By: Serena She
The rise of artificial intelligence has highlighted its fluency in literary arts, as its skills in crafting thoughtful, personal statements and poetic articles are undeniable. However, it still lags behind in problem-solving abilities, as its struggle with both simple and complex math problems evidences.
A.I.’s deficiency in the mathematical field is a stark contrast to the former purpose of computers—to compute, hence the name. Early computers and calculators were programmed to calculate arithmetic by rigidly following rules and structures. Based on a neural network modeled after the human brain, A.I. does not follow a fixed structure but, rather, learns and analyzes data as a human would. It rarely provides one specific answer, explaining its strength in language over math.
“Everything it generates is plausible,” notes Rayid Ghani, a Distinguished Career Professor in the Machine Learning Department and the Heinz College of Information Systems and Public Policy at Carnegie Mellon University. “And that’s exactly what it’s designed to do, is generate plausible things, rather than factually correct things, because it doesn’t know the difference between factually correct and plausible.”
Math can be tricky because it is not a single question with open-ended answers but a problem that involves many steps encompassing different mathematical skill sets. A.I. struggles, in particular, to solve word problems, which users commonly ask, because they require logic and reasoning in addition to arithmetic.
While math continues to be a researched area, computer scientists are developing A.I. with training similar to that of students—with textbooks and tests. Doing so can familiarize it with problems that people may encounter in the same textbooks. Open AI’s Chat GPT-4o evidently demonstrates A.I.’s improvement, which has an accuracy rate in mathematical thinking averaging 64%, compared to a mere 58% from the previous version, as reported by the New York Times. A.I. may not currently be adequate in math, but it undeniably has a high potential.
Image Credit by Tara Winstead