Sniggering at India’s AI efforts may be mistimed
Right now, everyone’s panicking about DeepSeek because it came out of China. But that’s the wrong way to look at it.
On Thursday, when technology minister Ashwini Vaishnaw said six homegrown Artificial Intelligence (AI) models are in the works and will go live by the end of this year, there were a few sniggers all around. What everyone knows is that the Americans started the AI race with expensive models most people know of as ChatGPT and Gemini. And now the China-built DeepSeek has overturned all assumptions that has the world stunned. So, how is India to get a foot in the door? To start at the beginning will help us understand this better. Yishan Wong, CEO of Terraformation and former CEO of Reddit, does the best job of sifting through it all.
Right now, everyone’s panicking about DeepSeek because it came out of China. But that’s the wrong way to look at it. Fact is, DeepSeek could’ve been invented anywhere. Let’s say, an American lab very unlike OpenAI didn’t have a big budget to buy the most expensive equipment. The team had to get creative. So, they found smarter ways to train their system, figured out a bunch of clever shortcuts, and — voilà! — they ended up with something as good as the big players who spent a fortune.
Normally, people would say, ‘Wow, what a clever idea’ and start trying it themselves. It would be a great week for technology, and no one would freak out. But because this came from China, everyone suddenly thinks it’s a crisis. People imagine the Chinese have discovered some secret weapon that the rest of the world can’t compete with. There’s this underlying fear that China’s government is now sitting on something incredibly powerful. But that’s not what’s happening here.
The geeks who worked on DeepSeek are solving the same kinds of problems as the American researchers are. And even the Indians are at work on it (we’ll come to that in a bit). In fact, the Chinese have even shared their work with the world by publishing a paper and releasing their model for anyone to use. This isn’t some grand geopolitical shift. It’s just a group of smart people saying, ‘Hey, we figured out a cool trick. Here’s how we did it—feel free to try it yourself.’
This is why Wong calls this AI’s Google moment. Back in 2004, when the number of searches on Google were going through the roof, the engineers didn’t bust the bank buying expensive supercomputers. Instead, they connected thousands of cheap computers to work together as one giant system. This allowed them to handle massive amounts of data for far less money. They didn’t just keep their clever approach a secret. They eventually shared it with the world by publishing papers that explained exactly how they built their system. This allowed others to copy their methods and make improvements.
DeepSeek is similar—but even faster. They shared their research and model as soon as they launched. It’s moving the field of AI forward. But it’s not like China has gained some unreachable advantage. In technology parlance, this is the Open Source way of doing things.
Does this stand to scrutiny with the Open Source community? Venkatesh Hariharan is one among India’s earliest Open Source evangelists. He said, “My prediction is that the Open Source model, powered by collaboration, community and shared ownership of knowledge will become a very powerful alternative to proprietary, closed source AI models like that of the extremely mis-named OpenAI.”
So how do Vaishnaw’s prediction fit in here? Hariharan points to Viswam.ai that hasn’t been talked of much in the mainstream media. The roots of the project go back to 2004 when the Swecha Foundation started work to develop an operating system in Telugu. And since 2020, it has been housed at IIT Hyderabad to develop AI models in Indian languages. Like DeepSeek, Viswam is Open Source. In much the same way, there is Sarvam, an AI model being built out of Bengaluru. Founded in 2023 by Vivek Raghavan and Pratyush Kumar who are familiar names in the tech community, the company has Nandan Nilekani batting for it. Until now, the venture capital community has pumped $53 million into it and the early outcomes sound promising.
When looked at from all these perspectives, the sniggers behind Vaishnaw’s back sound misplaced and perhaps a little too early.
On Thursday, when technology minister Ashwini Vaishnaw said six homegrown Artificial Intelligence (AI) models are in the works and will go live by the end of this year, there were a few sniggers all around. What everyone knows is that the Americans started the AI race with expensive models most people know of as ChatGPT and Gemini. And now the China-built DeepSeek has overturned all assumptions that has the world stunned. So, how is India to get a foot in the door? To start at the beginning will help us understand this better. Yishan Wong, CEO of Terraformation and former CEO of Reddit, does the best job of sifting through it all.
Right now, everyone’s panicking about DeepSeek because it came out of China. But that’s the wrong way to look at it. Fact is, DeepSeek could’ve been invented anywhere. Let’s say, an American lab very unlike OpenAI didn’t have a big budget to buy the most expensive equipment. The team had to get creative. So, they found smarter ways to train their system, figured out a bunch of clever shortcuts, and — voilà! — they ended up with something as good as the big players who spent a fortune.
Normally, people would say, ‘Wow, what a clever idea’ and start trying it themselves. It would be a great week for technology, and no one would freak out. But because this came from China, everyone suddenly thinks it’s a crisis. People imagine the Chinese have discovered some secret weapon that the rest of the world can’t compete with. There’s this underlying fear that China’s government is now sitting on something incredibly powerful. But that’s not what’s happening here.
The geeks who worked on DeepSeek are solving the same kinds of problems as the American researchers are. And even the Indians are at work on it (we’ll come to that in a bit). In fact, the Chinese have even shared their work with the world by publishing a paper and releasing their model for anyone to use. This isn’t some grand geopolitical shift. It’s just a group of smart people saying, ‘Hey, we figured out a cool trick. Here’s how we did it—feel free to try it yourself.’
This is why Wong calls this AI’s Google moment. Back in 2004, when the number of searches on Google were going through the roof, the engineers didn’t bust the bank buying expensive supercomputers. Instead, they connected thousands of cheap computers to work together as one giant system. This allowed them to handle massive amounts of data for far less money. They didn’t just keep their clever approach a secret. They eventually shared it with the world by publishing papers that explained exactly how they built their system. This allowed others to copy their methods and make improvements.
DeepSeek is similar—but even faster. They shared their research and model as soon as they launched. It’s moving the field of AI forward. But it’s not like China has gained some unreachable advantage. In technology parlance, this is the Open Source way of doing things.
Does this stand to scrutiny with the Open Source community? Venkatesh Hariharan is one among India’s earliest Open Source evangelists. He said, “My prediction is that the Open Source model, powered by collaboration, community and shared ownership of knowledge will become a very powerful alternative to proprietary, closed source AI models like that of the extremely mis-named OpenAI.”
So how do Vaishnaw’s prediction fit in here? Hariharan points to Viswam.ai that hasn’t been talked of much in the mainstream media. The roots of the project go back to 2004 when the Swecha Foundation started work to develop an operating system in Telugu. And since 2020, it has been housed at IIT Hyderabad to develop AI models in Indian languages. Like DeepSeek, Viswam is Open Source. In much the same way, there is Sarvam, an AI model being built out of Bengaluru. Founded in 2023 by Vivek Raghavan and Pratyush Kumar who are familiar names in the tech community, the company has Nandan Nilekani batting for it. Until now, the venture capital community has pumped $53 million into it and the early outcomes sound promising.
When looked at from all these perspectives, the sniggers behind Vaishnaw’s back sound misplaced and perhaps a little too early.
All Access.
One Subscription.
Get 360° coverage—from daily headlines
to 100 year archives.
Archives
HT App & Website
