Leaders flag AI's potential to transform education at India AI summit

Speaking on the second day of the India AI Impact Summit in Delhi, Chaudhary said, “I have the confidence that it (AI) is going to emerge as a very strong realisation among all the students of India. That is how AI is going to impact how students engage with learning. And, my hope – and I’m saying this as the minister of education for school – is that once you are able to continuously assess the journey of every student, maybe, someday you can do away with examinations, because examinations are a static assessment at one point. "

And, he added, "With AI, you can have a more intuitive understanding of every child who is special, every child who is unique.”

Chaudhary’s remarks came during a wider discussion among policymakers, venture capital leaders and academic researchers on India’s AI ambitions – from building sovereign large language models to rethinking classrooms and scaling up elite research talent.

Rajan Anandan, MD, PeakXV - a venture capital and growth investing firm, framed the challenge in global terms, arguing that India must aim beyond incremental gains.

“For India to really build an AI leader in the world… we have to lead the next generation. We have to build world models. We have to find new architectures,” he said, pointing to a sharp talent deficit compared to leading economies.

“The US and the UK have 3,000 top-tier AI researchers each. India has 200. How do we get to 5,000 to 10,000 world-class researchers? If we do that, we win,” said Rajan.

The conversation then turned to foundational models and technological self-reliance. Vibhu Mittal, CTO, InflectionAI, warned that countries that fail to build their own large language models risk long-term disadvantage.

“In education, there is a statement that says if you don't learn to read, you won't be able to read in order to learn. If we don't build our own LLMs, the inability to use AI will cause any country to fall further and further behind. AI is an amplifier. It can make a single person be as productive as a larger, much larger number of people. It can help us in some sense fill gaps that exist in every organisation and so I think it's very critical that India have its own LLM and its own models,” he said.

“There are other reasons to do this as well. Our knowledge, culture, sensitivities are all unique and I think every country, but especially a country as unique as diverse with a tradition as rich as ours, needs to have a model that India has built and India can control,” he added.

Beyond national capability, Mittal argued that AI could transform pedagogy itself, shifting the focus away from memorisation.

“We will no longer focus on rote learning. We will no longer focus on memorisation. We will focus more on analysis. We will focus more on pushing the boundaries,” he said.

Mittal suggested that even books, unchanged in format for centuries, may soon evolve into interactive systems.

“We have had books for 500 years, and they have not really fundamentally changed. I think in the next few years, we will see books being replaced by essentially the equivalent of AI models. Books that will allow you to talk to the book, talk to the author of the book, argue with the author of the book,” he said.

“Wouldn't you like to be able to read the Mahabharat from the viewpoint of either the author or from the viewpoint of Draupadi or Vidur, you want to be able to try and get descriptions, understanding, analysis of all of these things in ways that have never been possible,” he added.

Mittal even said that there may not be a need for the way Indian Institutes of Technology (IITs) teach.

“I think in some ways education will change to try and drive more curiosity, more drive, increase motivation because of the inherent joy in learning things. So from my point of view… let me be controversial and say maybe we will not need IIT's teaching the way they do today,” he said.

While industry leaders painted an ambitious picture, academics stressed that implementation would require careful design, accessibility and safeguards.

Manindra Agrawal from IIT Kanpur flagged India's digital realities and emphasising that AI integration must account for this.

“Integration of AI tools… using smartphones and popular applications like WhatsApp is going to be crucial. Every content, every model that is to be used for training the student should be available to all students. It's going to be a mammoth effort because it's going to require development of low-bandwidth models which can even work offline at times,” he said.

If such infrastructure is built, delivery could become universal, he argued.

“With all of these in place, which I'm sure will be there in the next year or two, the delivery of content – provided we can develop the right set of small models for them – will be universal and that is going to be the key in ensuring that quality education reaches to every single student of the country,” he said.

Sunita Sarawagi from IIT Bombay highlighted another critical dimension: control and oversight in educational settings.

“A very important challenge I see is the steerability of AI models. We want AI models to be regulated by teachers when they are used by students for augmenting their learning. I don’t see the current prompt-based steering of AI models as adequate for such important deployment, where we want to have human oversight in how AI is used for learning, so as not to degrade the cognitive development of students,” she said.