Robots could be one of humanity's 'greatest opportunities': Deepu Talla

I knew some of the startups but I did not know that there are so many startups working on robotics, it was like a breath of fresh air, he says

The new buzz word in the world of artificial intelligence is Physical AI, the coming together of hardware and software on which applications are built. Deepu Talla, vice-president (VP) and general manager (GM) for robotics and edge computing at chipmaker Nvidia, is working to make Physical AI a reality. In an interview on the sidelines of the Nvidia AI Summit in Mumbai, he spoke with Shivani Shinde about Physical AI, robots and more. Edited excerpts:

 

What does it mean when you say that the age of Physical AI has arrived?

 

The demand for robotics has long existed due to factors like an aging population, labour shortages and safety. Robotics today is focused on high-volume and repetitive tasks, such as automotive and semiconductor manufacturing, where the scale supports dedicated robot programming teams. The need for robots and autonomous capabilities has only increased, though progress has lagged due to complexities of the physical world.

 

AI has developed in three waves. The first wave, a decade ago, focused on consumer internet services with around 80-90 per cent accuracy. The second wave, emerging now, brings this approach to enterprise applications, offering significant growth for India’s software industry. This is because enterprises need tailored AI solutions for proprietary needs.

 

The third wave, encompassing physical AI and robotics, is the largest opportunity, with applications across manufacturing, warehouses, cities and agriculture.

 

There are two main technologies that have changed trajectory or reached a tipping point – GenAI (generative artificial intelligence) and digital twins – that address previously unsolvable robotics challenges, especially in accuracy. Technologies and platform like Nvidia’s Omniverse are making physical-world simulation more feasible, enabling robots to perform complex tasks.

 

I believe that the next 10 years are going to be significantly different in terms of how we can solve some of these problems.

 

Does this mean that a future of humanoids is getting closer?

 

Yes. The reason why humanoids are so popular, especially since last year, is the fact that this technology curve has suddenly changed due to GenAI and simulation.

 

When we think of humanoids, we picture robots which vary in form but fundamentally resemble humans. They must navigate, manipulate and most importantly think – a task where GenAI plays a key role.

 

Humanoids represent a vast opportunity, potentially allowing each of us to own multiple robots in the future, whether as endoskeletons, exoskeletons, or other forms. This field could be one of humanity's greatest opportunities.

 

Nvidia’ Omniverse platform allows for simulation of robotics. What is the accuracy that robots can reach?

 

I think we're still very early. However, the fact that we are able to finally take physical things and put them in simulation allows us to make progress orders of magnitude faster.

 

Digital twins now allow us to replicate exact motions and sensors, enabling extensive experimentation and building the infrastructure to train humanoids. It’s a challenging problem to solve.

 

What is your view on the argument that countries with huge populations, like India, are not ready for robots?

 

If there are people employed, then I think these robots can make them more productive. There are many cases where there is actually a labour shortage, even in India. In my interaction with startups and robotics companies here, they are focused on developing robots for tasks like welding which the younger generation wants to avoid. Some jobs, such as heavy lifting, are also hazardous and better suited to robots.

 

Additionally, as medical advancements increase human lifespans, robots could play a greater role in supporting our workforce.

 

You met with robotics startups here; what was your takeaway?

 

I knew some of the startups but I did not know that there are so many startups working on robotics, it was like a breath of fresh air. Some of them have come back from the US, while some of them are trying to address the Indian market. It's quite vibrant. I was actually pleasantly surprised that some of these companies have been working for five to seven years.

 

I think they should really focus on one area and solve the problem because it's such a hard problem to solve with robots. The physical world is unforgiving if you don't do it properly. That's why self-driving cars are still not yet on the roads. I would say they should go deep, focus on something, solve it and then generalise it.

 

The Indian IT services industry is partnering with Nvidia on the Omniverse platform. Can you elaborate?

 

The 90s and 2000’s were the heydays of the IT services industry. The Y2K phenomenon took the industry really up. Then came the mobile revolution and the industry suddenly had to move to the app ecosystem. This shift is evident in their growth trajectory too – it has slowed down.

 

Now is the next tipping point. The global solutions integrators (GSIs) instead of exporting software, now need to talk more about exporting intelligence.

 

When thinking of robotics, people often envision humanoids or industrial arms. At Nvidia robots encompass a broader scope. Every aspect of life, from cities and buildings to warehouses and retail stores, will integrate robotics. Consider a warehouse or factory: Multiple robots, including humanoids, arms and autonomous mobile robots, will coexist with humans. There will be sensors and cameras that will provide real-time situational awareness. There will be two types of AI. Real-time AI that processes immediate data from sensors and cameras. The other is non-real-time AI, which analyses stored metadata to identify trends and areas for improvement.

 

The future of physical AI involves integrating various robot brains, human oversight, and advanced sensors to create a seamless, efficient, and continuously improving ecosystem. The GSIs will integrate these components, creating digital twins of factories and warehouses to optimize operations. This opportunity is 10 times larger.

 

How do Nvidia’s India centres contribute?

 

For robotics we have teams across Bengaluru, Pune and Hyderabad. We have several thousands of engineers. They work across the complete stack. Our teams in India contribute to every platform, product that Nvidia is working on. We don't divide it based on region. We divide based on where's the best talent to do the task.