The State of AI in India and Some Remedial Measures

“There is a tide in the affairs of men

Which, taken at the flood, leads on to fortune;

Omitted, all the voyage of their life

Is bound in shallows and in miseries.

On such a full sea are we now afloat;

And we must take the current when it serves,

Or lose our ventures.”

―William Shakespeare (Julius Caesar, Act IV, Scene 3)

Since the invention of transistors in the 1950s, the world has witnessed technological advancements not only at a rapid pace but also at a pace that has gained momentum decade after decade. Our lives have been completely transformed by advances in computing power, genetic engineering, satellite imagery, digital technologies like the internet, e-commerce, e-banking, smartphones, and so much more. However, it can be argued that all the technological developments of these decades were only the means to reach the present juncture: the creation of artificial intelligence (AI). For the first time in the evolution of our species, we are debating the essence of what makes us human not in relation to nature but to a human-made entity. The potential to disrupt our world with this new technology is immense. Herein lies an opportunity that presents itself once in decades and one that India is, unfortunately, missing to lead from the front.

The three most critical components that have led to recent advances in AI by organizations like OpenAI, Meta, Microsoft, etc. are a) data, b) computing infrastructure, and c) an AI ecosystem. While Indian companies may not have as much data (digitized data of our regional languages and access to millions of images) as their counterparts in the U.S., this is not insurmountable. Buying data directly from consumers and digitizing our regional language repositories are tasks our companies can perform efficiently and cost-effectively. Thus, the first critical AI component is something India can fix en passant but the latter two need a complete overhaul of our present priorities.

Computing infrastructure is a highly strategic technology for India’s digital sovereignty. Currently, more than 60% of India’s cloud computing market share is held by three American companies: Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform. These firms provide reliable services and state-of-the-art cybersecurity measures. The services provided by these firms are apt for Indian startups, which predominantly rely on them for software-as-a-service (SaaS) applications. However, natively Indian cloud computing platforms are either unreliable, insecure, or lack the necessary software capabilities for developing and deploying applications at scale.

This high degree of dependence on other countries and companies is highly concerning. In fact, the dependence on the firms listed above is presently so stark that the highly sensitive digital infrastructure for CoWIN (Covid Vaccine Intelligence Network) as well as Aadhaar Data Vaults have been created using AWS. Granted, our relations with the U.S. are presently warmer than ever before, it is not prudent from a realpolitik perspective to be dependent on a single nation’s private firms for critical infrastructure.

The services of these firms are taken by the Government of India based on the strong cybersecurity measures provided by them. Yet, in June 2023 and October 2023, data, including sensitive personal information from CoWIN and Aadhaar databases, were leaked on a massive scale. These data leaks, the lack of alternatives for the government, and the cloud providers’ lack of accountability raise serious questions about India’s digital sovereignty. Additionally, the cost of cloud computing services provided by these firms is quite exorbitant. For example, just to train, tune, and validate a single deep learning-based AI model for a problem like disease diagnosis in medical imaging could cost upwards of a thousand dollars. To be able to train a model like the ChatGPT- 4 requires millions of dollars of compute power which most private institutions in our nation cannot afford. This high entry cost is another factor hampering the development of cutting-edge AI models in India.

Indian private players are quite reluctant at the moment to establish their own data centers and infrastructure for cloud computing. This is because of the heavy initial investment needed to establish them, which only a few companies can afford. Besides, the incentive for the government to help the industry acquire land and provide support in building infrastructure is also lacking since data centers create only a handful of new jobs. Large data centers also require the consumption of a lot of power. We presently inhabit a world in which carbon emissions for AI have already surpassed those for air travel. Even querying a large AI model like OpenAI’s ChatGPT-4 a few times by our digitally literate population has the same amount of carbon emissions as that of a car traversing the length and breadth of India many times. As another example, training a ChatGPT-3 which is orders of magnitude smaller than ChatGPT-4 takes approximately 700,000 litres of water to cool down the machinery in the data centers. Thus, establishing large data centers does not make financial sense for private players, and if done haphazardly, conflicts with the government’s goal to make India energy-independent through clean technology by 2047.

This brings us to the third most pervasive problem: the lack of an indigenous ecosystem to promote fundamental research in AI. There are only a handful of companies such as Google, Meta, OpenAI, Microsoft, and Mistral that have almost all the AI talent to build advanced foundational AI models (such as a ChatGPT-4) from scratch. Easy access to funding for computing resources, access to a lot of data, and a talented pool of graduates and PhDs from world-class universities means these researchers do not seek opportunities outside of the U.S. (or Europe, in the case of Mistral) as lucrative for them either professionally or monetarily. Consequently, we are always a few steps behind in the AI race, developing applications on top of the AI models being made elsewhere rather than contributing to the development of foundational AI models. For example, at a time when global corporations are working on deploying natively multimodal AI systems (like the newly released GPT-4o) that are capable of understanding and working with text, speech, and video simultaneously, our companies are stuck with building or tuning large language models (LLMs) for various applications.

