotions are often represented as multidimensional. In this paper, we adapt a Circumplex Model of emotion to the Active Inference framework by demonstrating a mapping of free energy into valence and arousal, relating valence to utility less expected utility and arousal to the entropy of posterior beliefs. Under this formulation, we simulate artificial agents engaged in a search task and assign emotional states to them. We show that experimental manipulation of priors and object presence results in common sense variability in these assignments.