I’m Cédric Colas, AI researcher and bricoleur.
I work at Inria and MIT on artificial curiosity and open-ended learning. My research started with curiosity-driven reinforcement learning agents that set their own goals (autotelic agents), and has grown into questions about language as a cognitive tool for agents, social learning in hybrid human-AI populations, and how cultural dynamics might sustain open-ended innovation. More recently, I’ve been working on LLM-guided program synthesis and LLM-based agents that learn to code and reason. I did my PhD at Inria’s Flowers Lab, where I studied language-guided autotelic agents.
The rest of the time, I make things. I’ve built a pianocktail that translates piano music into cocktail recipes, a thermal-printer spirit with its own inner life, and an interactive installation where visitors rewire neural networks by hand. I make dirty video mixers, generate imaginary French words, map geography to color, and breed visual universes through artificial evolution. I also shoot film photography.
