I'm a PhD student in an AI lab, working at the intersection of machine learning and ecology. Three years in, my everyday work looks very different from what I anticipated when I started.
When I started, picking up a new programming language or framework meant hours or sometimes days of reading documentation before I could do anything useful. A simple bug would cost me hours. Figuring out how to deploy a model efficiently was its own long round of trial and error. Today, LLMs explain unfamiliar tools, spot the bug, and suggest new ideas for experimentation, often in a fraction of the time.
This isn’t just about saving time. It’s a sign that the everyday work of an AI lab is being reorganized. This article is my view of what is moving and what is staying put.
What actually got faster
The popular story says AI is now doing the work of graduate students. There is some truth in it.…