On the stochastics of human and artificial creativity
Solve Sæbø, Helge Brovold
TL;DR
The paper tackles whether computers can exhibit genuine creativity and, more broadly, how to reach Artificial General Intelligence. It proposes a statistical, Bayesian representation of human creativity, framing it as a two-stage stochastic process with a proposal step that generates novel ideas and an evaluation step governed by a bias or world model, formalized via $P(M|D) \propto P(M) P(D|M)$. Drawing on Bohm, McCarthy, neuroscience, and chaos theory, it argues that creativity involves dynamic restructuring of biases and that neural and cognitive networks support both divergent exploration and evaluative refinement. When applied to modern AI—Large Language Models, diffusion models, and reinforcement learning—the framework finds that current systems produce novelties that "work" only within fixed biases and lack autonomous capacity to transform priors or evaluate usefulness, thus falling short of human-level creativity. The work highlights a promising path via predictive coding and world-model integration to realize artificial creativity, impacting how we assess and design future AGI systems.
Abstract
What constitutes human creativity, and is it possible for computers to exhibit genuine creativity? We argue that achieving human-level intelligence in computers, or so-called Artificial General Intelligence, necessitates attaining also human-level creativity. We contribute to this discussion by developing a statistical representation of human creativity, incorporating prior insights from stochastic theory, psychology, philosophy, neuroscience, and chaos theory. This highlights the stochastic nature of the human creative process, which includes both a bias guided, random proposal step, and an evaluation step depending on a flexible or transformable bias structure. The acquired representation of human creativity is subsequently used to assess the creativity levels of various contemporary AI systems. Our analysis includes modern AI algorithms such as reinforcement learning, diffusion models, and large language models, addressing to what extent they measure up to human level creativity. We conclude that these technologies currently lack the capability for autonomous creative action at a human level.