Ola co-founder and CEO Bhavish Aggarwal made news last week by unveiling a Made in India large language model (LLM) and generative artificial intelligence (AI) platform – Krutrim — on the lines of OpenAI’s ChatGPT, and Google’s Bard. However, Aggarwal is neither the first nor the only homegrown firm to come out with such an offering. There are several others who have ventured into the field earlier and are in the process of developing such models.
But will these models emerge as global alternatives to ChatGPT and Bard in terms of scale and capability, or will just serve niche segments? Analysts say that the answer will lie in the available data sets these players are able to tap and the number of servers they are able to deploy. Both require large investments and painstaking efforts.
For instance, Krutrim, which will come in two models — a base model and a Pro model — is currently being used by all Ola group companies for a variety of internal tasks, including customer support, voice and chat, and customer sales calls.
Vivek Raghavan, founder of home-grown generative AI startup, SarvamAI, which unveiled OpenHathi, the first Hindi LLM, before Krutrim, says that instead of having large LLMs like open AI, there can be smaller LLMs which are more efficient and can be deployed at lower cost.
“These will be very good at a specific task and in a specific language at extremely low cost. This is the unique technology we want to bring in. Smaller models that are domain-specific are very efficient,” he told Fe.
Nikhil Malhotra, global head, Makers Lab, Tech Mahindra, echoes the same view. “India has the potential to create its own model at a lower cost which will also take into account the country’s vibrant social fabric,” he said.
Tech Mahindra has also unveiled its LLM named the Indus Project, which aims to establish a foundation model for Indian languages. In the first phase, the company is focusing on creating an LLM for the Hindi language, covering 40 dialects. It plans to launch the LLM early next year.
The company has mentioned on its website that its goal is to “build an open source large language AI model to serve the needs of 25% of the world’s population”.
According to Malhotra, an LLM based on Indian languages would enable the implementation of specialised systems needed for different sectors in the country, such as rural finance, agritech, retail, and travel and tourism industry.
Zoho is another domestic startup which announced in June its plan to build its own LLM on the lines of OpenAI’s GPT and Google’s PaLM 2. Sridhar Vembu had said that the company is working on smaller AI models that are based on 7 billion to 20 billion parameters to solve specific domain problems for its customers.
Ramprakash Ramamoorthy, director – AI research, Zoho Corp, said, “Achieving technological sovereignty is crucial and cannot be fully realised if we continue to rely on imported graphic processing units (GPU) and other specialised hardware for building LLMs. Developing homegrown expertise and capabilities in the hardware arena is essential for the future of LLMs in India”.
Earlier this month, another domestic firm, CoRover.ai, unveiled its indigenous LLM named BharatGPT. This in collaboration with Bhashini, a national language translation mission under the ministry of electronics and information technology (MeitY), supports over 12 Indian languages.
CoRover provides AI virtual assistants, including chatbots, voicebots, and videobots, to numerous organisations such as IRCTC, LIC, IGL, KSRTC, NPCI, and has a user base of around 1 billion.
Ankush Sabharwal, CEO, CoRover said that just having India specific LLMs without solving real problems won’t help the business. “Models need to have the right use cases and best market fit, to be able to innovate and bring down the cost”.
According to Pareekh Jain, founder, Pareekh Consulting, it is premature to bet on Indian LLMs as their use cases and product differentiation is still not very clear. “The product market fit is yet to evolve. Global tech firms like Google and Meta have high quality data than others. Only if a firm has relevant data that fits with certain use cases, will the LLM be in demand,” he said.
