The unsolved problem of granular composition is coherent structure at meso and macro scales, not just generating clouds

Cloud, stream and spray models are effective because they agglomerate grains into the object (100 ms–8 sec) and meso (>8 sec) time scales. The deeper challenge is creating coherent multiscale behavior extending all the way to the macro time scale — where ‘long-term high-level forces are as powerful as short-term low-level processes.’ Simply making larger sound objects out of grains is not enough; the levels must cohere. This is why simplistic bottom-up strategies of ‘emergent self-organization’ tend to fall short: grains are too low-level to carry large-scale musical logic on their own. Roads calls building this multiscale coherence ‘one of the great unsolved problems in algorithmic composition.‘

Examples

Xenakis’s speculative hierarchical grain clusters, where each cluster point itself contains a sub-cluster; six of his nine ‘Music composition treks’ proposals used a high-level stochastic process to shape macroform while a separate low-level process computed the sounds.

Assessment

Why is generating interesting grain clouds insufficient for granular composition, and what does coherent multiscale behavior require beyond making bigger objects?