Abhishek Singh at Idea Exchange: ‘Whether it’s Nvidia, Anthropic, OpenAI or Google, companies are looking at India to hire AI engineers​’

The second is to showcase India’s prowess in AI — the ability of Indian engineers, human resource, IT companies. The real value from AI will come in building agentic AI, integrating AI agents with legacy IT systems and what can be called AI transformation services.

India has been at the forefront of providing IT and IT-enabled services, and people of Indian origin are part of almost every major big tech company globally. Given this advantage, we want to position India as a country where when anyone thinks of AI, they think of India.

Third is to attract the attention of investors towards India’s AI ecosystem, which includes setting up data centres, investments in data centres, AI startups and AI applications. We are expecting to source a lot of investment for companies in India. If Indian talent is able to get funds, then we will be able to create IP also in India.

Soumyarendra Barik: One conversation slowly eroding from public discourse is AI governance and regulation.

Why are you thinking this is going down? In fact, what I feel is that the discussion is not centred only around regulation, safety and security, primarily because some of these issues, which were flagged at Bletchley Park, have been addressed in many ways. That led to the setting up of AI safety institutes in the US, the UK, Japan, Canada and France.

P Vaidyanathan Iyer: Where does India currently stand in the global AI order and what leverage does it have to play a meaningful role in AI governance?

If you look at rankings and various indices, the most prominent is the Stanford AI Index, which places India at number three in the AI Vibrancy Index. We are behind the US and China and I must say that the difference, the delta between where the US and China are and where India is, remains huge. Our prime strength is talent, AI skills, our workforce and our AI startups.

The other advantage we have is infrastructure because AI requires data centres and energy. The third advantage that we have is in building AI applications, as we have traditionally been very strong in developing software and writing code.

The elements where we needed to work were compute, with regard to having our own GPUs (Graphics Processing Units), our design capabilities, fabrication capabilities and frontier AI models. Frontier AI models were invested as part of the IndiaAI Mission because we believe we have the talent and the capability of building our own sovereign LLM, which will be hosted in India. It will be trained in Indian languages, Indian context and cultural heritage and traditions. We have supported around 12 projects in the IndiaAI Mission and a few of them have come out with their models.

On the need to upskill for AI | Our engineers and skilled workforce in the IT sector will have to think of how to compete against $20 bots. We’ll need to use coding agents to become better coders ourselves. It will require a massive amount of reskilling

The model which Sarvam AI has released this week has been ranking very high on various indices, especially related to Indian languages, whether it’s for OCR, reading or speech-to-text or text-to-text. It has ranked much better than even OpenAI and Google’s Gemini.

The only area where we don’t have our own presence right now in the sovereign AI stack, is probably the GPUs, the ability to design our own AI compute and fabricate it in India. But we do believe that we have a very strong design strength in India. We are working towards a DLI scheme which will allow us to tap into this talent and ensure that we are able to design chips which can take AI workloads. Globally, only TSMC in Taiwan is the foundry where all these chips are taped out. We are much behind, at 28 nanometers right now, but we do believe that in three to five years we will be able to start designing our own chips.

Anil Sasi: When Anthropic released Claude Code and Cowork a few weeks ago, it had a significant impact on listed tech companies here. How do you envisage the risks posed by such innovations emerging in other markets, and how prepared are we as we move towards building our own sovereign large-language model?

This is a very real risk and this is what we need to guard against because the ability of Claude Code to write software is surely going to impact entry-level software coders. We have six million people who are employed in this sector and many of them are at basic entry-level. Routine jobs will be done ultimately better by these coding agents. Our engineers and skilled workforce in the IT sector will have to think of how to compete against $20 bots. We will need to use these coding agents to become better coders ourselves. It will require a massive amount of reskilling.

We are upgrading the curriculum in our engineering colleges. Engineers who are coming out from engineering colleges are not coming with just basic coding skills. They are coming with skills that can help them use these coding agents to build agents for specific use cases. It will require skills to be able to work on edge AI systems, because that is what the future is. If we are able to prepare our workforce for that, we’ll be able to survive.

On trust as India’s edge | Many countries trust our digital capabilities. We are regarded as an IT superpower… We do realise that there are countries like China also who might be doing the same thing but our goal is not to compete with China

Soumyarendra Barik: We have discussed the immediate harms of AI, such as deepfakes, misinformation. But beyond these, India’s IT services sector clearly needs to course correct; the Anthropic launch was a strong signal. What scares you most about AI in the long term?

Of course, this whole ability of AI being able to take on the jobs that humans do. For example, take SAP implementation. A task that required large teams 15 years ago can now be done by just three or four engineers. While we talk about skilling, reskilling and upskilling, we also have to think even if people reskill or upskill, will there actually be enough jobs for them?

Second is this whole quest for Artificial General Intelligence (AGI). While we’re investing a lot in building AGI, we’re investing trillions of dollars in frontier AI labs, how will they monetise it? If suppose AGI comes in and we are able to find a cure for cancer or something, maybe that’s something worth aspiring for. But without any such goal, if we are blowing up trillions of dollars, then it’s very difficult to see how it will pay back.

One doesn’t know how sustainable this model is and we have to guard against it. Luckily, for us, the leadership that we have, the direction that we got was to focus on using the technology for bettering the lives of people, ensuring that we are able to make the majority of Indians more efficient and productive by leveraging AI. So my fear, of course, is twofold. One, jobs and second, of course, the investments. How will they pay off?

