Advertisment

Data, chips, and India’s AI mission

The Rs-10,372-crore initiative seeks to position the country as a global AI powerhouse, enhancing its capacity for research and innovation.

author-image
VoicenData Bureau
New Update
image

Policy

The Rs-10,372-crore initiative seeks to position the country as a global AI powerhouse, enhancing its capacity for research and innovation.

Advertisment

Ever since the advent of ChatGPT in November 2022, the world has never been the same. Every industry and way of life has been touched by Artificial Intelligence (AI), a field that had until then remained a matter for researchers, techies and tech journalists to discuss.

Fast forward to today, and AI has emerged as an all-pervasive field that involves data, policies, governance, education, civil society and geopolitics—all under one umbrella. It is perhaps because of this that the Rs 10,372-crore India AI Mission, announced on 7 March, has proved to be a seminal moment.

Before we delve into the Mission and what it could potentially achieve, it is important to understand why nations around the world are looking to establish their own AI policies and strategies. AI today is feeding upon what we call the currency of the modern world—data. The latter is not just proprietary and sensitive—it is what populous nations such as India have as leverage over smaller economies. To safeguard this, though, cyber security strategies are just not enough.

Advertisment

AI has emerged as an all-pervasive field that involves data, policies, governance, education, civil society and geopolitics—all under one umbrella.

To leverage the data, global corporations and governments today need computing power, which in the modern-day AI world is supplied through graphic processing units (GPUs). The latter was once only known for producing ultra-high fidelity graphics resolutions on computers and consoles while playing games. Today, GPUs have emerged as the most efficient processors of vast troves of data. Companies that operate data centres, such as Mumbai-based Hiranandani Group’s Yotta Data Services, are procuring such GPUs en masse to offer paying clients a platform to get access to AI processing capability.

GPUs are perennially short-supplied today, and the availability of high-quality, structured and anonymised data is also slim at hand. It is because of this that the India AI Mission makes considerable sense and stands a chance to make India a significant nation to look at when it comes to the development of AI.

Advertisment

What does the mission offer?

To begin with, over the next five years, the Government of India will spend Rs 10,372 crore or nearly USD 1.2 billion to promote open-source AI datasets, create computing infrastructure with 10,000 GPUs, and more.

There are seven key breakdown areas in the AI Mission, which will be implemented by IndiaAI, a business division under the Ministry of Electronics and Information Technology’s (MeitY) Digital India Corporation. The seven key areas include compute capacity, innovation centres, datasets platforms, application development initiatives, future skill set, startup financing, and safe and trusted AI

Advertisment

Each of these seven subdivisions of the mission will get collective funding from the total corpus. While an exact framework for the business and a breakdown of total funding for each is still being made by IndiaAI within MeitY, at least two senior industry officials stated that the first ‘compute capacity’ vertical is likely to get at least 60% of the corpus since this is what will cost the most.

The compute capacity vertical, as per MeitY’s release, will “build a high-end scalable AI computing ecosystem to cater to the increasing demands from India’s rapidly expanding AI startups and research ecosystem.”

“The ecosystem will comprise AI compute infrastructure of 10,000 or more GPUs, built through a public-private partnership. Further, an AI marketplace will be designed to offer AI as a service and pre-trained models to AI innovators. It will act as a one-stop solution for resources critical for AI innovation,” it further stated.

Advertisment

Innovation centres, meanwhile, will be responsible for the development of India’s Large Language Models (LLMs) and domain-specific foundational models that could be applied in specific sectors such as healthcare, market regulation and more. The datasets platform will create a unified database, which will be open-sourced and made available to researchers, academia, startups and enterprises as required. This database will include non-personal, anonymised data from across all 22 official Indic languages.

image

The app development initiative, meanwhile, looks to boost startups working on specific problem statements. This will be promoted through the Mission’s framework. This initiative will “focus on developing, scaling and promoting the adoption of impactful AI solutions with potential for catalysing large-scale socio-economic transformation.”

Advertisment

Finally, the FutureSkills initiative will seek to establish an AI curriculum across all levels of senior education in India, from undergraduate to research. There will also be an as-yet-undefined startup financing part of the overall mission’s corpus that will be allocated for eligible ventures, which will likely not see overlaps with the Rs 1 lakh crore deep-tech fund for sunrise sectors that the Union Finance Minister Nirmala Sitharaman spoke about during the interim budget.

Each of these developments will tie up with India’s efforts to create a globally trending demand—responsible AI. The latter will see the Mission lay down guide rails for specific governance frameworks associated with building AI projects—and could be merged with India’s discussions globally so far at Bletchley Park in the UK, and the Global Partnership for Artificial Intelligence (GPAI) Summit, New Delhi.

Is this corpus enough?

Advertisment

Most industry stakeholders believe that the corpus of the Mission is enough to at least begin with—further revisions, as per demand, will always be on the cards. The costs seem to match, too, based on the present market rates of GPUs made by Nvidia, the brand that is supplying almost all of the world’s requisite GPUs.

For reference, the Nvidia ‘Blackwell’ GPUs, unveiled recently at Nvidia GTC 2024, have been touted to cost around USD 35,000-40,000 per unit. At such a scale, the Centre’s outlay to produce just the GPUs would be around USD 400 million or 40% of the entire corpus. Further budgets could be allocated depending on maintenance and requirements, as well as investments made by private sector enterprises.

The next most important factor lies within Bhashini, MeitY’s Indic language datasets platform. The latter already has functioning libraries that are available for researchers, and this unit, coupled with a special office catering only to this need, will likely be set up to handle all the data. There will also likely be the need for security checks and audits to verify the non-personalisation of data despite it being a government resource.

India’s AI mission could successfully replicate the US model of funding applied research to establish the education system as a frontier innovation base.

This will take care of two of the most important aspects of AI development. In research and academia, media reports have regularly flagged the lack of financial resources at eminent universities and institutes as being the biggest reason for the lack of applied research in the field of AI in India.

With the success model of the US being evident in terms of funding applied research at colleges and subsequently promoting them through venture capital funding and go-to marketplaces, India’s own AI mission could successfully replicate this, and establish the nation’s education system as a frontier innovation base.

Early signs of this progress can be seen as well. For instance, IIT Madras-backed BharatGPT group, as well as Vizzhy’s Hanooman LLMs promise to offer India domain-specific AI models that could be trained on locally relevant datasets. This will help take on biases and also promote responsible and explainable AI results—factors that will be crucial for enterprise adoption of AI products and services.

The present corpus, in its present shape, has enough room for most such ventures. How the real-world implementation of the Mission and its subsequent framework takes place, however, is yet to be seen. All of this is likely to further develop after the next government is formed, where AI is expected to take a front seat of importance.

feedbackvnd@cybermedia.co.in

 

Advertisment