By: Zach Zhang

AI, though talented at language subjects, is still struggling with math problems. This is unexpected, given computers in the past were really good at dealing with numbers, exceeding humans at the task. These computers were programmed and followed specific rules, yielding very powerful, yet very subject specific computers.

However, a short while ago, a new approach arose— a neural network modeled off of a human brain. Instead of being programmed with a set of rules, it learns and predicts words, like a human would. This approach makes AI good at writing but lacking proper computation skills.

Khan Academy’s Khanmigo uses AI to help teach subjects. But instead of using AI to solve math problems, it uses a separate calculator program to compute the numbers. Chat GPT has a similar approach to this. With large division and multiplication, a calculator is doing the work.

The famous Watson computer, developed by David Ferrucci, has beaten the best human players at Jeopardy. AI chatbots excel at writing tasks, but struggle with mathematical problem-solving.

On the flip side, these mistakes can turn out to be a wonderful way to instruct students. A high school teacher would often divide his classes into small groups of students, and let them compare the chatbot’s answers to theirs.

AI chatbots, being great at languages, are not programmed on a set of rules, so they frequently trip up on simple math problems.

Sources:
Steve Lohr, A.I. Can Write Poetry, but It Still Struggles with Math, 23, July, 2024
https://www.nytimes.com/2024/07/23/technology/ai-chatbots-chatgpt-math.html