File photo of Shashi Shekhar Vempati, co-founder of A4India.org, and Chairperson of the Apex Advisory at the Department of Science and Technology (DST), Government of India.
| Photo Credit: The Hindu
The increasing use of Artificial Intelligence (AI), especially in the media space, will remain in the long-term human rather than machine-led, despite apprehensions that automation may take over this space, according to Shashi Shekhar Vempati, co-founder of A4India.org, and Chairperson of the Apex Advisory at the Department of Science and Technology (DST), Government of India.
Speaking to The Hindu on a range of issues on AI, Mr. Vempati said that there were several “sticky” issues with regard to AI in the media space. “The first is, of course, of creative copyrights. There are several cases in the courts questioning the manner in which models have been trained, and how they are likely impact the rights of creators, especially if AI can generate creative content that is indistinguishable from the works of the human creator. We may likely see regulatory guardrails on this front,” he said.
“The second and more sticky issue is, ultimately, the essence of creativity is human. A fact that even Prime Minister Modi alluded to in his conversation with podcaster Lex Fridman. I believe that sticky creative content that will be trusted and consumed as a matter of habit requires the human touch. It cannot be fully automated,” he said.
“The advent of AI is of course going to change the workplace and nature of jobs for humans, especially in IT companies in both India and abroad, but not in the way many feel,” he added.
“Despite the high levels of automation, many functions, including software code generation, will require a degree of human oversight,” Mr. Vempati said.
He added that as far as Indian IT was concerned, the biggest area of impact of automation would be felt in low level tasks, including testing, quality assurance, etc., and with software code generation using AI becoming better and better, even software development would, to a large extent, get automated.
“It is also believed by some that with AI lowering the cost of software development and giving greater flexibility to enterprises to evolve their software systems to suit their business needs, we may see the clock turning back towards more in-house custom applications rather than off-the-shelf products,” Mr. Vempati said.
“Another substantial area of opportunity for Indian IT would be around annotating and enriching large datasets to make them AI usable. As AI use cases become more and more specialised, requiring deep domain knowledge, these value-added data services will be the natural evolution for Indian IT building on earlier cycles of ITO (IT outsourcing), BPO (business process outsourcing) and KPO (knowledge process outsourcing),” he said.
In terms of what had been the focus in India with reference to AI, Mr. Vempati said that it was guided by “our societal concerns”.
“India’s primary AI focus has been on overcoming the language barrier, and rightly so. The ‘Mann ki Baat’ corpus, which is a highly reliable and accurate corpus of text and audio in multiple Indian languages and dialects, was the basis for many of the early efforts focused on Indian models for AI. Bhashini by the MeiTY (Ministry of Electronics and Information Technology), which PM Modi has used for his speeches during the 2024 elections, stands out, and Anuvadini by the AICTE (All India Council for Technical Education) is being used to translate textbooks across all levels of education from schools to universities using AI. AI4Bharat at IIT (Indian Institute of Technology) Madras has released a wide range of models and datasets based on Indian languages, encompassing text and voice. The BharatGen initiative at IIT Bombay, funded by the DST, is looking at frugal Indian language models. Sarvam AI, incubated out of IIT Madras, has been roped in by the UIDAI (Unique Identification Authority of India) for a range of use cases. mVaak out of IIT Kanpur has built an audio/voice model leveraging DD (Doordarshan)/AIR (All India Radio) archives for a range of use cases,” Mr. Vempati said.
There are, he added, several startups focused on other kinds of AI models, including Paralaxiom in Bengaluru, working on a vision AI model for public spaces in the Indian context; Pienomial in Pune, which has a knowledge model for the Life Sciences industry aiding highly accurate and reliable research; and Innoplexus, also out of Pune, that has amassed an array of patents in AI-led drug discovery, among others.
In terms of challenges for Indian AI startups, Mr. Vempati said that apart from funding and access to computing power, there are challenges for access to public datasets in the Indian context. “We at AI4India.Org have done our bit to motivate Indian organisations, institutions, corporates and public bodies to pledge to make their datasets accessible to AI model developers and researchers through what we are calling ‘Data Daan’ (http://datadaan.org), as an ethical voluntary movement to share data, in stark contrast to the manner in which LLMs (large language models) have been trained violating copyrights,” Mr. Vempati said.
“Ultimately, it’s more important for India to create conditions for the diffusion of AI across all layers of the economy and society, so we are able to reap tangible benefits in healthcare, education, and governance by replicating the success of other DPGs (digital public goods) such as UPI (Unified Payments Interface), etc.,” Mr. Vempati said.
Published – March 23, 2025 06:48 pm IST