← home

About

In the summer of 2016 I followed my favourite professor, an Aristotle scholar, to London, and spent it reading Aristotle and Confucius side by side. I was hoping to write about artificial moral agents: whether a machine could have something like virtue, and which virtues we’d want it to have if it could. This was a year before the paper that introduced the transformer, the design behind today’s language models, so it was a thought experiment with footnotes. It became my senior thesis.

I didn’t expect that within a decade I’d be setting AI models against each other in moral philosophy, from my laptop, for a few cents a run. Symposium is that experiment. (The thesis argued for virtue ethics. Asked the same question, the frontier models mostly prefer consequentialism; only Claude sides with me, three times out of five.)

I spent the years in between making sure the words in contracts meant exactly what they were supposed to, which turns out to be good training for working with language models, which have their own ideas about what words mean. Now I learn the way Socrates recommended: by being wrong in public. The other experiments are in Projects, and what I learn from them, along with whatever else I’m thinking about, goes into Thoughts. If you spot a mistake in any of it, Socrates would want you to tell me.