Studying Algorithms with ChatGPT
- Chrissy Gonzalez
- Jul 29
- 3 min read
As a Computer Science student at NYU Tandon, I feel like I should be all about generative AI, but the truth is that I’m not sure how best to use it. I went back to school because I wanted to learn CS fundamentals and get better at programming, which so far has required a great deal of plodding effort. I worry that if I give in to using AI, my logic and reasoning skills will never improve. What if, for me, the only way to remember what I’ve learned is to grind it out all semester? I’m wary of taking easy shortcuts. Luckily, school is a great place to learn from others, and my classmates are less hesitant to experiment with generative AI.
The first time I used ChatGPT for learning was Spring 2024. A friend in my Algorithms study group pasted all of our previous problem sets into the chat input and asked it to generate more in the same story-based style. The results were pretty amazing – ChatGPT wrote pages of new questions that sounded different from the originals, but they also made sense and seemed challenging. When we tried to study with the new questions though, we quickly recognized the originals. I remember us saying things like, “Oh, baseball game? That’s the dungeon maze.” It was fun, but it didn’t advance our ability to solve problems much. We had hoped for a whole new set of algorithm problems, but given that ChatGPT is an LLM, it made sense that it simply wrote new storylines. Chatbots can produce such convincing text, it’s easy to forget that they don’t understand what they’re writing about. They’re just stringing highly probable words into sentences.
The second time I tried GenAI for learning was closer to the end of that semester. We’d just covered a tricky subject called dynamic programming. I was struggling with a homework problem that came directly from the textbook and, although I’d found an answer online (I know, I know), I didn’t understand it. I remember feeling a little desperate by the time I turned to ChatGPT for help, but it just repeated what I already found on my own. It did provide some examples that helped me understand a simpler form of the problem, but in this situation, ChatGPT’s real value to me was as an emotional support. Having an additional resource helped me not to panic about being confused.
By the end of Algorithms class, I decided that although ChatGPT had potential as a learning tool, it wasn’t what I needed for help in a problem-solving theory class. I was impressed by how quickly GenAI could produce text-based results with the right structure, so perhaps I just needed a text-shaped task for it to solve? I could ask it to write study guides or outlines, although that might rob me of the chance to synthesize new information by summarizing it on my own. I would say that one semester of experimentation wasn’t conclusive, but the experience of trying to study with ChatGPT did provide some valuable information that will inform my next try. Hopefully someone in my class next semester will have some good ideas.
References
What is an LLM (large language model)? | Cloudflare. (n.d.-d). https://www.cloudflare.com/learning/ai/what-is-large-language-model/
Cormen, T. H., Leiserson, C. E., Rivest, R. L., & Stein, C. (2009). Introduction to algorithms (3rd ed.). MIT Press.