Data Sovereignty and Cloud Infrastructure: Where Does India Stand Currently?

Can computing infrastructure truly uphold sovereignty if part of the data resides on cloud systems managed by entities that may not comply with Indian laws, questioned Yotta Data Services CEO and Co-Founder Sunil Gupta, during a session at the NVIDIA AI Summit 2024 on October 24, 2024.

“[India has] local people who are actually making motherboards and government of India is promoting that. That is a part of sovereignty. Now, can you make it in your own servers and then build your own cloud?…Why we are talking about data localization and data sovereignty? The point is, even if your data is in India, you may still be working that data on cloud infrastructure which may be operated by somebody else, who may still not necessarily be following the laws of the land, who will be conflicted when the law of the land is conflicted with the law of their country. So, are you sovereign on your cloud? Of course, then it applies to data sets, where your priorities are in terms of developing AI,” said Gupta.

He said that sovereignty should have its own layer that considers complete independence as well as a strategic alliance with the right countries.

AI tools must be democratized despite risks

EkStep Foundation’s Chief Technical Officer, Pramod Varma, pushed for the democratization of AI tools, learning, differentiation, scaling, etc. This, he said, was despite perceived risks because “not doing so is the biggest risk in India.” He gave the example of how India jumped at the Unified Payments Interface (UPI) opportunity whereas the USA delayed and eventually released FedNow.“The risk versus the benefit is distinctly different for us compared to the West. The US launched FedNow, which is somewhat like UPI. Their value [benefit] was low and the friction [risk] was high. For us, we were waiting for UPI. India just went on and in such a rampage, in one sense, taking in UPI. We loved it and we used it. So where do you draw the line? That depend on the context, depending on the risk versus benefits,” said Varma.

Machine unlearning and model editing likely to impact the future of AI

When asked about possible factors that will impact the future of AI, Mayank Vatsa, Professor at IIT Jodhpur, talked about machine unlearning that works on teaching a model to forget a concept or an identity or a class. For example, you may ask an AI model to look at the image of a fruit and forget the apple but remember the oranges.

Another practice is model editing, which involves teaching a model to modify itself when it understands a concept wrong.

India is the use-case capital: According to Varma, India has many opportunities for AI applications owing to the variety of issues India deals with in terms of education, agriculture, etc.

Advertisements

“We have tons of problems to solve. I think that allows us to really apply [AI] in a frugal way; to be able to actually change the lives of the people. It could be healthcare, it could be agriculture. We have many efforts in the digitization where AI is being applied,” he said.

Also Read:

_{Images are for reference only.Images and contents gathered automatic from google or 3rd party sources.All rights on the images and contents are with their legal original owners.}