I’m Dyana, a visiting student from the US collaborating with Frank, as well as with Charles Kemp. Broadly, I am interested in what makes human learners unique from machine learners. Humans have an incredible capacity to learn arbitrary rules, concepts, and complex systems like language. Machine models, such as large language models, are becoming increasingly better at imitating human performance, and the level of performance they can achieve is surprisingly high considering they are just statistically modeling their input data. However, they are still clearly lacking in many regards. Would closing this gap be just a matter of training beefier statistical models? Or are there priors and/or systems that humans have built-in that give us a leg up in learning?

When I’m not contemplating such existential questions, you can often find me eating, making food to eat, or contemplating what to eat. Sometimes I do these things out in nature on camping, backpacking, or bikepacking trips!