Sakana AI Replicates Picbreeder
SakanaAILabs · x · 2026-07-15
Sakana AI published a study replicating the 2008 Picbreeder open-ended evolution experiment using Large Vision-Language Models (VLMs). Their approach involves a "tribute" rebuild of the original website, allowing CPPN images generated by VLMs to be: - Branched out for further evolution - Scored and evaluated - Individually inspected and reviewed This work attempts to explore the components of "open-ended generation/evolution," investigating whether large model agents can recreate an emergent creative process similar to Picbreeder.
Related event: Sakana AI Replicates Picbreeder Experiment with VLMs(2 posts)→
More from Multimodal
- MiniCPM-V 4.6 now runs locally on iPhone with no cloud dependency — amos_gyamfi · 2026-07-21
- Creator says they no longer shoot with a camera, but with prompts — taherdhanera · 2026-07-21
- PixVerse demo turns into a full sci-fi dark comedy set on Mars — aliscodes · 2026-07-21
- Alibaba’s Qwen-Audio-3.0-TTS-Plus takes #1 on Artificial Analysis Speech Arena — airesearch12 · 2026-07-21
- Reddit users say anima_turbo generates brighter images and runs much faster — DeliciousMoraxMora · 2026-07-21
- Qwen launches Qwen-Image-3.0 with a focus on richer content and authentic detail — ilreb · 2026-07-21