The AI wave that we are witnessing right now and missing because of the problems identified above needs urgent remedial measures. Despite the challenges, the good news is that we have an example from our own history to learn from: the establishment of the Indian Space Research Organization (ISRO). In the 1950s, the Government of India witnessed the launch of space programs by technologically advanced nations like the U.S. and the U.S.S.R. and quickly understood the potential of space technology for India’s progress. Consequently, the Indian National Committee for Space Research (INCOSPAR) was established in 1962, which became a precursor to the ISRO. Just like AI is now, space research at the time needed a huge amount of investment for setting up the infrastructure, while an ecosystem to promote talent in this research area had to be nurtured. In just a few years, India became a leading power in space research by independently developing indigenous systems for rocket launches, propulsion, and satellite communication. With time, India gained a strategic and sovereign advantage by developing indigenous space technologies and is now catering to the needs of advanced nations. Our world-class indigenous systems that are second to none are providing cheaper satellite launches, weather updates, and navigation information to many other countries.

The establishment of ISRO and its achievements were only possible because the government stepped in to invest huge amounts of financial resources and gave a free hand to renowned scientists like Vikram Sarabhai to create a pool of talented individuals. This led to the return of many Indians who studied abroad while Sarabhai and others also nurtured talent from within the country.

To replicate ISRO’s success in AI, the government must undertake three initiatives without delay. First, the government must collaborate with leading industry players and provide financial incentives to set up advanced data centers across the country. Identifying locations where green energy can be utilized for operating these data centers, taking into account other factors as well, such as job creation, proximity to the submarine communications cables, and cost-effective subscription plans, is crucial. The establishment of this computing infrastructure, with the support of companies like Nvidia, which manufactures the world’s best GPUs, is essential for democratizing AI research in our universities and industry. With time, India should develop capabilities to fabricate its own semiconductors and GPUs with a focus on their increasing import demand in other developing countries, to insert ourselves in the global AI hardware supply chain.

Second, the government must create a consortium of AI researchers of Indian origin to identify foundational areas of AI innovation and research critical for India’s sovereignty and economy. This initiative should be backed by government funding and aimed at creating AI hubs across universities and industries to stimulate collaborations across them. This will lead to the establishment of an AI ecosystem within the nation where domestic talent could be nurtured to work on these foundational AI models. Competing with the best AI models from other countries will safeguard our sovereignty and prepare us to tackle the societal and economic upheavals caused by AI. The applications of these models should be focused on the grand challenges facing our nation such as water scarcity, renewable energy, air pollution, and personalized healthcare. This will also enable India to become the hub of the global AI supply chain focused on meeting the sustainable development goals of various developing nations.

Third, since the first two initiatives will take a few years to go from sowing the seeds to fruition, until then the government should engage with private players like Amazon, Microsoft, Google, etc. to subsidize their cloud resources for consumers. India already has significant leverage to undertake this because it is one of the largest (and growing) markets for these U.S. tech companies and these tech companies are filled with employees of Indian origin. The government should simultaneously promote data portability and interoperability of cloud resources at different levels with these companies. This will work like a catalyst to significantly advance our present AI capabilities, enable the entry of several new private entities in this space, and address the needs of institutions facing financial constraints in adopting AI. This “AI subsidy” will go a long way in making sure that we can establish many more AI hubs across the country with university-industry collaborations and make valuable social and economic impact.

We inhabit a world where AI researchers are presently divided on a question fundamental to the existence of our species and our planet: Will AI become conscious and supersede human intelligence? In these interesting times never before witnessed by our species, India, being the most populous nation, has a significant responsibility at hand. We will only be able to play our part concerning the future of our species and our planet if we are undertaking research at the forefront of AI. Even if it turns out later that AI does not develop superhuman intelligence, it is certain that AI’s adoption in every area of human activity will absolutely transform how we live and work, and pose newer societal and cultural challenges. India will be better prepared for these if we can anticipate them before they knock on our doorstep. This can only be done by building the capacity to conduct foundational AI research to keep us at the forefront of AI’s evolving trajectory rather than responding to the disruptions being caused by other technologically advanced nations.

On the other hand, the prospect of AI evolving superhuman intelligence makes our times even more crucial, unprecedented, and singularly rare. Most private firms such as the ones listed above are investing only a small fraction of resources (less than 5%) in AI safety experimentation. India can lead by conducting large-scale experiments to ensure that AI research is not going in a direction averse to the interests of our species and the planet. If AI is indeed heading towards superhuman intelligence, India must be at the forefront to build safeguards around this technology, preventing its misuse, and averting its proliferation in the hands of nefarious entities. Learning from our own history, under no circumstances should we live under a system imposing an “AI apartheid” upon us. The only way to ensure a seat at the highest policymaking table is to ride this AI wave and carve our path through the stormy currents ahead.