Opinion: India Isn’t China, We Must Stick To The High Road When Developing Our Own DeepSeek AI

Last Updated:January 28, 2025, 19:43 IST

There’s no doubt that China’s AI app DeepSeek has created ripples by becoming the most downloaded free AI app on Apple App Store. Even Sam Altman, the developer of ChatGPT, has publicly appreciated the Chinese feat

There are still many unanswered questions about how DeepSeek developed its model.

Let’s not get distracted by the headlines. Many are and will sing hosannas to China for levelling the field with US big tech giants or for edging past them on the way to developing one of the most significant technologies of the current era.

There’s no doubt that China’s AI app DeepSeek has created ripples by becoming the most downloaded free AI app on Apple App Store. Even Sam Altman, the developer of ChatGPT, has publicly appreciated the Chinese feat.

There’s no doubt that it has been developed at a fraction of the cost of what was spent on ChatGPT, sending global markets into a tailspin. There’s even talk that China’s AI app could have outperformed its more pedigreed US competitors in its ability to “mimic human reasoning, processing information step-by-step, making it skilled at tackling complex scientific problems”.

But here’s the rub: there are still many unanswered questions about how DeepSeek developed its model. What we know is that DeepSeek leverages older-generation Nvidia H-100 GPUs, which has significantly reduced its cost of production and dramatically increased return on investment.

But is it possible that DeepSeek has withheld information on how many top-flight Nvidia chips it has access to and used to train its models? An article in the Fortune magazine quotes Alexandr Wang, the CEO of AI company Scale AI, who claims to have information that DeepSeek secretly acquired access to a pool of top-of-the-line 50,000 Nvidia H100 GPUs. If this is indeed the case, then DeepSeek’s claims may be the deep fake news of the year.

But as things stand, we have no choice but to believe DeepSeek that its decision to go with the older Nvidia H-100 GPUs was not out of choice but necessity. Yes, necessity borne out of a US semiconductor export restriction meant to deny China a chance to outrace the United States in acquiring AI supremacy. Necessity may be the mother of invention, but is it also the mother of innovation?

An expert writing on the subject observed that “while one might be able to run DeepSeek R1 on a laptop and get it to output a good answer to a tough maths question after, say, an hour, other AI platforms with access to the latest GPU or AI chip technology might be able to evolve faster into performing the same task in seconds”.

As many will tell you, for many business applications of AI, the time it takes a model to produce an output matters. The quicker the output, the better. And it is here, in the speed sweepstakes, where Chinese companies starved through sanctions of the latest, most advanced computing chips may not be able to compete.

Aside from the technological, there are also larger ethical issues that merit attention. There are issues around data sovereignty. Storing data in China will always raise privacy and compliance concerns. DeepSeek’s very broad data-sharing policy with third parties leaves users with little control. And then there’s limited accountability as DeepSeek makes users responsible for outputs and account security.

China may have captured the national headlines, prompting many in India to legitimately ask questions about why India has lagged. We are, after all, the kings of ‘jugaad’. But when (and not if) India builds its own Large Language Model on an ISRO-like shoestring budget, it must take care not to compromise on data ethics and data security.

India isn’t China and therefore it must stay committed to taking the high road, irrespective of the cost.

News opinion Opinion: India Isn’t China, We Must Stick To The High Road When Developing Our Own DeepSeek AI



_{Images are for reference only.Images and contents gathered automatic from google or 3rd party sources.All rights on the images and contents are with their legal original owners.}