AI in India - What The Country is Up To?

The US media often gets caught up in self-centered narratives, but we firmly believe that learning about global AI affairs, cultural nuances, and political variations can benefit us and our readers a lot.

In January, Ben Eum presented an in-depth historical analysis of AI in South Korea in 2023. Today, Tasmia Ansari brings us the latest from India, building on her insightful August overview of the nation's AI landscape. Dive into this update to discover the key developments in India's AI journey since then!

From Metaverse Hype to Indigenous LLMs

Six months ago, India's generative AI trajectory started gaining momentum after companies got distracted by the new, shiny, human-mimicking technology. A little over a year ago, the same companies that hailed metaverse were open to abandoning the previous projects, all in pursuit of building a large language model. 

One of the Indian tech titans, Tech Mahindra, rushed into building in-house Indic large language models (LLMs) after leaving 60+ metaverse global projects stranded. By launching a project named Indus, the IT services and consulting company made a genuine initial effort to build an open-source model for the Hindi language and its 37+ dialects. The plan was to move ahead in a phased manner to cover other languages and dialects, but the project's first phase, slated for January, has vanished with an unexplained delay.

Even though the IT leaders akin to Tech Mahindra continue to test the waters, the open-source AI landscape is thriving independently. Many of these culturally rooted models have mushroomed, with Meta's LLaMA as the foundational bedrock for the most. As robust as the foreign models, like LLaMA and Mistral, might be, their proficiency in generating coherent text in Indian languages remains noticeably deficient. Hence, Indian developers have undertaken the task of imprinting India's mark on the global generative AI landscape.

From the southern parts of the country emerged the likes of Tamil LLaMA, Telugu LLaMA, and MalayaLLM, joined by the admired Kannada LLaMA, a personal favorite of AI visionary Yann LeCun. Odia Llama represents the East, and a Marathi LLM stands tall from the West. Indian developers, the second-largest contributors on GitHub in AI, are leaving no linguistic stone unturned.

While most of these local language models are based on Meta's product, there are ventures apart from the infamous Indus striving to build models from scratch. BharatGPT and Sarvam AI are renowned projects pushing the Indian AI ambitions. 

BharatGPT, a research consortium led by the Indian Institute of Technology (IIT) Bombay, with seven other IITs and Google Cloud, has unveiled Hanooman, a suite of language models. The suite's initial quartet of open source models (1.5 billion, 7 billion, 13 billion, and 40 billion parameters) are set to be released next month.

Reportedly, these multimodal tools will engage in conversation across 11 Indian languages, with aspirations to extend support to all 22 official languages of India.

Sarvam AI, on the other hand, partnered with Microsoft. Beyond the partnership, the software giant's chief, Satya Nadella, announced other plans showing generative AI's impact on the Indian economy. He has boldly proclaimed that Microsoft will upskill two million Indians in AI by 2025. Additionally, the company plans to extend the reach of Shiksha CoPilot, an AI copilot initiative by Microsoft Research India, to a hundred schools by the end of the academic year. Shiksha CoPilot leverages Microsoft Azure OpenAI Service and Azure Cognitive Services to analyze and organize content from textbooks. 

Speaking of partnerships, AI leader NVIDIA has also recognized the country's generative AI potential and signed deals with Indian conglomerates Reliance and Tata Group to develop AI-powered supercomputers and AI clouds. On his worldwide AI quest, Jensen Huang, the CEO of NVIDIA, also graced the shores of India and backed the idea of India having sovereignty over its AI capabilities, recognizing the country's distinct and sovereign needs.

Previously in the Global AI Affair series:

• AI Development in South Korea - 2023

• China – the most famous US rival (from January 2023 to early July 2023)

• India – the second-largest talent pool for AI and ML (from January 2023 to early July 2023)

• The UK – emerging leader in Europe (from January 2023 to early July 2023)