Keep Learning Like the AI Models Do
After being a TA at OMSCS for nearly four years, this is what I wish every ML learner could hear
About three years ago, I graduated from what’s probably the most well-known online computer science master’s program in the U.S. , OMSCS (Georgia Tech Online Master of Science in Computer Science). Some people call it life-changing. Others call it a lot of Python and work. I call it three years of late nights, imposter syndrome, and a surprising amount of coffee. I finished the machine learning track with a 4.0, which sounds fancier than it felt.
For the past five years, I’ve also been a TA for two courses. Mostly grading. It’s been a chance to stay close to the material and, hopefully, be a small part of someone else’s learning experience. I never thought of myself as a strict grader, but I’ve always tried to be fair. If something doesn’t quite add up, I feel it’s worth pointing out. Not to be difficult, but because that’s how I learned too. Honest feedback, even when it’s not easy to hear, helped me the most when I was on the other side.
This year, though, I’ve found myself pausing more often. I’ve caught myself thinking, maybe it’d be easier to just give out the points and move on. Less stress. Quieter weekends and more free time with my daughter. And I’ll admit, it’s tempting.
A big part of it is the regrade requests. There have been more of them lately. They tend to show up as posts on the class channel. Short, direct messages asking for points back. Not a lot of questions, not much discussion. Just a request to change the grade so they can move forward. I totally get it. Life is busy. Everyone’s juggling jobs, families, other classes. Sometimes, just getting through the week feels like enough.
I’ve been thinking about that a lot, how hard it is to make space for learning when life is already full. I remember feeling that way too. Some weeks, just submitting something before the deadline felt like a victory. But looking back, the moments that stayed with me weren’t about grades. They were the times I got curious. When I dug into something just to understand it better or tuning my models just make it more accurate, even if it wasn’t required. Those moments made the long nights feel worth it. This is a reason I keep blogging about those ML topics. Honestly, I love those struggling moments.
And I guess that’s where I’ve landed. If we want to build machine learning models that really work, maybe we need to learn the same way they do. A model doesn’t get tired or frustrated when it’s wrong. It doesn’t argue with the loss function or ask for points back. It just keeps going, little by little, getting closer to something useful. Something true.
I think that’s what I’m trying to hold onto as a learner myself. Staying curious. Trying again, even when it’s hard. Learning from feedback, without taking it personally. Bit by bit, getting closer.
At least, that’s the hope.
