AI Alliance NCR: India doesn’t have the capacity nor talent to build foundational models, says Lightspeed’s Hemant Mohapatra

(L-R) Tanuj Bhojwani, Head, people + ai and Hemant Mohapatra, Partner, Lightspeed India

India at present doesn’t have the capacity and talent to build foundational artificial intelligence (AI) models from the grounds up, though it will become one of the key countries to develop use cases of AI going forward, industry experts said at the Moneycontrol and CNBC-TV18 AI Alliance NCR Chapter in Gurugram.

Hemant Mohapatra, Partner at Lightspeed India said that while he would love the country to wage the wars where they can win, there seems to be the question of capacity and talent when it comes to building foundational models for AI.

Foundational models are a form of machine learning model used for building generative AI tools, and is capable of performing a wide variety of general tasks such as understanding language, generating text and images, and conversing in natural language.

“I do not think we currently have the right amount of highly trained talent to win certain kinds of wars. If they come back altogether from wherever they exist, we’ll have the talent. Then we’ll only need to worry about the capacity like GPU clusters, CPU clusters regulation and policy,” he said.

Mohapatra added that the war for building foundational model is extremely intense globally at present. Even to train a world class foundation model could require upwards of hundreds and millions of dollars, and that layer is getting commoditised very quickly.

“The layer already has diminishing returns. To train a model on $120 million wouldn’t mean it is 20 percent better than one which used $100 million. I think that war is run by folks who can raise that much capital, let them build that,” he said.

Tanuj Bhojwani, Head, people + ai believes that India will become one of the frontrunners when it comes using and building population-scale use cases of AI.

“AI and DPI (digital public infrastructure) is about what super power can we give to the average citizen,” he said.

Bhojwani added, “My conviction is that India has all the problems in the world, how can we not use AI? By January 2031, India will be the country that will use AI the most.”

Moneycontrol had earlier reported that despite technology product and IT services companies training thousands of employees in AI skills over the past year, India is in the midst of a talent shortage in core AI products and services.

According to data from specialised staffing firm Xpheno, the active pool of senior AI engineers who are actually building core AI products and services in India is less than 2,000. Core AI products and services include AI-based apps, tools and platforms, to name a few.

The total active accessible Indian talent with the minimum required skills, experience and exposure to core AI is a little under 21,000, Xpheno said, referring to the segment of upcoming AI engineers.

This is quite a contrast to the 650,000-700,000 people that tech industry body Nasscom claimed are trained on AI across top-tier tech companies.