Here are some thoughts about my future plans:
I think I will start to write things about different topics in my Substack.
Since a while ago, I wanted to celebrate two somewhat related Finnish masterpieces, the book Painful intelligence: What AI can tell us about human suffering by Hyvärinen, and the movie The man without a past by Kaurismäki. At some point, I will write it.
I am fascinated by the deep questions at the intersection of neuroscience, machine learning, and neuro-inspired artificial intelligence - questions that cut to the core of how natural and artificial systems process information, act in the world, and generate meaning. Where does agency, an act of intentional behaviour, come from, and how does a physical system move from collection of unorganised components to elaborate part-whole hierarchies, and from there to the rich phenomenon of goal-directed behaviour? Then how does the same agent formulate a well-defined problem for its survival out of myriad amounts of environmental stimuli it receives? Can we trace the origins of reward and intrinsic motivation in biological brains, and can such principles be formalised in artificial ones - through frameworks such as empowerment, minimising the entropy of the agent’s state visitation estimated using a latent state-space model, and maximising occupancy of future paths of actions and states?
I am equally intrigued by the relationship between cognition, consciousness, and their physical substrate. Does the brain “compute” in the way a digital computer performs an algorithm (in the way that computational functionalism assumes), or is its mode of operation closer to the adaptive feedback-driven dynamics of cybernetics contingent to its physical substrate? Marr’s theory of computation, algorithm, and implementation invites us to think in layers, but perhaps these layers co-evolve in physical computing systems such as brain, blurring the neat separations we often impose. This raises further questions: what counts as physical computation? Must computation always involve representation, or can it be embodied in the very physics of dynamical systems?
These lines of inquiry naturally connect to the field of Neuromorphic Engineering — where hardware and software are designed not just to emulate but to co-evolve with biological principles of intelligence. Neuromorphic systems may provide a more energy-efficient, sample-efficient, and robust substrate for intelligent behaviour, grounded in principles like predictive and prospective coding, and even the least-action principle as temporal mechanisms. In this sense, the study of neuromorphic systems is not only an engineering pursuit but also a philosophical one: it asks us to rethink what it means to compute, to perceive, to act or behave, and perhaps even to be or exist in time.