LLMs, Enterprise Adoption and Accountability to Fuel India's AI Ambition As the country aspires to become an AI hub, the technology is expected to add around USD 450-500 billion by 2025 and USD 967 billion by 2035 to the Indian economy according to a TeamLease report

Who would have thought in 2014 that the world would be smitten by an artificial chatbot in 2024? OpenAI, the creator of ChatGPT, recently shared that the organisation had over 250 million weekly users alone for work, creativity, and learning purposes. India has come a long way since 2014- when homegrown AI start-ups raised a modest USD 127.7 million- to raise a staggering USD 599 million in 2022.

Evolving at a breakneck speed, the Artificial Intelligence (AI) market has grown exponentially worldwide in the past decade. According to Allied Market Research, the global market was valued at USD 153.6 billion in 2023 and is expected to reach USD 3,636 billion by 2033. In India, Large Language Models (LLMs) and AI adoption in enterprises have been leading the charge.

At Entrepreneur India's flagship event, Entrepreneur Summit 2024, prominent leaders from the AI ecosystem discussed the potential of LLMs, ethical considerations players should keep in mind, a growing appetite for AI and Generative AI (GenAI), and why AI should not be considered as a hype.

(Clockwise- Ankush Sabharwal, Amitabh Nag, Dr. Pradeep Upadhyay, Praveen RP, Ramesh Parthasarathy, and Dr. Ajay Singh)

Leading Bhashini, the Ministry of Electronics and Information Technology and Digital India Corporation initiative, Amitabh Nag shares that while they cater to 22 scheduled languages of India, Bhashini intends to cover the remaining languages. "We have created this society and introduced artificial intelligence (AI), expecting machines to learn in a short span what humanity has developed over thousands of years. Rather than a comprehensive approach, we are progressing step by step. There would be specific use cases which we will be tackling using the basic Research and Development(R&D) technologies," shared Amitabh Nag, CEO, Digital India's Bhashini Division and Director, Digital India Corporation.

Bhashini powers IRCTC's AskDISHA and has helped the chatbot converse in Gujarati apart from Hindi and English. It further aims to add 10 more languages to it. He further notes that to build a vocabulary for a specific use case, one can either take the LLM route- have a billion parameters and hope the required data is included in that or build a step-by-step glossary. "In the real world, when working with AI, there is never a single solution. We need to work on both, rather than embracing just one or the other," Nag added.

Grand View Research shares that the global LLM market size was estimated at USD 4.35 billion in 2023 and is projected to grow at a CAGR of 35.9 per cent from 2024 to 2030. During his 2023 visit to India, OpenAI CEO Sam Altman voiced his concern regarding the country's LLM journey. He felt that while India can try to build ChatGPT-like tools, it will most likely fail. Presently, initiatives and projects such as OpenHathi, BharatGPT, Krutrim, Project Indus, and Bhashini are leading India's LLM dreams.

With discussions over a potential AI bubble and a decline in AI start-up funding, what is the current viewpoint of industry players and experts? "I would say we should stop judging AI now. We should stop considering what is high and what is reality. Consider this as a necessary tool. Do we question whether mobile, internet, or computer is required? It's given now, right? I don't think there would be any field where AI is not being used or would not be researched," shares Ankush Sabharwal, founder, CoRover.ai and BharatGPT.

"If not in use, it would be used in the near future," Sabharwal added. An IBM survey revealed that around 59 per cent of Indian enterprises have actively adopted technologies like AI, making India among the countries with the most extensive AI adoption.

Also read: India Solidifies Commitment to Generative AI With BharatGen

Dr. Ajay Singh, global director and head of AI, HCLTech while echoing Sabharwal's views, shares that system neutrality remains a challenge. "Corporate customers are ready to adopt GenAI, but the only challenge is the system neutrality. What kind of data is being used for the training in the backend? Is there any proprietary stuff? Will this model be deviated? Will there be any drift in the accuracy? What about hallucinations? Biasness certainly is a major concern," Dr Singh notes. While AI has been accepted and appreciated by users and enterprises, GenAI has a different take. "AI models still do have some kind of confidence score. Gen AI models by default, do not carry that," he added.

Ethical biases and cybersecurity concerns have flocked the ecosystem largely. The dataset used for developing a model requires high quality, not just quantity. An inaccurate dataset can lead to skewed outcomes, low accuracy, analytical errors, systematic prejudice, and discrimination against specific groups of people. "You have to put filters and have enough guard rights in place to avoid these biases from LLMs. I think the software engineering lifecycle today itself needs to get a notch up to catch these rather than try to audit these and find them after it has gone out," shares Ramesh Parthasarathy, SVP, Technology, Freshworks.

Praveen RP, chief operating officer– Generative Business services, Happiest Minds Technologies shares, "There were people who took certain decisions to be better than competition but ended up hurting brand reputation." He echoes the need for setting up ethical internal committees, training people effectively on ethics and responsible AI.

Dr. Pradeep Upadhyay, corporate vice president and AI & Generative AI solutions leader, WNS Global Services calls for holding organisations accountable when things go wrong. "Transparency and explainability are crucial as AI increasingly takes autonomous decisions on our behalf. It's important to understand the logic behind those decisions, and that's where explainable AI comes in. Models like SHAP (Shapley Additive Explanations) can be applied, providing insights into why a particular decision was made. This is especially vital in sensitive sectors like healthcare and banking, where understanding the reasoning behind AI-driven outcomes ensures trust, accountability, and improved decision-making," he adds.

As the country aspires to become an AI hub, the technology is expected to add around USD 450-500 billion by 2025 and USD 967 billion by 2035 to the Indian economy according to a TeamLease report.