My first AI experience
My first contact with AI happened because of a school assignment in Taipei, 2020 — and a lucky draw in a class raffle.
At that time, I looked back on quite a few years of dreaming about inventing a useful protein. Designing new molecular machines was my dream, and just two years earlier Frances H. Arnold had won the Nobel Prize for designing new proteins. A few of these were already industrially produced and mixed into household appliances — laundry detergent, for instance, for improved cleaning.
It was then that a paper was published using frontier technology to solve a task first stated decades earlier by another Nobel laureate. The paper made quite a jump on the CASP14 leaderboard, which tracked progress on the task. I nearly overlooked it — first, I was intimidated, doubting my ability to grasp the sophisticated statistical methods; second, my PhD program had me swamped.
Fortunately, fate intervened during a class raffle where each student was assigned a paper to review. I drew exactly this one: “Improved protein structure prediction using potentials from deep learning.” It took me three days to read, but finally I understood the technology behind this landmark advance.
The technology was called a neural network, represented on page 2 of the article. I redrew it for my homework — and the rest, as they say, followed from there.