I am seeking a deeper and more universal understanding of how adaptive optimization underpins perception, knowledge formation, reasoning, planning, and creativity. Theory-based learning, compositionality, and out-of-distribution generalization are topics that especially fascinate me.

To build the next generation of artificial intelligence that could provide ingenious solutions, uncover novel theories, and, in the long run, significantly augment the scientific journey, we need new approaches. Open-ended settings like curiosity-driven exploration, curriculum learning, hypothesis testing, and long-term control/planning are promising ways to push models to their limits and yield relevant experiences to learn from. Children and scientists are exemplar learning agents in the world, and algorithms inspired by their behavior have good chances of furthering this area of research.

Finally, these cognitive phenomena are being studied in relation to both biological and artificial systems, I like wondering about the intricate relationships and possible synergies between the two.