By Sujit Bhar
A small Chinese company has developed a pretty impressive Artificial Intelligence (AI) large language model, DeepSeek, that ticks all the boxes and has been said to give Open AI’s Chat GPT a run for its money. The after-effect of this launch was so big that giant technology company Nvidia at one point lost nearly $600 billion from its market cap. Nvidia has been recovering since, but only just. As if a cat has been set among pigeons, other tech giants are also growing weak in the knees.
In a massively interconnected world, it was only expected that a rival large language model would emerge, and a Chinese involvement was only natural. So what is the big story in this? Two critical issues need to be considered while drawing any comparison.
The first is the question of funding. OpenAI, which developed Chat GPT’s several stages, was promoted to the hilt by Microsoft. It was a real greenfield project and the giant software company had funded it up to its gills, to an extent of over $1 billion. Open AI founder Sam Altman had once boasted that it was expensive because of the basic research and training models involved and the time consumed. He had pooh-poohed suggestions of other, cheaper models evolving from Asian countries.
Meanwhile, in 2024, Liang Wenfeng released DeepSeek-V2, starting a price war even within China. It has been claimed that Wenfeng had used no more than $6 million in funding to create and train his model. Nobody in the West believed it, especially because they said the Chinese company did not have the technical wherewithal—read advanced processor—to force-develop this model, especially when these processors were “banned” in China.
The argument—and this is the second critical issue—was that while not officially “banned,” the Chinese government has effectively restricted the sale of Nvidia’s most advanced AI processors to Chinese customers due to US export controls, essentially preventing them from buying the latest Nvidia chips for critical applications like AI development. It has been said that China is encouraging Chinese companies to use domestically produced alternatives like those from Huawei and Cambricon through “window guidance” which is a form of strong unofficial pressure.
Now it is being said that Wenfeng and his company developed DeepSeek-V2 and conducted his training with old Nvidia chips, for which they wrote fresh code and installed domestically developed patches that made it work better. It is a matter of time before companies such as Huawei and Cambricon catch up with the market and develop their own matching chips. This news has sent shivers down the spines of western tech giants.
Surely, there are limitations to DeepSeek, especially through government controls which force it to withdraw from contentious political and such other uses that are looked down upon or similarly “banned” by the government. However, there is no doubting the fact that its potential has been proved.
The fight is on, and it is being fought out in the open. Meanwhile, India is feeling left out. The worldwide perception is that India has managed to grow into a tech giant and has enough engineers and grey cells to handle large problem issues that involve AI. Frankly speaking, it has. However, the best Indian brains are aching to leave its shores, land in an ecosystem where it is more important to innovate than to prove his/her Sanatan Dharma identity. This was boldly shown to the world in the fact that while a Chinese large language model was bringing the West to its knees, India was promoting a Mahakumbh Mela and trying to handle the ignominy of over a hundred illegal Indian immigrants being sent back by the Donald Trump administration in chains.
To be frank, India’s priorities have gone haywire. Top-tech honchos, including Infosys founder NR Narayana Murthy and Larsen & Toubro Chairman SN Subrahmanyan have called for 70-hour and 90-hour work-weeks, respectively—it is another matter that Tesla boss Elon Musk has come out with a crazy demand for a 120-hour week—but there has been little talk about why India, a tech giant, has yet to develop any all-pervading software system such as Windows or Facebook or even the design of an advanced chip. We have to remember that Vinod Dham, who is known as the Father of the Pentium Chip for his contribution to the development of Intel’s Pentium microprocessor, gained fame and renown only after he migrated to the US from India and then joined Intel.
That Satya Nadella is the chairman and CEO of Microsoft and Sundar Pichai is the big boss at Google has been stated over and over again, but they, too, carried their IIT backgrounds and training to US shores before gaining prominence and fame. Had they stayed back in India, they probably would have created another Infosys and become billionaires, but the world would not have seen innovation as they have promoted in the US.
In fact, DeepSeek has created a fear of sorts among Indian companies and government offices. A directive from the Union finance ministry about AI tools like ChatGPT and DeepSeek restricts the use of AI tools in offices. This is the proverbial cat among the pigeons. While a recent survey has pointed out that 50 percent of internet users in India use AI platforms, with ChatGPT being the most popular, the ministry’s ban is for office work that its employees carry out. The directive warns that these chatbots could pose a threat to government documents and data security.
“It has been determined that AI tools and AI apps (such as ChatGPT, DeepSeek, etc) in the office computers and devices pose risks for confidentiality of (government) data and documents,” says the instruction.
Interestingly, ChatGPT is not banned in India. In fact, most promotional work today has been handed over to this language model, and while the quality of work has plummeted, nobody is complaining, because it is cheap.
And the big story for India lies in this small word: cheap. Sure enough, Indian generic drug manufacturers have created a world market with its knock-offs of patented drugs, but basic, hardcore research is a matter of perseverance, time, and of course money. We cannot do away with these factors just because we believe in a quick buck.
Also, the entire ecosystem in the country has been geared to encourage quick-fix “jugaad” type solutions. There have been several articles and speeches, even from top government functionaries, praising this “jugaad”. This, accompanied by our recent excessive stress on the past has, seemingly, blinded us when we try to look ahead.
Maybe we will get our own language model in the near future, but let us not tom-tom our frugality, but highlight our technical acumen instead. Someday, maybe, some Indian software will finally rule the world.