Why interpretability keeps pulling me in
I keep coming back to the same question in almost everything I work on, whether it's a quantum circuit, a neural network, or eventually, a person: what's actually happening underneath the behavior we can see?
Most of my technical work right now sits in AI interpretability, trying to understand why a model produces the output it does, rather than just accepting that it works. It's tempting to treat a well-performing model as a black box and move on. But the interesting part, the part that actually matters, is in the mechanism. What's driving the decision. What would change it.
That same instinct is a big part of why I'm working toward psychiatry. The parallel isn't metaphorical to me, it's the same underlying curiosity applied to a different kind of system. A mind is not a neural network, and I don't think the comparison should be stretched too far, but the core question is the same: something complex is producing a behavior, and the behavior alone isn't the full story.
I don't have a tidy conclusion here yet, this is genuinely still something I'm working through, both in my research and in how I think about where I'm headed. But it's the thread that ties together most of what I'm building right now, and I wanted to put it into words somewhere other than my own head.
More soon.