Being global leaders in artificial intelligence (AI) is strategic to India, Nilekani said, in a recent conversation at a conference organised by the All India Management Association, in Delhi. Even as large language models (LLMs) are becoming commoditised, India should not take its eye off the real national objective, which is to find ways to apply AI to benefit the entire population. That is the area in which India has an opportunity to lead the world, says Nilekani. Here are some takeaways:
DPI (digital public infrastructure) and AI are complimentary. DPI has laid the foundation in terms of large systems and data, which are required for AI applications. Therefore, AI will be built on top of DPI, and DPI, in turn, will get turbocharged by AI. It is a very synergistic relationship.
What DeepSeek—the Chinese rival to OpenAI’s GPT and ChatGPT that developed by Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co—has done is that it has demonstrated that one does not need billions of dollars to create good large language models.
That is a big breakthrough, and an important example of how the barriers to entry of creating these AI models are significantly coming down. With the Government of India’s AI mission, expect to see some very significant models coming from India in the next one year.
While a billion dollars might not be viable for India, if it can be delivered at $50 million, “sure, why not?” LLMs will be commoditised, with many more vendors in the market, and while India can actually build the models, “we should not take our eye off the ball, which is how is this useful to society? What is it going to do that makes a difference?”
At $50 million, building of core models is now at a quality and cost point that many private companies will be able to afford it.
The challenge with AI is how one makes it work at population scale, and in a way that the end-user access is affordable to everyone. Tasks such as inferencing and reasoning and so on are very compute intensive and can be expensive.
If we want a farmer on Odisha to tap an AI agri app and get a response in her native tongue to improve his productivity, can that be done at one rupee per transaction? “That a whole different ballgame. And that that’s where we have to go.”
Three examples of areas where AI could have a multiplier effect
Consider the massive diversity of languages and dialects in India. Perhaps the only way to bring AI to everyone in this context is to make voice communications with the internet effective. That would dramatically expand access—it takes the challenge of literacy out of the picture because a user can just speak in her native tongue.
AI4Bharat, a not-for-profit effort at IIT Madras, supported by both Nilekani and the government of India, has built the world’s largest database of Indic languages that is now being used by everyone else to train their models. That is similar to how DPI is a core infrastructure.
Second, in education, if one can improve literacy and numeracy of children that will have a big impact. EkStep, another non-profit funded by Nilekani and his wife Rohini, has developed tech-based solutions to perform micro diagnostics of where exactly children are with their learning outcomes at any given stage. This is being rolled out in three states. Third, in agriculture, Nilekani is involved in developing an open agri network “to bring all agri knowledge at your fingertips with reasoning and inference capability.”
The Government of India has a project called Vistaar (Virtually Integrated System To Access Agricultural Resources), which builds on the same infrastructure. Many of the states are interested in joining in.
“So, if you can have every Indian communicate effortlessly, if every child can learn better with AI, and if a farmer can improve his earnings, good enough.”
Also read: Is India’s talent pool ready for India Inc’s AI requirements?
Population scale applications
The early population scale applications will probably come from the government. For example, the various benefit plans under the PM Kisan initiative, have already begun to tap AI. The government taking the lead ensures we know how to do this at scale affordably. That is an important tenet of DPI, and UPI is a very strong example of that—17 billion transactions a month, 400 million active users, 50 million merchants, “and you can make a one rupee payment.”
India will apply that same thinking to AI. For example, NPCI has implemented an Indian language voice command for a payment so a user can speak to the phone to make a money transfer.
Further, under the AI million, the government is ensuring competitive price discovery of GPU compute, making those costs more attractive, and supporting LLM strategies of private stakeholders. Therefore, the government may even “pioneer” early adoption, but the private sector will strongly follow.
AI in helping India’s energy transition: Digital Energy Grid
All over the world, electricity is a growth market because consumption is going up, be it in cities or in specific uses such as AI data centres. The next 20 years are being seen as the age of electricity, the demand for which is expected to rise six-fold.
Distribution is changing from a centralised power generating unit to distributed rooftop solar electricity. This will bring forth millions of households that could have excess electricity, which they can store and even sell back to the grid. “[The] very person in the world who has all this at home can become a producer, consumer, buyer, seller and storer of electricity.”
Coordinating this massive network of consumers and sellers will require a digital energy grid on top of which one can apply AI to every decision.
Also read: Human intelligence must guide artificial intelligence
AI and jobs
“I see it as an opportunity, but then I tend to be optimistic about everything.”
Only a few jobs will be totally taken out [by AI]. There will be many jobs where a lot of the tasks will be taken out, but not all the tasks. So, humans will still have work to do.
AI will help humans to become much more productive and effective in their jobs. This could be a “net positive.” And AI will also create new jobs that we didn’t anticipate. In India, there aren’t enough teachers, doctors, nurses or enough of any skilled resource.
If AI can amplify the ability for people so that more people can learn, more people can get healthcare more, more farmers can get better information, then it’s a net positive. From an ideological point of view as well India’s task ahead is to create inclusion at scale. “So AI also should do inclusion.”
Humans in the world of AI
“The bigger question to ask ourselves is what do people do in the world of AI?”
Human skills of empathy, motivation, leadership, collaboration and so on will become even more valuable and our institutions have to build those capabilities in people. With all the AI in the world, if we fail to get people to work together, not much will be achieved.
Second is first-principle thinking—being able to step back and go to first principles and analyse something that again, AI may not be able to do effectively.
Then there is sheer human creativity. In a world of AI, one must discover ways to be useful, relevant, and satisfied with life.