When Sam Altman visited India in June of last year, he was asked what it would take for an Indian company to create something similar to ChatGPT. His response to that query went viral since it appeared to downplay the potential applications of artificial intelligence for Indians. Altman had stated: Indians will not succeed in creating AI similar to ChatGPT. Almost a year later, efforts are still underway to construct Large Language Models (LLMs) tailored to India. However, recent developments in this field, particularly in the past 10 days, indicate that efforts to create something akin to ChatGPT are, in fact, at a standstill.
Particularly noteworthy are two items: the tool Krutrim, which Ola introduced, and the recent guidelines published on Friday by the Ministry of Electronics and Information Technology (Meity).
First, let’s discuss Krutrim. Krutrim, a homegrown AI tool similar to ChatGPT, was introduced to the public last month and is hailed as India’s answer to LLMs. The hype surrounding the AI tool that was made in India subsided in a matter of days. As people began interacting with it, they quickly discovered that the AI tool is incomplete. It provided a great deal of false information to people, was frequently hallucinating—that is, making up information—and in one instance, outright stated that OpenAI was the company behind its creation. Many conjectured that Krutrim is nothing more than a sophisticated shell over ChatGPT as a result of this. Because Ola felt the allegations were serious enough, it responded with a tweet.
Whatever the case, Krutrim’s launch and the controversy surrounding its performance demonstrated that it was unready for prime time and, if this was the best an Indian company could do in the LLM space, then it is unlikely that we will be able to compete with ChatGPT or other similar companies anytime soon.
Having said that, let’s put things in perspective and show a little kindness to Krutrim. Because Google’s Gemini also failed the public test shortly after Krutrim did. And how! Beyond its issues with historical personages, Google Gemini sparked controversy in India when it misrepresented some opinions and viewpoints about Prime Minister Narendra Modi as fact.
Government brings down hammer
Thus, LLMs are not perfect. We shouldn’t be too hard on smaller businesses and startups like Ola if a giant like Google can’t get them right in these early stages of the AI industry. But in the wake of the Krutrim and Gemini debacle, the Indian government appears to be operating under the old adage “every problem is a nail”—appropriated from someone who owns a hammer.
This is also most likely the reason behind the Indian government’s recent release of a stringent advisory that will probably have an effect on how Indian businesses proceed with developing ChatGPT-like AI tools in the nation. Furthermore, in a more constrained setting, if the best work that Indian companies have done thus far in a free-form setting is similar to Krutrim, that might not be possible
Meity released an advisory on Friday, stating that the use of under-tested, unreliable AI models—such as generative AI and LLMs—as well as software and algorithms that are made available to users on the Indian Internet must be done with the express consent of the Indian government.
This caused a stir, with numerous technologists, specialists, and AI industry members denouncing the new advisory as harmful to India’s endeavours to develop ChatGPT-like technology.
A few days later, on Monday, Rajeev Chandrasekhar, the Union Minister of State for IT, provided clarification. He said that rather than focusing on startups, this rule primarily targeted large platforms. According to Chandrasekhar, the advice is intended for significant platforms, and requesting permission from Meity is limited to established platforms; startups are not covered.
However, not everybody is persuaded. Rather, they think that Indian startups and AI-related businesses would relocate their operations abroad, especially now that Dubai is also reportedly making significant investments to develop an AI-ready policy and infrastructure environment.
“Bad move by India,” stated Aravind Srinivas, CEO of Perplexity AI, citing a news article on a new advisory. He wasn’t by himself. Bindu Reddy, CEO of Abacus AI, shared similar opinions, saying that India had just bid farewell to its future. The Indian government now has to give its approval before any business can use a GenAI model. That is, deploying a 7b open source model alone no longer requires permission.
Was Altman right?
Returning to Sam Altman and his challenge to Indian tech firms and technologists, Altman has sort of been validated thus far. Though it’s too soon to say that Indian businesses can’t create something along the lines of ChatGPT.
Other LLM initiatives are being carried out in India; some of them might even be more extensive than Krutrim. These things require time as well. However, it’s evident that developing a dependable LLM takes more than a few months, and Indian businesses face significant difficulties, the most significant of which are personnel and infrastructure-related (large farms of high-end GPUs, for example). For background, consider this: this year, Mark Zuckerberg of Meta is purchasing 350000 H100 Nvidia graphics cards with high-end cash. After that, AI models will be trained on these GPUs.
Srinivas brought up these two main points in a recent interview. But recent events indicate that these are not the only obstacles India’s AI plans must overcome. Additionally, there is currently policy space. It is possible that all Indian AI companies could be severely impacted if the Meity advisory remains in place and is strictly enforced.