India doesn't need hand-down versions of AI products: Intel's Viswanathan

The aspirations of India and the Global South are no less than those of other countries when it comes to artificial intelligence (AI) development across all layers of hardware and software, Santhosh Viswanathan, vice president and managing director (VP & MD) of Intel India, said on Tuesday.

 

“We are an aspirational society. I am not saying that such places need not have a low-cost alternative. Price is a function of value, not the other way around. The Global South, and especially India, does not require ‘hand-down’ versions of the latest AI products,” Viswanathan told Business Standard on the sidelines of the ongoing AI Impact Summit.

 

India requires AI products that adapt to societal needs and adopt a frugal approach to using high-end, high-cost technologies, including frontier models, he said.

 

“But it does not mean that we don't expect the same bar of quality or the same level of products that the rest of the world demands, right? So, I think that balance has to be brought in,” he said.

 

Going forward, the journey of AI the world over will spread across various functions, and will no longer be as homogeneous as it was, as developments begin to happen in inference, agents, and application-building, Viswanathan said.

 

Intel’s capabilities, especially around newer products such as Core Ultra Three, deliver the power, performance, and usage levels needed to help frontier models achieve more, he said.

 

While advanced chips and graphics processing units (GPUs) are widely used around the world, companies are also realising that these high-end hardware are not the default for running advanced AI models and inferencing, Viswanathan said.

 

“(Companies) are slowly putting their heads around that it's not like one size fits all. Like, a GPU doesn't mean it's the only thing. The GPU, CPU (central processing unit), and NPU (neural processing unit) work together to make AI possible. It is a combination of these three that’s going to create more efficiency for us,” he added.

 

Countries such as India will also need to rethink AI infrastructure and can no longer afford to focus solely on developing the latest frontier models and expect to lead the AI race, the Intel India MD said.

 

“We have to take an approach that is practical and real for our society, which says yes to leveraging the large (AI) models but also building sovereign AI modules, and that's what I think the government is doing,” he concluded.