The upcoming IndiaAI Datasets platform, which is a key pillar for the Rs 10,372-crore IndiaAI Mission, will provide anonymous, non-personal data to help businesses train their AI models. Government departments will have the opportunity to monetise the datasets through the platform. Abhishek Singh, additional secretary in the ministry of electronics and information technology (MeitY) and the CEO of IndiaAI Mission, talks to Jatin Grover about India’s potential in AI, the US AI chips restriction impact, how the country is contributing to the US growth story, and global AI governance. Excerpts:
In the AI landscape, where do you think India would be in the next two years?
The whole world expects India to be in the lead and we have the potential. According to Stanford AI Index, India is Number 1 in AI skill penetration. We have the talent, but need the datasets platform, which the India AI mission is going to provide. With the strength of our researchers, we should be able to move forward.
Have we taken up the matter with the US to relax AI chips restrictions? Is that a concern?
We are in talks with the US and will find a way out. The ministry of external affairs (MEA) is taking it up and it is also being discussed at various levels. But it is not going to be a major bottleneck for our AI plans. They (the US) are more wary of the way China is moving and less so of India. One reason is that India is currently not in a position to compete with the US, which is much ahead. But the US does see India as a trusted partner and its potential. The big-tech companies are very much dependent on Indian software developers. India is actually contributing to the US growth story.
What will the first version of the IndiaAI Datasets platform look like?
Datasets will ultimately be populated by multiple departments. Version 1 will have the basic core datasets and basic functionalities of the architecture. Over a period of time, more datasets will come from various departments and more functionalities will be added. The first version will be a minimum but viable product with core, prime and basic functionalities.
What kind of fee will companies have to pay to access the datasets platform?
Every department will decide how they are going to give the dataset. Some may make it free, and others may charge a licence fee. It will be at the discretion of the departments that are populating the platform.
Are we developing the National Data Management Office?
The India AI Kosh — the data portal being built — will ultimately take the form of a digital data management office. A team of 30-40 people, including techies as well as management and data experts, is currently working on building the platform. They will constitute the data management office.
Global AI models are already there. What is the market potential of our own foundational models?
The market potential in India will be huge because Indian foundational models will primarily be trained on Indian datasets and in local languages. For example, the internet will become more voice-enabled, and there will be many people who will be wanting to access services by using a voice command. Voice-based large language models (LLMs) will be a huge market. We have 900 million people who are online, but the balance 500 million may also want to use an online service.
Are we looking at exporting the AI models as well?
Once something state-of-the-art is built, it should be available for the whole world.
AI computing is a critical requirement. Any updates on developing our own graphic processing units (GPUs)?
The India Semiconductor Mission is working towards that. We are building our own chips, setting up fabs, and have design-linked incentive schemes. Hopefully, we could also design a GPU-kind chip.
Is there a cap on startup applications looking to access the AI compute infrastructure set up with over 18,000 GPUs?
The portal is ready and will be live soon. Startups can apply for the same. The approval will be near real-time. There will be a basic hygiene check for what they are taking it (compute) for.
On AI governance, do you think there is a need for common standards globally?
AI by nature is a global technology. There has to be a global framework for AI governance. The UN high level advisory body has given a direction to that and different countries are looking at it. Every country tries to put forward their own priorities, but then there is a common minimum benchmark, which every country will have to agree to in order to make a global governance framework successful.
