I've been working as a programmer for five years, and I recently made a late application to study International Affairs as a graduate student at Carleton. I have just been accepted and awarded the Joubin-Selig Scholarship. Provided I can complete two online economics courses in August with acceptable grades, I will start in the Fall.

This is a two year masters program with a summer co-op option, and I'd like to use the co-op (and my future career) to help governments with AI. 

I believe I have an unusually strong understanding of machine learning for someone whose interests primarily lean towards the humanities and social sciences (my undergrad is in philosophy from Oxford).

I am fluent in Python, I have built several small neural networks, and I have a reasonable understanding of the attention-based transformer architectures that are used by large language models (LLMs) like ChatGPT (I have yet to actually go through Attention is all you Need line-by-line alongside the pytorch transformer implementation, but it's on my todo list). I have spent time with researchers at Mila (informally), I remain in touch with the field, and I am confident that I can go more deeply into both academic literature and work on practical implementations of models where useful.

I'm unclear on how best I'd be useful at present, but one possibility is to increase technical literacy on a team to a point where they could have useful conversations with computer scientists. Concretely, this might involve leading a group through online courses like Andrew Ng's materials on Coursera, and facilitating dialogue with interested folks from research labs. 

I believe that some grasp of how neural networks are built is necessary to understand the ways they may not act as intended, and that this understanding is crucial to many, perhaps most, major decisions regarding AI that government will face. Outside of perhaps a few industry labs, technical researchers are not primarily selected based on their ability to engage with safety/governance issues, so I believe governments require sufficient technical know-how to critically assess their opinions when making governance decisions. 

I suspect my technical background could plug some important gaps in an AI governance team's knowledge. If you work for the Canadian government and you'd like to work with me, please send me a DM!