AI has been making waves all over the world, but there are still people in India who aren’t sold on its capabilities.
Physics Professor HC Verma, whose books Concepts of Physics are a part of the preparation of IIT-JEE for most Indian engineering students, has said that he doesn’t believe that AI can replace human ingenuity. He also said that AI would struggle to answer the Physics questions that he sets.
“I challenge AI to solve the questions I set in exams,” he said in an interview to a YouTube channel. “Lets see if AI can come up with their solutions. AI is powerful, but it can only operate on available data and information,” he added.
HC Verma also said that while AI also can’t have original thoughts. “A human being can come up with original thoughts and ideas, but AI can’t come up with novel ideas and innovations. It’ll keep regurgitating old information. Only human beings have the ability to create something from scratch,” he said.
HC Verma is one of the most recognizable names in Physics in India. A professor at IIT Kanpur, his books are a staple for the preparation for the IIT entrance exam. He’d retired in 2017 after a 38-year career at IIT Kanpur. Since retiring, HC Verma has made several media appearances and spoken to YouTube channels.
But Verma’s latest comments might not find a lot of cachet with the AI community. Several AI models have attempted exams such as the IIT-JEE to varying levels of success. All the way back in October 2023, two researchers had administered the IIT-JEE exam to ChatGPT and discovered that while it wouldn’t make the cut for a seat at the prestigious institutions, it could achieve a score at the 80-90 percentile, putting it ahead of 80 percent of human students who took the test.
And AI has improved dramatically since then. OpenAI’s latest o3 model scored 96 percent on Competition Math, which is a math test that is a feeder for the US Math Olympiad team. Also, o3 scored 87 percent on test that had PhD level questions, and even managed to score around 25 percent on Research Math, which has such hard questions that it takes the world’s best mathematicians and Fields Medal winners days to solve. As such, it’s highly likely that AI systems will be able to solve IIT-JEE level problems in the not too distant future.
Verma, though, might be right about AI being unable to create original ideas. While there are plenty of examples of AI performing sophisticated tasks, like diagnosing diseases and solving Mckinsey case studies, AI doesn’t seem to have yet come up with something that can be called truly original. But these are still early days in AI, and if this pace of development keeps up, it might not be long before AI systems — through AGI and ASI — end up creating original research as well.