Startups are driving the development of nation-centric models in India’s AI revolution

Startups are driving the development of nation-centric models in India’s AI revolution

Many start-ups in the nation are providing AI solutions with an emphasis on India, despite the fact that the majority of AI models now in use worldwide are based on the English language.

The CEO of Ola, an Indian ride-hailing app that rivals Uber, announced this month a new artificial intelligence (AI) startup that aims to create an entire AI system for India from the ground up.

With the goal of making AI India-centric, this is a relatively new area for the nation.

During a live broadcast of the new AI company’s launch, Krutrim, the creator Bhavish Aggarwal, who is also the CEO and co-founder of Ola, stated that a huge number of the AI models in use today are steeped in Western culture and the English language.

It can comprehend and produce material in 10 of the 22 official languages of India, for which it was designed exclusively.

Mr. Aggarwal asked Krutrim, whose name means “artificial” in Sanskrit, to compose a Bengali poem on the monsoon rains in order to showcase some of the technology’s possibilities.

Krutrim, which will be completely available for use next month, was developed using billions of pieces of data that are exclusive to India, according to the business.

According to Mr. Aggarwal, “AI will define the future paradigms of economy and culture.” “And India needs to become a global leader in AI in order to become a true leader of the world.”

It states that Krutrim is intended for usage in a variety of contexts, including business communications and education.

The creation of the company coincides with the widespread use of AI throughout the world and with developments that have raised calls for and worries about AI regulation.

With the emergence of generative AI this year, which can produce fresh text, video, image, and audio content, there have been advances and controversy on a worldwide scale.

The most well-known instance of this is ChatGPT. ChatGPT is a chatbot that can generate text to suit users’ individual needs and engages in conversational interaction. It was developed by the AI research company OpenAI.

According to observers, these technological advancements offer a chance to improve economies and develop efficiency.

India might add $359 billion to $438 billion in the fiscal year between April 2029 and March 2030 if it fully utilizes generative AI technology, according to  inquiry. This would represent a 5.9–7.2% rise in the country’s baseline gross domestic product.

According to this, the application of AI in industries like retail, financial services, and IT will have the biggest influence.

AI has been cited by the Indian government as a key facilitator of the nation’s digital economy.

The industry is seeing an influx of investment. According to Report, the total investment made in AI start-ups in India last year was $3.24 billion, making it the fifth biggest amount worldwide.

“AI is pervasive across Indian organisations, influencing both business functions and industry sectors, as they increasingly embrace the transformative power of AI to drive innovation and enhance overall operational efficiency,” says Sameer Dhanrajani, chief executive at AI consulting and advisory firm AIQRATE and at 3AI, a platform for AI and analytics leaders and professionals.

Beyond utilizing AI, though, India, with its booming IT industry, has the chance to contribute to its development and design systems tailored to the nation of over 1.4 billion people, each with distinct and varied linguistic and cultural traits. Although Google and OpenAI do have databases in Indian languages, they mostly use English data.

That’s where businesses like Krutrim come into play.

India has been adopting AI, and in an attempt to capitalize on the trend, a variety of start-ups have formed in the nation offering AI solutions.

By creating their own large language model (LLM), or foundational model, businesses like Krutrim hope to go beyond this.

“Many businesses, big and small, are vying to accomplish that,” says Jaspreet Bindra, the creator of the consulting firm The Tech Whisperer.

One of the biggest providers of IT services globally, Tech Mahindra, is among the other Indian businesses working on LLM creation.

Tech Mahindra’s model, named Project Indus, is intended to be able to comprehend forty distinct Indian languages. There are numerous additional important languages in India besides the country’s 22 official languages.

Launch This month, announced the introduction of BharatGPT, an Indian generative AI platform for which it has partnered with Google Cloud on a technological level.

Another Bengaluru-based startup, Sarvam AI, is also producing LLMs.

“There has been a lot of discussion around India-centric models,” says Mr Bindra. “The inspiration from that comes from countries that have already created their own models – China has over 100 such models.

“The UAE pleasantly surprised the world by creating Falcon and then Jais, which are world-class open-source AI and LLMs. Therefore, the question in India is that why can’t India – with all its IT prowess, with all the human resources that it has – create its own LLMs?”

However, it is “not very easy to create an LLM of the scale, size and performance of a ChatGPT”, Mr Bindra says.

He says that building a ground-up, full stack LLM can cost billions of dollars.

Finding the talent to build the technology is one of the other challenges.

For instance, computer scientists in Bengaluru and San Francisco created the software program Krutrim.

However, collecting the massive amounts of data needed to create such models is one of the most challenging aspects of the process, according to Mr. Bindra.

“More important than language is context,” he says. “We need models with Indian context – Indian healthcare information, Indian data from radio and TV channels, Indian land records data, legal, education. All of these are going to be important to create the right kind of models for India,” he says.

Ultimately, if India can develop models that are accessible to the mass population, it will give a massive boost to the economy, he adds.

“India-centric AI models are crucial due to the country’s diverse linguistic, cultural and socio-economic landscape,” says Deepika Loganathan, co-founder and chief executive at HaiVE. Tech, an AI provider.

“Tailoring AI models to India’s unique context ensures they are more effective and inclusive.”

However, India’s work in this area is still at a relatively nascent stage, she adds.

“While some strides have been made in this area, notably in language processing and localised applications, there’s a need for more extensive work,” says Ms Loganathan.

This includes “developing data sets that reflect India’s diversity and addressing local challenges through AI”.

The main challenges include shortage of skilled AI professionals, infrastructural constraints and data privacy concerns, she says.

“Additionally, there’s a need for more comprehensive regulatory frameworks to govern AI use.”

Due to AI’s exponential advancement and expanding capabilities, India and other nations must decide how to regulate the technology while raising ethical questions, security concerns, and possible employment ramifications.

“A multifaceted approach” is necessary to address these issues, according to Ms. Loganathan. This would entail creating public-private partnerships, investing in infrastructure, improving AI education and training, and creating clear AI policies.

Industry insiders, however, are still upbeat about the potential advancements AI may see in India by 2024.

“With increasing investment in AI research and development, growing government support, and the rising adoption of AI across industries, we are likely to see significant advancements,” says Ms Loganathan.

“I anticipate more collaboration between academia, industry and government to drive AI innovation, addressing both domestic and global challenges.”