Indian IT firms exhibit varying strategies in AI; Infosys and Tech Mahindra favor small AI models to lower costs, while TCS and others focus on existing tools. Smaller models dominate due to cost efficiency and specialized applications, amid security concerns with large models. Currently, revenue from generative AI remains unreported by major players, contrasting with Accenture’s significant gains.
Diverging Paths in AI Development among Indian IT Firms
India’s leading IT services firms are adopting contrasting strategies in their AI initiatives. Companies like Infosys Ltd and Tech Mahindra Ltd are focusing on developing their own small AI models, aiming to help clients reduce costs. In contrast, Tata Consultancy Services Ltd, Wipro Ltd, and HCL Technologies Ltd are leaning towards enhancing existing foundational AI tools available in the market.
Understanding the Scale of AI Models
The efficacy of an AI model is largely influenced by the scale of data utilized for its training. Small AI models, known as Small Language Models (SLMs), operate on limited data sets and cater to specialized functions, while larger models, referred to as Large Language Models (LLMs), are fed more extensive data for broader applications. Most firms leverage internal datasets to construct these smaller models effectively.
Infosys, the second-largest software services provider in India, reported revenues of $18.6 billion and actively develops small language models tailored to its clients. During a press conference on January 16, Chief Executive Salil Parekh stated, “In Generative AI, we have built four small language models for banking, IT operations, cyber security, and broadly for enterprises.”
In an earlier press call, Parekh elaborated, “The reason for the small language model, we believe, we have some very good data sets within Infosys. And we are taking some, let us call it, clean data sets from outside the industry…”
Chairman Nandan Nilekani echoed this sentiment, advocating for the effectiveness of small language models tailored to specific data. He remarked, “Small language models trained on very specific data are actually quite effective… I think they don’t have to build these gigantic ones” during an interview in the Financial Times.
Tech Mahindra’s Shift to Small Models
Tech Mahindra has joined Infosys in transitioning to smaller language models, focusing on specific client needs. CEO Mohit Joshi explained, “We have since moved from these LLMs to creating small language models and tiny language models.” These models, he noted, enable solutions for precise issues without extensive computing costs or carbon emissions.
Joshi illustrated their potential by referencing a small language model’s ability to enhance search functionalities on a desktop without external aid, such as ChatGPT.
TCS Concerns Over Data Security
Tata Consultancy Services (TCS) has developed its proprietary AI model, WisdomNext, but has not categorized it as either large or small. An anonymous executive highlighted concerns regarding data security with LLMs, stating, “Biggest problem in using large AI models is that it poses data security issues…”
Ultimately, the preference among clients is skewed towards smaller, in-house AI models that offer not only cost-effectiveness but also enhanced data protection. This sentiment was echoed by Abhishek Kumar, an analyst at JM Financial, who remarked, “Customers want small language models for specific problems.”
Cost Efficiency of Smaller Models
Sriram Raghavan, Vice-President of IBM Research AI, noted the financial advantage of smaller models, emphasizing the high hardware costs associated with AI. He suggested that small models can provide substantial cost savings, estimating reductions of up to 50 times for specialized use cases: “We are talking a real order of magnitude difference.”
As of now, India’s IT giants have not reported any revenue from generative AI, a sector gaining traction since the introduction of ChatGPT, which showcases advanced human-like capabilities across various content forms. In contrast, Accenture, the largest global software services firm, reported generating $900 million from generative AI, constituting 1.4% of its total revenue of $64.9 billion.
India’s IT industry displays a clear divide in AI strategies, with some firms opting for the development of specialized small models tailored for specific tasks while others focus on leveraging existing large models. This divergence highlights differing philosophies on data security, cost efficiency, and innovation within the sector. As companies navigate this landscape, their approaches will significantly impact client solutions and operational strategies going forward.
Original Source: www.livemint.com
Leave a Reply