Building an AI strategy - AI Square

Atmanirbhar India's Artificial Intelligence Square

This is the second post in a new series titled “Open Ideas for India”, where I hypothesise on technology interventions that can make an impact in India.

---

Artificial Intelligence (AI) is a major disruptor in the making. Individuals, communities, companies, countries, and humanity at large should tactically adapt to and strategically plan for this disruption.

Like in other aspects, India’s future in AI remains highly uncertain. India is capable of rising to the very top of the AI pyramid with its talented workforce and data-generating demography. It is also capable of languishing at the middle of the stack without the right leadership or focus on innovation.

A lot has already been said about AI strategy, including India’s AI strategy. And yet no clarity remains on the top three or five things to be done. It also remains unclear who should be doing these top things - the government, the IITs, the IT industry, or the startups. We are thus in the dangerous twilight of much having already been said and little done.

In this backdrop, I propose something unusual - to build, a building. Not just any building, but the tallest building in the country - the AI square: Atmanirbhar India’s (AI) Artificial Intelligence (AI) Square. A 108-storey building, costing about 5-10 billion USD, that should be a physical statement and symbol of vision, intent, and action.

Why a building? There are three fundamental reasons:

1. Thinking really big
   
   The characteristic of the AI revolution of the last decade has been the unexpected and exponential scale in the amount of data and computation required, the size of AI models, the scale at which these models are deployed and the wealth they create.
   
   It is reasonable to state that the scales of today will look puny in a few years. For this singular reason, plans for AI strategy in the country require thinking big. Erecting the tallest building in the country (and probably the 8th tallest building in the world) will first be a recognition of the towering scale needed to address AI.
   
   By virtue of physical scale, the building will concentrate the brightest minds of the country. The current pandemic notwithstanding, innovation requires people to come together and tinker in a garage-like mode. Solving AI for India would require a galaxy of such garages with many shining minds coming together.
   
   The building should directly affect the learning of 10,00,000 learners a year, host 1,000 AI fellows chosen with a national program like the NTSE scholarship, incubate 1,000 different startups innovating in AI, be the home to 100 exceptional faculty members, and house at least 10 exaflop compute power. At this scale, purpose-built infrastructure itself would be an impetus for quality of work.
   
   The building must be topped off with a world-class conference centre providing unmatched views of the surroundings.
   
   

2. Building from scratch
   
   India has its institutions - the elite technical colleges (IITs and IISc), the government bodies such as CDAC, industry associations such as NASSCOM and CII, leading IT services companies such as TCS and Infosys, and leading MNC tech companies with India labs such as Google, Microsoft, and Amazon, and a host of startups and incubators.
   
   A strategy for AI innovation would naturally be based around these institutions. All these institutions have strengths to contribute and value to gain from an AI revolution in the country.
   
   But they also come with structural weaknesses of varied kinds. The IITs have faculty juggling between teaching, mentorship, administration, and research, within structures that are outdated. Bodies such as CDAC have not kept pace with the rapid progress in AI to meaningfully lead or even contribute to strategy. The IT services companies are in a slow process of up-skilling their employees in AI while battling their chosen quarterly revenue targets. The MNC labs are technically most competent but have high-velocity high-value trajectories of their own organisations to drive. Opening a centre of excellence in an IIT or starting a program on AI for social good at an MNC lab are useful attempts but ones that are likely be insignificant in the longer run.
   
   In contrast to reinforcing existing institutions while inheriting their weaknesses, we need to build fresh, build from the ground up, and build a new coalition of partners who bring their respective strengths to the table.
   
   

3. Integrating full-stack
   
   The last decade has quietly changed the technology landscape. A tech company is today the largest car company by market cap and builds processors that go into them. An online e-commerce company is building the compute infrastructure for the world. Tech companies own the largest communities of human interaction with the ability to influence elections and ban presidents.
   
   The traditional boundaries between hardware and software, product and services, offline and online, truth and falsehood are rapidly blurring. And in this new reality, companies and countries cannot afford to work and own only part of the stack.
   
   We need a full-stack approach with vertical integration from silicon to software to user base to data. This vertical integration should be physically manifest in a building, housing cutting edge compute clusters and data centres, hardware teams, AI scientists, startups and companies that take technologies to the market which is ready with a large demand.
   
   The idea of full stack also has a strategic manifestation. In the past, a weak military-industry-academic complex has slowed technological progress and made India dependent on imported strategic tech. We need to address this with an integration of researchers, government, and strategic sectors (defence, space, and cyber-security) within the same building shedding the cultural boundaries between them.

---

AI SQUARE

AI Square with its 108 floors

---

AI Square will provide the physical manifestation of an AI Strategy, a visual representation of coherent thinking of an aspiring nation. But for it to succeed, we need a core set of principles to work with. Here are five that I recognise as being important.

