VLM Replicates Picbreeder's Open-Ended Creation
matdryhurst · x · 2026-07-12
Sakana AI presents research on open-ended creation, asking whether agents can exhibit creativity without being told what to create.
In collaboration with MIT and NYU, their GECCO 2026 paper, "In Search of the Ingredients of Open-Endedness: Replicating Picbreeder with Large Vision-Language Models," reproduces Picbreeder, an early collaborative evolutionary image website. The original site let users continuously select "interesting" images, evolving them over generations into unexpected results like faces, animals, vehicles, and skulls. The team adapted this process into an exploration experiment for vision-language model agents, having them generate and select images without predefined goals to observe VLM performance in an open-ended creative space.
A reshared comment points out that the experiment's true value isn't just about "which model is better," but that it provides a rare historical benchmark for comparing how humans and AI explore open-ended environments.
Related event: Sakana AI Replicates Picbreeder with VLM Agents for Open-Ended Creativity(14 posts)→
More from Multimodal
- Reddit user chains Ideogram 4 and Krea2 to mimic bbox-based image positioning — v3lh0t05c0 · 2026-07-22
- Ultimate Face Fix: Open-Source Multi-Face Repair Node for ComfyUI — Merserk13 · 2026-07-22
- Getting Started with AI Video: Solving Consistency and Censorship — cynicalnewenglander · 2026-07-22
- Storyboard-first workflows are making AI dance videos and influencers more consistent — aftahi_ai · 2026-07-22
- Runpod MCP and Claude help spin up image and video generation workflows — 802high · 2026-07-22
- Midjourney prompt turns a bee into a glitching pixel explosion — michaelrabone · 2026-07-22