As Deepseek and ChatGpt Surge, is Delhi behind?

Two years after ChatGpt took the world by storm, China’s Deepseek has caused a ripple of the tech industry by disrupting the costs of developing generative artificial intelligence applications.

But as global races for AI hegemony get hot, India is lagging behind, especially when creating its own basic language models that will be used to drive things like chatbots. It seems to have happened.

The government claims that homemade equivalents to Deepseek are not far away. We provide startups, universities and researchers with thousands of high-end chips needed for development within 10 months.

The gusts of the global AI leader have also recently spoken about India’s capabilities.

This month, Open Eye CEO Sam Altman said India should play a leading role in the AI ​​revolution after initially rejecting it. The country is now the second largest market in Openai with users.

Others like Microsoft have committed $3 billion (£2.4 billion) to cloud and AI infrastructure. Jensen Huang of Nvidia also spoke about India’s “unparalleled” technical talent as the key to unlocking future possibilities.

With 200 startups working on generation AI, entrepreneurs are well underway.

But despite having important factors for success, India risks falling behind without fundamental structural modifications to education, research and state policies, experts say.

China and the US already have a “4-5-year head start”, investing heavily in research and academia, developing AI for military applications, law enforcement and now large-scale language models tech analyst Prasantroy told the BBC.

Although it is in the top five globally in Stanford’s AI Vibrancy Index, ranked in metrics such as patents, funding, policy, research, and more, India is still behind two superpowers in many key areas.

China and the US were granted 60% and 20% of the total world AI patents between 2010 and 2022, respectively. India was less than half percent.

Indian AI startups received a small portion of the private investments made by US and Chinese companies in 2023.

Meanwhile, India’s state-funded AI mission is worth a trivial $1 billion compared to the incredible $500 billion allocated by the US to Stargate. This is a plan to build a large-scale AI infrastructure in the US. By 2030.

Deepseek’s success demonstrates that AI models can be built on old and inexpensive chips – India can take comfort from a lack of “patients” or long-term capital from either industry or government You can, says founder Jaspreet Bindra of consulting to build AI literacy in organizations.

“Even though I’ve heard about Deepseek developing a $5.6 million model, there was even more capital behind it.”

The lack of high quality India-specific datasets required to train AI models in regional languages ​​such as Hindi, Marathi and Tamil is another issue, especially given the diversity of Indian languages.

But for all of that issue, India is far beyond the weight of talent. 15% of the world’s AI workers come from the country.

But, as Stanford’s AI talent immigration research shows, the problem is that more and more people are choosing to leave the country.

This is because “basic AI innovations usually come from deep R&D in universities and corporate labs,” says Bindra.

And while India does not have a supportive research environment, there are few deep technology breakthroughs born from the academic and corporate sectors.

The huge success of India’s payment revolution was due to strong cooperation between the government and the industrial industry. Similar models need to be replicated for AI push, he says.

Unified Payment Interface (UPI), a digital payment system developed by government organizations, revolutionizes digital payments in India, allowing millions of people to trade by clicking a button or scanning a QR code. It’s now.

Bengaluru’s $20 billion outsourcing industry is home to millions of coders, ideally at the forefront of Indian AI ambitions. However, IT companies have never actually shifted their focus from cheap service-based tasks to developing basic consumer AI technologies.

“That’s a huge gap they left to fill the startup,” Roy says.

However, he is not sure if the startup and government missions can lift this quickly enough. He added that the 10-month timeline set by Minster is a knee-headed response to Deepseek’s sudden appearance.

“I don’t think India will be able to produce at least some Deepseek for the next few years,” he added. That’s a view that many others share.

However, India can continue to build and tailor applications to existing open source platforms such as Deepseek “to leap our own AI advances.”

However, in the long term, the development of basic models is important for having strategic autonomy in the sector and reducing import dependence and threat of sanctions, experts say.

India also needs to increase its computing power or hardware infrastructure to run such models. This means manufacturing semiconductors.

Much of this has to fall into place before the gap between the US and China becomes significantly narrower.