Siddharth Upasani: While Indian companies are still playing catch-up, much of the global AI investment boom has bypassed them. Could this have long-term implications for India? Second, with reports that OpenAI has disbanded its mission alignment team, which focused on ensuring AGI benefits humanity, does this raise concerns about the direction in which we’re headed?

AGI was never our goal and I never felt that AGI is something that is worth aspiring for. There are many elements in AI, machine learning, deep learning, and natural language processing that have too much value within India, which can help us move forward.

With regard to the AI investment boom having missed India, I think that’s a very short-term assessment. A lot of Indian startups, Indian entrepreneurs, are in fact relocating to India because of the investment climate. Every fund that I speak to, they are earmarking certain quantum of their funds for investing in Indian AI startups. You have already heard about big tech announcements by companies like Google, Microsoft and Amazon, which will invest almost $70 billion in data centre infrastructure. As we move forward, I foresee many such investments coming into the Indian AI ecosystem, not only in infrastructure, energy and compute, but also for Indian startups working on foundation models, application layer and use cases.

Sandeep Singh: You had said that the majority of the skills that we have are at basic coding level. In that case, do you think we are prepared to position ourselves as an AI talent hub for global companies?

Because we have a large number of people who have basic IT skills, this does not mean that we don’t have AI talent that the whole world is scouting for. If you look at the absolute number of AI engineers, we would be far ahead than any other country in the world. Of course, we do have the challenge of upgrading those with basic skills, but we are regarded as the AI talent capital of the world. When many companies are looking at hiring AI engineers, they are looking at India, whether it’s Nvidia, Anthropic, OpenAI or Google. Google DeepMind’s bulk of work happens in India. So, we do position ourselves as a talent capital when it comes to AI and the capability that we have.

Amitabh Sinha: From a government perspective, are there specific focus areas where you would want AI applications to be developed to serve certain national priorities?

Our focus has been on building India’s stack and digital public infrastructure and we believe that using technology for improving governance has yielded a lot of benefits, whether it’s in implementation of programmes like direct benefit transfers or preventing leakages, ensuring that subsidies reach people, access to services becomes easier, and giving identity platforms to people or digital payments.

What we do believe is that with AI, the layer of DPI and the access of DPI multiplies multifold. Look at how Bhashini has enabled services in all languages. Now with voice-enabled capabilities of Bhashini APIs, coupled with GenAI services, we are looking at a scenario where voice-enabled services will be possible. If through voice people can query information by unstructured data input and get their output, it will be transformational because the majority of the governance challenges in the last mile is because of asymmetry in access to information.

Aanchal Magazine: Can stronger privacy rules coexist with innovation in AI?

Definitely. AI is trained on data and you can’t use personal data to train AI applications. You need to allow only anonymised non-personal data for training AI applications…otherwise, if AI models start getting to know about each individual, the risks of misusing the information goes up.

Having a global governance framework becomes very important. Having empowered institutions with the regulatory capacity, the tech capacity to be able to test such things with regard to privacy norms, becomes very important. That’s where summits like this also become very important because it brings everyone together.

Sukalp Sharma: Beyond the IT sector workforce, how challenging would it be to make the legacy workforce in other sectors ready for a future that will have heavy AI use? What is the government’s plan and vision?

AI skills will be required in almost every sector. I’ll give you an example. Last year, under the future skills pillar of the AI Mission, we launched a fellowship programme for AI projects for undergraduate, postgraduate students and research scholars. When we announced it in July, only engineering students were eligible. Within two months, I received queries from five students at AIIMS who were working on AI projects. When we examined their work, it was fully justified. This shows that AI cuts across domains and disciplines.

Students graduating today need basic AI skills. That is why we have worked with the UGC and AICTE to ensure AI, data science and data analytics become part of the curriculum, so graduates are job-ready. At the school level, CBSE has introduced AI from Class V onwards and already has it in Classes 10 and 12. We also launched the Yuva AI programme to provide basic AI skills to everyone.

Harikishan Sharma: Math education is necessary for anyone entering the field of AI. As the CEO of IndiaAI Mission, how do you see the current status of math education in India, particularly in schools? What should be done to improve the quality of math education, given India’s ambitions in AI?

Essential math statistics skills for building AI applications and math skills are required in almost every domain. What we need to do for improving math education is to first ensure sufficient numbers of math teachers and that they are trained with the relevant pedagogy skills, ensuring that the curriculum remains balanced.

I foresee that AI tutors will address the challenges of shortage of quality teachers, especially science teachers, math teachers, not only in the remotest parts of the country but even the best of the schools where the teachers don’t have the ability to personalise their teaching to the ability of the students.

Soumyarendra Barik: Since G20, India has aspired to be the leading voice of the Global South. Yet, we compete with China but we are miles behind in most sectors. Even in AI, China appears to have a significant lead. How can India begin to close this gap?

When we built digital public infrastructure and the India stack, the whole world at the end of G20 was looking at India. India has built the MOSIP platform to provide ID to 28 countries and several countries are replicating our UPI models, Digilocker models. The advantage that we have over other countries is trust.

Many countries trust our digital capabilities. We are regarded as an IT superpower and Indian engineers enjoy a very strong reputation.

We do realise that there are countries like China also who might be doing the same thing, but our goal is not to compete with China. Our goal is to build production solutions that work in India, and then we make it available for the global community. Just like the world embraced India’s DPI at G20, I am sure they will be looking at India’s AI innovations.