E-Paper
Centre earmarks ₹5k-cr in AI plan for computing power
The Indian government has earmarked about ₹5,000 crore to provide over 10,000 Graphics Processing Units (GPUs) as part of its ₹10,000 crore-plus IndiaAI Mission, a senior official announced on Thursday.
The Indian government has earmarked about ₹5,000 crore to provide over 10,000 Graphics Processing Units (GPUs) as part of its ₹10,000 crore-plus IndiaAI Mission, a senior official announced on Thursday.
Speaking at the Global IndiaAI Summit, Abhishek Singh, additional secretary in the ministry of electronics and information technology (Meity), said the funds are part of the ₹10,371.92 crore approved by the cabinet in March for the mission. Of this, ₹2,000 crore will be meant to support Indian start-ups in developing indigenous AI-based solutions.
“A lot of investments will come from investors, VCs [venture capitalists] and all, and that will add to the pool of funds available for the AI start-ups,” said Singh, who is responsible for IndiaAI.
The government plans to subsidise access to computing power rather than own the infrastructure outright. Singh explained this approach aims to help researchers and startups access compute at lower costs, enabling more experimentation and scaling.
“It is not that we will buy the chipsets and we will build a public-owned, public-run infrastructure. We will want the investment to come from the private players but part of the cost of access to that compute will be subsidised by the government [for researchers],” he said.
The Indian government, Singh said, felt that Indian start-ups and researchers who are building AI-based solutions, applications and models to solve problems in healthcare, agriculture, education or social sector should have access to compute at a subsidised cost.
To give access to the compute infrastructure, Singh said that it will largely be trust-based where end users (start-ups, researchers) will choose what kind of infrastructure they want but the government will introduce some checks and balances to prevent misuse.
Tenders to procure compute capacity will be floated soon, Singh said.
The government wants to focus on the bottlenecks in the AI ecosystem, especially those related to compute, datasets and skilling, Singh said.
Singh said that to catch up with countries like China and the US, India needs to invest in research ecosystem and compute resources. While India is the “tech garage” of the world and has the highest penetration of IT talent, deep tech and AI need research acumen that goes beyond coding skills, he said. “We have the potential, we have the skillset, and hopefully with the investment that is happening, we will be able to catch up in due course,” he said.
Regarding datasets, Singh said that the government wants to augment the synthetic datasets used by start-ups with real datasets in Indian languages so that Indian foundational model can be built.
In a separate panel, Kavita Bhatia, COO of India and group coordinator in the ministry, announced plans to soon release a request for proposals for the IndiaAI Datasets Platform. This platform will provide researchers and startups access to non-personal data from government ministries.
The IDP is one of the seven pillars of the IndiaAI Mission that will give start-ups and researchers access to non-personal data. It will function as a central repository of datasets from government ministries and departments for better decision-making strategies and AI applications.
“They will be able to use this data to train the indigenous models,” Bhatia said. The platform will also offer data annotation, labeling, and curation services.
On this platform, data annotation, data labelling and data curation services will also be made available, she said, and added that compute functionality will also be integrated. Models from Bhashini, Meity’s AI model used for translation into Indian languages, will also be offered on the platform so that when solutions are built in English, they can also be easily offered in other languages, she said.
