Evolutionary Computation: A Paradigm of Natural Generative AI and Creative Synthesis
Authors: Abhishek Gupta, Yaxin Shi, Ying Wu, Melvin Wong, Ivor Tsang, Thiago Rios, Stefan Menzel, Bernhard Sendhoff, Yaqing Hou, Yew-Soon Ong
Organizations: Centre for Frontier AI Research (CFAR), Agency for Science, Technology and Research (A*STAR), Singapore · School of Mechanical Sciences, Indian Institute of Technology (IIT) Goa, India · School of Computer Science and Technology, Dalian University of Technology (DLUT), China 116024 · College of Computing and Data Science (CCDS), Nanyang Technological University (NTU), Singapore · Honda Research Institute Europe (HRI-EU), Offenbach am Main, Germany · Chief Artificial Intelligence Scientist with the Agency for Science, Technology and Research (A*STAR), Singapore · Data Science and Artificial Intelligence Research Centre, School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore
Generative AI relies on statistical learning and local gradient-based guidance, often resulting in artifacts that may be more emulative than genuinely creative. This position paper advances a perspective on Evolutionary Computation (EC) as a paradigm of Natural Generative AI (NGAI), in which stochastic variation, diversification, and natural selection present an alternative foundation for generative exploration and creative synthesis. EC provides an array of variation operators with varied generative capabilities: parent-centric operators induce population distributions analogous to conventional generative models, while disruptive operators can produce structured evolutionary leaps to out-of-distribution artifacts (e.g., programs, designs, or AI agents themselves). This lens reveals a unique theoretical association between certain disruptive variations in multitask EC and a product-of-distributions population structure. A mathematical result derived herein shows that a product structure also occurs under a local second-order approximation of an established generative adversarial program for creative extrapolation, inspired by Martindale's cognitive theory of artistic creation. The simultaneous existence of multiple fitness landscapes in multitask EC inherently moderates selection pressure, allowing these novel artifacts to survive and seed further innovation. Structured disruption and selection moderation are thus argued to be key elements to support creative synthesis. Agnostic to any specific algorithmic realization, the perspective positions NGAI as an emerging engine for open-ended scientific exploration.
Figures & tables
Fig. 1: Illustration of the EDA workflow where modern GenAI algorithms, as opposed to the biologically-inspired genetic operators of classical EC, are used for solution variation .
Fig. 2: Visualization of offspring generated by GenAI-EDA (with explicit population modeling using a VAE) and the parent-centric SBX operator (under different distribution parameter values) . Complex parental distributions are used to assess generative behavior in a general setting. Notably, the SBX operator is gradient-free and does not require the training of a statistical model.
Fig. 3: Visualization of offspring generated by explicit probability mixture models (Gaussian mixtures) and a parent-centric variation operator (SBX) in a multitask environment. PopAs(t) and PopBs(t) represent parent populations for Tasks A and B at generation t , while PopA(t+1) and PopB(t+1) represent the offspring populations for these tasks. The mixing of colors within each convex hull indicates inter-task transfer.
Fig. 4: Visualization of offspring produced in a multitask environment with disruptive genetic operators versus population sampling from a product of generative models. The red circle marks an evolutionary leap outside the parent convex hulls due to the structured disruption of the variation operator.
Fig. 5: LLM2TEA for multitasking with LLM-guided natural language solution encoding [ 80 ] . Generated offspring illustrate gradual out-of-distribution design synthesis through the composition of task-specific features (of cars and airplanes) encoded in the form of textual prompts.
Fig. 6: Trajectory of MTEC during joint aerodynamics optimization of airplanes (top row) and cars (bottom row). C~d represents the drag coefficient of generated designs as the aerodynamic performance indicator to be minimized. As evolution progresses (from left to right), the genotypic traits emerging in the population of one task are transferred to offspring of the other task (highlighted in black) due to the compositional effect of inter-task crossovers. The schema expressed in natural language for blending a sedan with an airplane appears in the early stages of the airplane domain and persists across subsequent stages of both tasks, eventually producing a car with airplane-like features at the bottom right.
Fig. 7: Examples of LLM2TEA-generated designs that intersect distinctive features of car AND airplane shapes, analogous to the evolutionary jump depicted in Fig. 4 .
This paper investigates the creative process of automated design and artistic evaluation using an evolutionary system. We consider how a multimodal artificial intelligence (AI) model can communicate and guide a combined generative and evolutionary computational system. This creates a framework for the evolution of aesthetically pleasing complex 3D organic forms by integrating genetic algorithms with the visual reasoning capabilities of large-scale AI foundation models. The framework shifts the artist role from that of intensive direct selection to one of system design; transferring detailed step-by-step curation to an AI agent capable of multimodal aesthetic judgement. This framework enables the human artist/designer to rapidly traverse large areas of multi-dimensional evolutionary parameter space to find creative outcomes based on their semantic targets. Detailed audit trails of the AI's aesthetic reasoning are generated for each experiment. Interactive visualisation tools, together with AI-generated summaries and evolutionary narratives, enable deep exploration into each evolutionary experiment and providing a transparent insight into the AI-guided process.
Dylan Banarse, Stephen Todd, William Latham +1
Google DeepMind · London Geometry Ltd. · School of Computing at Goldsmiths, University of London
The saturation of high-quality pre-training data has shifted research focus toward evolutionary systems capable of continuously generating novel artifacts, leading to the success of AlphaEvolve. However, the progress of such systems is hindered by the lack of rigorous, quantitative evaluation. To tackle this challenge, we introduce CreativeBench, a benchmark for evaluating machine creativity in code generation, grounded in a classical cognitive framework. Comprising two subsets -- CreativeBench-Combo and CreativeBench-Explore -- the benchmark targets combinatorial and exploratory creativity through an automated pipeline utilizing reverse engineering and self-play. By leveraging executable code, CreativeBench objectively distinguishes creativity from hallucination via a unified metric defined as the product of quality and novelty. Our analysis of state-of-the-art models reveals distinct behaviors: (1) scaling significantly improves combinatorial creativity but yields diminishing returns for exploration; (2) larger models exhibit ``convergence-by-scaling,'' becoming more correct but less divergent; and (3) reasoning capabilities primarily benefit constrained exploration rather than combination. Finally, we propose EvoRePE, a plug-and-play inference-time steering strategy that internalizes evolutionary search patterns to consistently enhance machine creativity.
Zi-Han Wang, Lam Nguyen, Zhengyang Zhao +4
Southern University of Science and Technology · Tsinghua University · Xi’an Jiaotong University +3
Many of the most consequential advances promised by AI, from curing diseases and inventing new technologies to formulating new scientific theories, are fundamentally creative acts. Yet creativity is uniquely difficult to learn because it faces a creative paradox: learning to repeat past examples of creative success makes those specific examples more likely to recur, inhibiting the very creativity desired. This work argues that creativity is better modeled as an explicit process rather than as a property of individual artifacts, and proposes to learn the underlying generative principles behind the creative process. In particular, to escape the paradox, we introduce creative priming (CP), wherein a trained priming model generates a prompt-specific primer that assists a response model in being creative, explicitly decoupling the creative process from response generation. The priming model learns through GRPO, rewarding primers that lead to responses that are novel, high-quality, and diverse, thereby reinforcing the process that yields creative responses without reinforcing any specific response.
Alon Albalak, Sebastian Gabriel, David B. D'Ambrosio +2