‘Study math if you want to understand AI’: IISc Bangalore professor explains how to learn artificial intelligence

Professor Prathosh AP believes you should learn mathematics before you try to understand AI. Speaking on The Ground Truth podcast by Vizuara, the IISc Bangalore professor highlighted his thoughts on artificial intelligence.

Prathosh says remembering and reproducing information no longer carries the same value. He believes AI has made it easier to implement ideas, helping researchers build more freely. Yet, a deep understanding of AI still requires serious study and strong mathematical foundations.

He suggests a learning path spread across three semesters. Students should study mathematics first, followed by the foundations of machine learning. They can then approach the latest methods with a better understanding.

“Even though you have so many AI tools, the right kind of mentorship at the right points in your life matters a lot,” he said.

He compares using AI with driving without knowing how a car engine works. People can use prompts without understanding the technology underneath. However, contributing new algorithms requires deeper learning.

“You can drive a car without being an automobile engineer. You don't understand how an internal combustion engine works, but you can still drive,” he said.

“If you just want to use prompts and use AI for something, fair enough. There is no judgement there. But if you really want to contribute to algorithm building, there is no alternative to studying,” he added.

Prathosh also questions the usefulness of mathematical proofs that humans cannot understand or verify. He believes discovery matters beyond obtaining solutions.

Ultimately, he wants artificial intelligence to help solve major human problems, including cancer. Without such benefits, he considers the effort little more than a game.

“I tell my students that if you want to understand AI, it's a three-semester journey. Study math in the first semester. Study foundational ML in the second semester. Then you can understand the state-of-the-art methods,” he added.

Mathematics as a language

The profession has a simple message for learners: learn two languages. One should help them communicate. The other should help them think

His advice connects communication with clear reasoning. He describes mathematics as a precisely-designed language, rather than merely a collection of calculations. For him, its value lies in helping people express ideas clearly and discover unexpected possibilities.

“If you think about it, mathematics is, at the end of the day, a language. It is nothing more and nothing less,” he said.

Prathosh explains that researchers may start with a small problem and develop a method. Expressing that method mathematically can reveal uses far beyond the original problem.

A solution first intended for one task could help address many others. He says this ability to apply ideas more widely comes from mathematical precision.

He also stresses that reaching an answer is only part of learning. The steps between a question and its solution matter equally. Working through those steps can reveal other methods and deepen understanding. Simply receiving an answer does not offer the same experience.

His own interest in mathematics grew through strong mentorship during engineering college. A mathematics professor taught him throughout eight semesters and introduced him to IISc and IITs.

Those lessons helped him appreciate mathematics and its engineering applications. He continues to value good mentors, even when powerful AI tools are available.

For engineering students, he connects clear thinking with building useful things in the real world. He encourages them to start with a problem and work backwards. Research should also consider how ideas might become technology that people actually use. However, he acknowledges that not every research paper can immediately become a product.