ELLIS PhD Student · Berlin
Alejandro Hernandez Artiles
MPI for Human Development · MPI for Intelligent Systems
Supervised by Iyad Rahwan & Bernhard Schölkopf
About
Hi! I’m Alejandro, an ELLIS PhD student at the Max Planck Institute for Human Development and the Max Planck Institute for Intelligent Systems, co-supervised by Iyad Rahwan and Bernhard Schölkopf.
I believe science and art are evolutionary processes, where ideas are mutated, shared and selected over time by a population of agents. My research therefore explores open-ended AI for cultural and scientific discovery. Inspired by cultural evolution, I develop methods that help AI systems continually explore original, coherent ideas in domains without clear reward signals, such as early-stage scientific ideation.
Outside research, I enjoy creative writing, literature, cinema, video games and opera.
I really like high perplexity ideas! If you are interested in my research or would like to explore ideas or collaborations, feel free to send me an email!
News
- The Alien Space of Science was accepted to the NeurIPS 2026 Main Track! 🎉 Paper
- New preprint: Time Machine Experiments, where we use historically-bounded language models to study perceptions of moral decline.
- Presenting Lacuna: A Research Map for Machine Learning at the AI4Science Workshop at ICML 2026.
- Started a research internship at Tiptree Systems on AI for scientific discovery.
- Presenting the Cultural Alien Sampler at the NeurIPS 2025 Creative AI Track in San Diego.
Selected publications
-
NeurIPS 2026
The Alien Space of Science: Sampling Coherent but Cognitively Unavailable Research Directions
Advances in Neural Information Processing Systems (NeurIPS), Main Track
Sampling scientifically coherent research directions that lie beyond the ideas existing research communities are most likely to produce.
-
Preprint 2026
Time Machine Experiments: Using Historically-Bounded AI for Inquiry into the Human Mind
arXiv preprint · * equal contribution
-
NeurIPS 2025
Cultural Alien Sampler: Open-ended Art Generation Balancing Originality and Coherence
Advances in Neural Information Processing Systems (NeurIPS), Creative AI Track
A concept-sampling method that balances originality and coherence in art generation, evaluated against GPT-4o and human art students.