Other

⚠️⚠️ This page is under construction ⚠️⚠️

Things I am playing with

My own movies recommender system: Building on my predoctoral research, I created a simple recommender system using embedded representations of movies. A simple demo version of the system is available here and I wrote a blog post showing how to estimate these representations.
Exploring the limits of large language models. Recently I documented the performance of multiple large language models on a classic creativity test applied by psychologists (i.e. alternative uses task). This project was part of the Inverse Scaling Challenge where people were encouraged to find tasks for which the performance of the model decreased as its size increased. Although other people found much more interesting and relevant tasks that exhibit an inverse scaling pattern, this pattern seems to also be present for the alternative uses task. Here is a link to the repository.

Being mostly an empirical researcher, I like to reflect on the ethical and philosophical foundations of empirical methods. What underlying assumptions about the world are different empirical methods using? I started thinking about these topics in the context of the History of Economic Thought (with the great Jimena Hurtado) and have gradually moved closer to the recent literature on fairness, accountability, and transparency in machine learning.