Self-organization of complex morphological patterns from local interactions is a fascinating phenomenon in many natural and artificial systems. In the artificial world, typical examples of such morphogenetic systems are cellular automata. Yet, their mechanisms are often very hard to grasp and so far scientific discoveries of novel patterns have primarily been relying on manual tuning and ad hoc exploratory search. The problem of automated diversity-driven discovery in these systems was recently introduced [26, 62], highlighting that two key ingredients are autonomous exploration and unsupervised representation learning to describe "relevant" degrees of variations in the patterns. In this paper, we motivate the need for what we call Meta-diversity search, arguing that there is not a unique ground truth interesting diversity as it strongly depends on the final observer and its motives. Using a continuous game-of-life system for experiments, we provide empirical evidences that relying on monolithic architectures for the behavioral embedding design tends to bias the final discoveries (both for hand-defined and unsupervisedly-learned features) which are unlikely to be aligned with the interest of a final end-user. To address these issues, we introduce a novel dynamic and modular architecture that enables unsupervised learning of a hierarchy of diverse representations. Combined with intrinsically motivated goal exploration algorithms, we show that this system forms a discovery assistant that can efficiently adapt its diversity search towards preferences of a user using only a very small amount of user feedback.

Summary and Contributions: This paper gives interesting evidence for the need of modularity for better exploration and diversity in goal-driven systems. In particular, the papers shows that a single neural network is worse than a dynamic modular architecture of sub-networks on diversity metrics. The paper formulates the notion of meta diversity, a search algorithm to hierarchically (tree structured) build modules that route decisions to either create new networks to handle incoming observations or decode it outcomes. A continuous version of the game-of-life system is used (Lenia) as the experimental platform. In this system, different starting states, rules and interventions lead to vastly different outcomes.

This paper proposes a hierarchical method for representation learning and goal-directed search in morphogenetic systems, and is evaluated on a particular type of cellular automata (Lenia). This method allows for identifying diverse and "interesting" regions of space in the dynamical system, also supporting small amounts of human feedback to identifying preferred regions of space. R1 and R4 praised the novelty of this approach, with R3 also finding it interesting and highlighting its implications for other areas of research. The reviews initially had some concerns with clarity, but these were satisfactorily addressed by the rebuttal. Another issue, highlighted by R1 and R4, was that the system has only been evaluated on a single dynamical system, and so its applicability to other domains is unclear.

Self-organization of complex morphological patterns from local interactions is a fascinating phenomenon in many natural and artificial systems. In the artificial world, typical examples of such morphogenetic systems are cellular automata. Yet, their mechanisms are often very hard to grasp and so far scientific discoveries of novel patterns have primarily been relying on manual tuning and ad hoc exploratory search. The problem of automated diversity-driven discovery in these systems was recently introduced [26, 62], highlighting that two key ingredients are autonomous exploration and unsupervised representation learning to describe "relevant" degrees of variations in the patterns. In this paper, we motivate the need for what we call Meta-diversity search, arguing that there is not a unique ground truth interesting diversity as it strongly depends on the final observer and its motives.