Ashoke Agarrwal: Building AI India

By Ashoke Agarrwal

 

Ashoke AgarrwalOne of India’s achievements of the past two decades has been the ongoing building of Digital India. The world acknowledges the excellence of the public digital infrastructure that India has built through Aadhar and UPI.

 

Can India now take a leadership position in the nascent but fast-developing age of Artificial Intelligence (AI)?

 

It is currently fashionable to think of AI mainly in terms of Generative AI- AI that puts together words and images in practical or amusing ways.

 

However, two other forms of AI – Predictive and Prescriptive AI – are likely to prove equally, if not more, valuable and powerful in terms of their impact on human civilisation.

 

While it is already too late for India to strive for a leadership position in Generative AI, it has the potential to be a leader in Predictive and Prescriptive AI.

 

Many experts predict that China will lead in developing cutting-edge AI, given the tremendous amount of data its large and increasingly affluent population generates. And more importantly, because of the unfettered, centralised access that its government has to the data. On the other hand, India, too, has a large, increasingly better-off population. Still, given its noisy democracy, media and civic society, centralised access to data by fiat is neigh impossible.

 

However, the State already has access to a tremendous amount of data in various silos – commercial and personal. Think of company and tax filing, court filings and judgements, police records, land records, birth and death records, applications and petitions, legislative records, etc.

 

Many of the above records are probably musty paper sitting in cardboard boxes in damp storage rooms. More importantly, they belong to bureaucratic fiefdoms that will loathe to give them up. And most importantly, bringing them together into an integrated State-owned dataset threatens the privacy and rights of individuals.

 

The solution would be to set up a central institution through an Act of Parliament – say, National Data Centre (NDC). The NDC would have the power to acquire data sets from all listed bureaucracies – Central and State. The NDC would then be responsible for digitizing, cleaning, and integrating the datasets. However, before the integrated dataset is available for public and private organizations to use as training sets to develop Predictive and Prescriptive AI, NDC will have the onus to anonymize the dataset to protect the rights and privacy of individuals and other entities. The NDC could also run an outreach program that would facilitate individuals and other entities to share their data in prescribed formats with the NDC assurance of anonymization before use. NDC would pay the individual and other entities for making this data available to them.

 

With suitable communication campaigns and incentives, voluntary direct information sharing by individuals and other entities could become a significant data flow. Integrating this data with the data available from Government sources could lead to NDC sitting on the world’s most potent dataset.

 

NDC would generate revenues and move towards self-sufficiency by making integrated, anonymized datasets available to various users.

 

While the world-class, sustainable, and responsible datasets that a properly mandated and empowered organization like the NSD could create are necessary to build AI leadership, more is needed.

 

The two other ingredients are computer power and talent. India has the talent and, if adequately rewarded, will stay and work in India.

 

It is the area of computing power that India needs another significant initiative. An AI developer or lab needs a lot of “compute”- essentially large, advanced clusters of graphical processing units (GPUs like Nvidia A100). While all three big cloud-computing companies – Amazon, Google and Microsoft – the so-called hyper scalers- offer such facilities, all these servers sit outside India. Quite rightly, given privacy and security concerns, India will not allow the kind of datasets outlined above to reside on servers outside India.

 

Therefore, the other significant initiative that the Government of India needs to undertake in its drive to build AI India is to persuade hyper scalers to locate advanced cloud computing servers within India. One way to do it is to offer them incentives, but a better, more sustainable way would be to convince them about the seriousness of India’s push in Predictive and Prescriptive AI.

 

If India is going to fulfil its potential of being a developed economy within the next few decades, it will need to build leadership in some of the critical technologies of the future. Given its unmatched potential to integrate and harness robust datasets and the ready pool of talent it commands, Predictive and Prescriptive AI could make India a leader in the coming Age of AI.