New Delhi Investment：Indian ChatGPT？ Krutrim launch and Meity directive show AI future in India is stuck

Last year, in June, when Sam Altman was in India, he was asked: what would it take for an Indian company to build something like ChatGPT. His answer to that question went viral, because in a way it was dismissive of what Indians could do with AI. Altman had said: Indians can try to build AI like ChatGPT, but they will fail. Nearly a year later, there are attempts ongoing to build India-specific Large Language Models (LLMs)New Delhi Investment. But the recent happenings in this space, most notably in the last 10 days, show that the efforts to build something like ChatGPT are indeed stuck.

Two bits in particular standout: Krutrim, a tool launched by Ola, and the recent directives issued by the Ministry of Electronics and Information Technology (Meity) on Friday.

Let’s talk about Krutrim first. Touted as India’s own LLMs and the home-grown AI tool like ChatGPT, Krutrim opened to public last month. Within days, the excitement surrounding the made-in-India AI tool fizzled out. People started interacting with it, and they soon realised that the AI tool is half-baked. It gave people a lot of inaccurate information, it was hallucinating a lot — as in, it was making up information — and in one instance it even blurted out that it had been created and developed by OpenAI. This led many to speculate that Krutrim is nothing but probably a fancy wrapper on top of something like ChatGPT. The allegations were serious enough for Ola to respond, which it did through a tweet. The company said: "We investigated the issue and found the root cause to be a data leakage issue from one of the open-source datasets used in our LLM fine-tuning."

Irrespective, its launch and the subsequent brouhaha over its output showed that Krutrim was definitely not ready for prime time, and it also showed that if this was what an Indian company could do best in the LLM space, then probably we are not going to compete with the likes of ChatGPT anytime soon.

Having said that, let’s be a little kind to Krutrim and put things in perspective. Because just days after Krutrim failed the public test, Google’s Gemini too flunked it. And how! Beyond its trouble with images of historical people, in India, Google Gemini caused uproar after it blurted out some opinions and views on Prime Minister Narender Modi as facts.

Now, LLMs are fallible. These are still early days in the AI space and if a company like Google can’t get them right, we should not be too harsh on smaller companies and startups like Ola. However, after the Krutrim and Gemini fiasco, the Indian government seems to be moving with that classic motto of someone who has a hammer: every problem is a nail.

This is also probably what led the Indian government to come out with a strict advisory that is likely to impact the way Indian companies go about building AI tools like ChatGPT in the country. And if, in a freewheeling environment, the best Indian companies have done so far is something like Krutrim, in a more restrictive space, that might not even be possible.

On Friday, Meity issued its advisory saying, "The use of under-testing, unreliable Artificial Intelligence model(s) (like) LLMs, Generative AI, software(s) or algorithm(s) and its availability to the users on Indian Internet must be done so with explicit permission of the Government of India."

This led to an uproar with many technologists, experts and people associated with the AI industry calling the new advisory detrimental to India’s efforts to create something like ChatGPT.

A few days later, on Monday, Union Minister of State for IT, Rajeev Chandrasekhar, clarifiedChennai Stock. He said this rule primarily targeted large platforms rather than startups. "Advisory is aimed at the Significant platforms and permission seeking from Meity is only for large platforms and will not apply to startups," Chandrasekhar wrote.

But not everyone is convinced. Instead, they believe that Indian startups and companies doing work on AI would move their efforts outside the country, particularly at a time when Dubai is also seemingly investing heavily in creating an AI-ready infrastructure and policy space.

Aravind Srinivas, CEO of Perplexity AI, quoted a news report on new advisory and said, "Bad move by India". He was not alone. Similar views were expressed by Bindu Reddy, CEO of Abacus AI, said: "India just kissed its future goodbyeâ€æ Every company deploying a GenAI model now requires approval from the Indian governmentJaipur Stock. That is, you now need approval for merely deploying a 7b open source model."

Other tech experts also highlighted that the advisory was likely to impact work done on AI tools in India. "This Advisory may be well intended but is an overreachâ€æ It interferes with innovation," Mishi Choudhary tells India Today Tech. "The fact that clarifications have to be issued for an Advisory tells us it was rushed and intended to address merely one issue caused by Google's Gemini. AI needs regulations and guardrails like labelling but not like the current one."

Choudhary is a technology lawyer and online civil rights activist and is also the founder of SFLC.in.

Neil Shah, VP at Counterpoint Research, agrees there is a need to be cautious about AI development. But he also says that it will slow down AI development in India. "The Indian regulators are making sure to prioritise ethics driven AI over economics driven AI which could slow down the overall AI proliferation in India until the AI driven platforms get the output accurate without any misinformation and usable," he said.

Thomas George, president, CyberMedia Research, said that the government will have to balance its concerns with the needs of the industry. "It is imperative that there is a delicate equilibrium between fostering transparency and innovation in the AI sector while simultaneously safeguarding privacy and intellectual property rights," he says.

Coming back to Sam Altman and his challenge to Indian companies and technologists, so far Altman has been kind of proven right. Although it is too early to declare that Indian companies cannot build something like ChatGPT.

There are other LLM efforts being done in India and some of them may even have a larger scale than Krutrim. These things also take time. But it is clear that building a reliable LLM is not something that a company can manage within months, and Indian companies face key challenges — most notable challenges related to talent and infrastructure, such as humongous farms of top GPUs. Just for context: Mark Zuckerberg over at Meta is spending top dollars to buy 350000 H100 Nvidia graphics cards this year. These GPUs will then be used to train AI models.