‘Infra is supercritical for innovation, research’

4 min read . Updated: 14 Mar 2023, 10:46 PM IST Abhijit Ahaskar

In an interview, Vishal Dhupar, managing director, Asia-South for Nvidia, delves into the role of generative AI, the value of computing infrastructure, and how well-placed India is to exploit these disruptive AI technologies.

NEW DELHI : ChatGPT has put generative AI on every company’s roadmap. Training such large language models also requires large amount of computing power and companies like Nvidia, which drive some of the most powerful supercomputers, are set to play a key role. In an interview, Vishal Dhupar, managing director, Asia-South for Nvidia, delves into the role of generative AI, the value of computing infrastructure, and how well-placed India is to exploit these disruptive AI technologies. Edited excerpts:

What role is Nvidia playing in the adoption of generative AI?

A lot of people think about Nvidia as a chip company, and it’s true that we make chips . But we are the only semiconductor company that builds platforms. We build platforms by understanding a domain, creating the tools, libraries, algorithms, which are specifically designed to solve that problem, and translating all that into a framework, so that the community can then apply it in a quicker and more efficient, and effective way. They can use portions of the Nvidia stack depending on competency or they can use the full stack.

If you look at it from generative AI, there are companies like Mid Journey, Stable Diffusion (Stability AI), etc., which are basically taking the entire stack and applying it because the purpose is to build solutions quickly so that they can basically generate revenues. Then there are others like Microsoft who want to basically use certain tools from us and integrate them into their tools to accelerate their performance.

What role can India Stack or AI mission play in driving faster and more meaningful adoption of generative AI applications?

They are extremely critical. At one layer, all of us are speaking English. But that is like sub-10 % of the total population. And then, we are basically dealing with different segments of the pyramid. Society is so different and we speak multiple languages. So, this whole initiative by the government of India called the national language translation mission (NLTM), is so critical because you are trying to get access to the data, convert it into datasets, apply algorithms, convert that into models, and some of them are extremely large models; and we can already see that multiple bodies that serve the citizens are applying it.

Is India as serious about AI technologies as firms in North America and China?

One thing that is very evident is that India is very clear that AI is transformative and it can solve its challenges. On one hand, that’s why you have clear-cut documentation and strategy paper. All these initiatives whether it’s the National Supercomputing Mission; whether it’s NLTM; whether it’s India Stack are definitions that the work is in progress. We’re seeing a lot of activity where academia in conjunction with pivoted medical institutes, AIIMS, and Tata Memorial are working hand-in-hand to solve critical problems using technology. One thing where we can do a lot more is research. We believe if more infrastructure gets applied, we will see more breakthroughs even in research work.Supercomputers will play a key role in driving adoption and training large AI models. Do you think India is slipping on this front?

Infrastructure is supercritical, whether it is in the physical world or the technology world. It is a critical component of making innovations, breakthroughs, research, etc. I think the initiative by the Indian government under the National Supercomputing Mission is profound. It needs to be accelerated and applied more. We have already seen how much better our weather forecast has become thanks to the infrastructure that’s available. I think a lot of speed needs to be added to it. Another thing that is really taking place simultaneously is the indigenous effort being put into it. I think this whole thing about self-sustainability is supercritical to our nation. And we have seen it in different initiatives, whether it is from the semiconductor industry perspective or building computers. Even if the volumes are not as big as what some other countries consume, we’re becoming self-reliant and those are again good signatures of what is to come.

Has India stepped up the adoption of the metaverse?

The 3D world is upon us. People are working on it. Even here in our country, there are many companies that are talking about inserting digital avatars. On the industrial side, there is also a lot of excitement. A lot of factories will be created in virtual worlds; they will be simulated how they operate in the real world before being implemented in the real world. So that’s a really large opportunity. Simultaneously, those that exist will look to get more automated, cut down on mundane activities, and automate to reduce inefficiencies. We can already see that taking place. So, metaverse is only going to gallop with generative AI because now you can apply it to your text, image, videos, and coding to everything else and that’s a compounding impact.Do you see India emerging as a semiconductor design and manufacturing hub anytime soon?

I think the intent and our needs are so substantial that it is a very worthy cause. And hopefully, we’ll translate that into actual success. In addition to serving ourselves, we’ll be able to serve the larger world. Talent is at our core and so many companies basically have benefited from the talent our nation has in designing. So the design aspect is available in plenty. Now we need to basically also bring the other side of the coin, which is the manufacturing part of it. With the right determination, talent, and investments we should reach that goal.

What can we expect from the AI developer conference, GTC 2023, slated for 20-23 March?

We are talking about large language models, diffusion, and reinforcement learning. There are many who want to understand how they can apply it and what they need for it. We will have a lot of sessions at GTC dedicated to this. There will be a lot of conversation about how you apply generative AI to the content universe. Health is going to be an important part. Covid has awakened everyone to what is required to get the right drugs by understanding the structure of proteins. Also, 3D needs to be interoperable and there needs to be a standard for it. Just like how web2 has HTTP. How will you simulate and collaborate simultaneously if the 3D applications don’t work together? Nvidia is spending a lot of energy on that aspect.