Jack Thompson does empirical research & analytic philosophy, and writes about it.
I study computer science, cognitive science, & philosophy at Princeton University (jackthompson[at]princeton[dot]edu). Currently, I’m doing AI introspection work with with Kirsten Ziman of Princeton Neuroscience Institute; AI risk & alignment research independently; and developing tools for learning with generative AI.
Areas of interest: artificial sentience, functional decision theory & superrationality, AI “thought control”, Pearl causal inference, interactive & independent learning, risk, and effective altruism.
My rules of writing
I don’t use AI to ghostwrite my posts. While I’ll often use AI in my research & learning process (quick simulation code, finding sources, helping me learn a new concept), the actual content of my posts is written by me. (Yes, I use em-dashes—I like using em-dashes and I used them before ChatGPT, darnit!) If I use LLMs for any of the text content, it will be quoted rather than slipped into the main text (for instance, to show how an LLM would understand a particular viewpoint).
I frequently publish small results. This is meant to be an open research log and place for discussion & analysis. Each post is another piece of the puzzle working up to a longer-term project, rather than a single big finding or two a year.
Empirical findings & rigorous analysis over “musings” and empty conjecture. I feel like there are enough substacks where people just dump their thoughts into the void or ramble on about subjects without making concrete, analyzable claims. I will try my best to ground every post here in (a) research results from some simulation, (b) mathematical or philosophical analysis, even if uncertain or tentative, or (c) significant personal experience.
No politics. There are other blogs for that. For my type of research, politics is the mind-killer. Culture-war comments will be removed.
