Biological cells can be regarded as interacting agents whose collective dynamics give rise to adaptive behaviour across multiple organisational levels, from single cells to tissues and whole multicellular organisms. This article asks whether intrinsic rewards in artificial neural systems can support adaptation, functional specialisation and higher‑level self‑organisation without a shared external objective. We review empowerment, curiosity, learning progress, information gain, unsupervised skill discovery, mutual‑information estimation and the use of world models for computing intrinsic rewards. Particular attention is paid to failure modes where such objectives do not sustain exploration or generate increasingly complex behaviour. We argue that more capable systems may need complementary objectives, communication, memory, learning at multiple temporal scales and environmental constraints. Building on this perspective, we outline three experimental directions: (1) a resource‑constrained environment where otherwise stable behavioural attractors become unsustainable, allowing us to test whether environmental constraints can mitigate characteristic failure modes of intrinsic objectives; (2) a network of recurrent agents each receiving its own intrinsic reward, to study self‑organising group dynamics; (3) a hierarchical world‑model agent in which exploratory motor competence develops before goal‑directed behaviour, testing whether intrinsic learning can lead to adaptive organisation at progressively higher levels.
Review