Dr. Anirban Ghosh, Associate Professor, AI Technology Institute (AiTI), Department of Electronics and Communication Engineering, SRM University AP.
Since inception, the AI race has prominently rewarded one thing: size. The biggest models, the largest data centres, the deepest pockets and in that sense, India has been running behind. But with a call from the leading AI developers across the globe about pacing the race and strengthening guardrails, a different question deserves attention – is it really the scale that deserves sole attention, or should the focus include other aspects too?
In this context, let us reflect on what guardrails really demand. They ask for systems that can be tested under different scenarios, include explainability,and are trusted by ordinary people. That is a multi-objective problem that is much harder to solve,which raw scale and money alone cannot address. It depends on knowing how people use technology, in which languages, and with how much confidence. On these counts, India starts from a position of strength in contrast to what most rankings suggest, due to its demographics. Let us try to understand how.
If you consider the experience of deploying technology at population scale spanning billions, India has mastered it. Theindigenous public digital systems for payments and identity,catering tohundreds of millions of people daily,are burning evidence of the same. The country knows how to reach users who are diverse in every sense of the word (language, competency, need, etc.), price-sensitive, and spread across big cities and small villages. An AI tool that works equally well on a modest phone for a person speaking a regional language as well as the latest smart device used by a tech-savvy person is a very different engineering challenge from the one that is currently being solved. This is a multi-fold problem, and whoever solves it first will have built something the whole world needs. For India, with its diversity, the testbed for such systems already exists; the only thing that matters now is whether we can build it. The second fold in the problem is language, and India holds a clear advantage here too with 22 scheduled languages and hundreds of dialects. The current models are largely trained in English; however, models that can handle a diverse set of languages reliably are not a niche product but a test of whether AI can serve most of humanity. In this regard, efforts under the national AI mission have begun, helped by tens of thousands of shared GPUs offered to startups and researchers at subsidised rates.Furthermore, recent national AI governance guidelines put innovation first, while providing for institutions such as an AI Safety Institute to test and audit systems, laying the foundation for a trusted ecosystem. The final cog in the wheel is the human in the loop. The testing, auditing, and monitoring of AI systems will need thousands of skilled people, whereIndia’s rich demographic dividend provides an opportunity. We already have systems where the technology services sector carries out quality assurance at enormous scale for clients across the world; the natural next step is to extend that expertise to the realm of AI. As opposed to the concern about automation due to the advent of AI, it can provide the sector with a new lease of life. A more measured pace in the development of AI thus provides something precious, which is time to retrain and adapt.To summarize, a country that gets the balance right between these conflicting requirements early could become the preferred place to build and test for the wider global south.
So, as is seen, there are opportunities galore, but nothing is guaranteed, and the risks are real. A slowdown,if universally agreed,could set the ceiling as per the advantage of the pioneers and leave little wiggle room for the late entrants. Thus, a breathing space helps, but we need to double down on our efforts; otherwise, the gap will only widen. We might be late to the party, but with the unique attributes of our nation and the favorable policy changes,we can make our efforts count in the global conversation rather than being a bystander or a follower. India still largely depends on imported chips and, for many products, on foreign models, but with the introduction of dedicated national missions, efforts are being made to address limitations on both fronts. In addition, government skill development programs are launched to provide a deliberate push to AI upskilling. The policymakers are stringent in building trust into the native systems from the first day rather than introducing it as a patchwork later.
India may or may not build the world’s largest AI model;either way, it does not really matter. In a race that is beginning to value reliability and trust more than just efficiency and raw power, the country is uniquely positioned with expertise in building operational and trustworthy systems for the masses. Thus, it has the opportunity to build an AI systemthat earns the trust of a billion users and sets the standard that others follow. A slower race does not have to mean a smaller opportunity,or a late entrance a missed one; in fact, for India, it may be the right moment to run its own race, and to run it well.
