AI cracking the IIT entrance test is a giant leap for machine-kind

AI has leapt beyond mimicking patterns and can now be used in fields where reasoning is crucial

- SRIKANTH NADHAMUNI & ARUN SEVAKULE

When ChatGPT first appeared in late 2022, it could charm us with words but stumbled with numbers. It could draft a sonnet in the style of Shakespeare but might guess wildly if asked to multiply 17 by 24. Two years later, the tables have turned. Last month, Google's Gemini Deep Think and an experimental OpenAI model won gold medals at the International Mathematical Olympiad, solving five of six problems and matching the scores of the world's brightest teenage prodigies. Days later, Google's Gemini 2.5 Pro topped India's famously tough IIT Joint Entrance Examination.

These headlines are striking not just because machines can now ace exams designed to pick the top one in a thousand students, but because AI has acquired the ability to reason.

From language to logic: The first generation of large language models were, at heart, articulate parrots. They could produce convincing sentences because they had been trained on vast libraries of text. But they lacked a systematic way of thinking. Confronted with problems that demanded multiple steps in mathematics, science or logic, they crumbled.

Two innovations changed that. The first was reinforcement learning. Instead of training AI only to mimic patterns in data, researchers began to reward it for following coherent chains of reasoning. Models learned not just to provide the right answer, but to show their work step-by-step. This nudged them away from guessing and towards disciplined thought.