There have been several theories floating around the internet claiming that India has arrived late to the AI party. Even with the alleged late arrival, most of the AI development in India is focused around building use cases of AI, rather than the core technology, by adopting AI models that have already been developed by the western countries.
These claims, however, may not be completely true.
To begin with, when one looks at the curriculum, premier institutions in India, such as the IITs, have been heavily focused on the theoretical aspects of AI. Many of the prominent contributions in the field of AI have also been made by professors from these institutes that have been the bedrock of innovation for several decades.
But where are we lacking? Anirbit Mukherjee, assistant professor at department of computer science, University of Manchester, said that India should focus more on AI theory. “If even a quarter of the mathematical talent in India were to get into AI theory, it would cause a tectonic shift,” he said in conversation on LinkedIn.
He believes that India’s core theory communities in mathematics, statistics, and physics departments are not actually interested in getting into AI theory. “It’s freaking cool – and more Indians should be doing it,” he added.
‘Honourable mentions’
Nikhil Malhotra, the chief innovation officer of Maker’s Lab, which is building Project Indus, said, “Most LLMs produced in India are built on top of the already-available LLMs. They cannot be called fundamental research or foundational LLMs.” In another comment, he wrote, “Who has challenged the original algorithm? While transformers are a great piece, they have flaws in terms of compute and carbon,” reiterating that most of the research in India is done on fine-tuned models.
Sourav Das, researcher at IIIT Kalyani also weighed in with his thoughts. “How many of them have made an algorithm, theory, or model from scratch,” questioned Das, saying that everything is available on the internet and the researchers are just exploiting the resources. “There is no invention in India, just reusing the things that are already there,” adding that all the fine-tuning is just getting “honourable mentions”.
A lot of AI development currently is being driven by young developers who are building AI models on top of existing ones such as LLaMA and Mistral, but nothing concrete has come up yet. Though there are initiatives such as Ola’s Krutrim, Sarvam AI, Tech Mahindra’s Project Indus, and BharatGPT that are focused on building models from scratch, a lot of work still needs to be done.
On the other hand, “Issue that the sceptics don’t realise is that there’s not much capital available in India for the youth to take it to the next level,” rued Sreekanth Sreedharan.
This issue was also highlighted by several others in the conversation talking about how a lot of investors are not interested in investing in deep research, but just application-based startups that will mint money easier. “India can’t compete in AI foundational research unless the investment behaviour changes,” added Rishabh Bhardwaj.
Similar thoughts were shared by Hakim Hacid, executive director and acting chief researcher at Technology Innovation Institute (TII). “You need a lot of funding to sustain open source and we believe that not everyone will be able to do it,” Hacid told AIM.
Innovation Requires an Entire Ecosystem
Several researchers from IITs point out that the institutes have mastered the art of publishing papers, however only a miniscule amount of such research is actually fundamental. It might be true that we don’t actually need more research on LLMs, but research on something that replaces the current paradigm of AI research.
Some experts argue that there is a need for a push from the industry, along with the government, for fundamental research in AI and focusing on AI theory. “The students need incentives [such as placements, internships, and media coverage] to solve difficult problems,” said Abhishek Gupta.
To build the ecosystem, it is necessary to change the curriculum, while also building an ecosystem which supports groundbreaking research in the field, and not just incentivising AI wrappers.
The recent QS World University Rankings currently feature 72 universities recognised for providing top-notch data science and AI courses. Among these are four Indian institutions that made it to the top 50 list, namely, IIT Bombay (30), IIT Kanpur (36), IIT Kharagpur (44), and IISc (45). Additionally, IIT Guwahati was part of the top 72 universities.
These premier Indian institutions’ inclusion in the rankings may suggest a trend of higher education institutions increasingly integrating data science and AI into their academic offerings.
To skyrocket this, Amit Sheth, the chair and founding director of the Artificial Intelligence Institute at the University of Southern Carolina (AIISC), has been continuously working with Indian academic institutions to drive research in the country. He has proposed Ekagrid, a private research university with an ambition to be ranked among the top in the world and contribute to India’s research-driven ecosystem as Stanford and UC-Berkely have done for Silicon Valley.
This team includes experts from 11 of the 25 top universities in the world. The project is still in its initial stages and is looking to raise funds. “The Prime Minister was very prompt and quick to understand the need for this project and provide actionable guidance,” he told AIM.
“There needs to be a lot more investment in AI research, which is still very low,” Sheth said. “But India is still doing great despite less funding.”
The post Why India Needs to Build its Focus on AI Theory appeared first on Analytics India Magazine.