1. Providing education for all
   
   The most foundational lever of change is education. This is particularly important for AI technologies which are not only new but rapidly evolving.
   
   Education in AI should be seen at multiple levels. The general populace must build a common-sensical knowledge of what AI is, what data-driven algorithms can and cannot do, and the general trends into the future which will affect their lives. This also includes those in the corridors of power, such as the bureaucracy. Then, there are professionals in different areas - science, finance, sports, law, tourism, agriculture, and arts - who need to understand what AI tools are relevant to maximise their impact at work. Then, there are the professionals in the IT industry who need to be up-skilled in AI. And finally, there are students who need to learn both the basics and advanced topics in AI to become the torch bearers in the future.
   
   There are several pitfalls, many already laid out, that are to be avoided in education in AI.
   First, is the misconception that AI is like earlier technology shifts, from say mainframes to PC, from PC to mobile, or from one programming language to another. Data driven programming is a fundamental shift in the approach, economics, and outcome of software.
   Second, the default approach to learn AI today is to parachute into learning a language such as Python with a few libraries. This is a quick start, but one that can come to a quick halt if individuals do not keep pace with the evolving landscape. Instead, innovation in AI requires deep foundations in a wide spread of topics from mathematics, to efficient software, to systems engineering, to hardware design, and perhaps to quantum physics.
   Third, education is foundational, so foundational that it must be almost free. The Square should focus on providing almost-free education at scale to the entire country. And in the process it should drive out of market the plethora of educational companies who today offer expensive, sub-standard courses. Indeed, the Square should aim to reach all gainfully employed citizens within a period of 10 years.
   
   

2. Moving from open-source to open-benefit
   
   Much of the technology behind AI is being open-sourced. This has enabled networked effects and rapid progress in the field.
   
   However, one must be careful in equating open-source with open-benefit. An open-sourced AI model does not imply that it is directly and freely usable. Training an open-sourced AI model requires data and compute, and so does specialising it for a specific domain. Thus, there is many a slip between open-sourcing a model and people benefiting widely from it.
   
   An example here is the state of AI for Indian languages. AI models of machine translation (MT) are being open-sourced. But for one to build a MT system for example from Odia to Tamil requires curation of a lot of data, careful annotation of the data, training on expensive hardware, and then evaluation on carefully created benchmarks. Many of these are missing, today.
   
   The Square must create scalable recipes to take the kernel of open source and convert it into usable systems that lead to open benefit. In particular, the Square must identify for each of the 136 crore citizens, the 6 crore small-scale businesses, and the 36 states and union territories of the country, the AI tools that would openly benefit them. Research and innovation must be prioritised on broad-basing the impact of AI throughout the nation.
   
   

3. Retaining data sovereignty
   
   Data is the king in every AI strategy. India with its demographic strength, diversity in languages and cultures, affordable access to communication media, and rich social networks, is extremely data rich. Clearly, there is much commercial and strategic interest in getting access to this data.
   
   In an increasingly digital world with blurring international borders, sovereignty will not be based on constitutions of the country. It will be based on who has access to a country’s data and consequently power to algorithmically influence huge networks of people.
   
   It is essential that the Square act as a focal point in establishing the value of the data generated in the country, recognising how it is being currently used, and how it must be regulated. The Square should host a new Ministry of Artificial Intelligence, whose goal it must be to provide the optimal legislative frameworks for retaining India’s data sovereignty.
   
   Data sovereignty must not be confused with closing one’s economy to large corporations. This would be counter-productive. The citizens must be provided the best digital services on offer in the world. But companies providing these services must be accountable to the elected government and also invest the accrued wealth back in the country.
   
   

4. Insisting incessantly on innovation
   
   AI is expected to widen economic disparities. The haves and have-nots are going to move further apart. Efforts can be made to slow this process with steps such as the above three - providing education to all, moving from open-source to open-benefit, and retaining data sovereignty. But it is unreasonable to expect the concentration of wealth to reverse.
   
   While building usable services will drive the growth and adoption of AI, wealth creation will be based primarily on innovation. In this context, the Square must be the place to create wealth in and for the nation through innovation in AI. There should be no hesitation in recognising this as the primary objective to success.
   
   All functions of the Square - startups, R&D labs, AI fellowship programs, faculty labs, strategic efforts - must be primarily judged on innovation at the cutting edge. Whether it is the number of publications in top-tier conferences, the number of patents filed, or the number of first-of-a-kind products launched, innovation should drive people.
   
   While sounding obvious, this focus on innovation has been a major lacuna for the country. We are often reluctant to invest in the long-term big bets that are necessary for innovative solutions. It requires a major mindset upheaval with both institutional support and the right education.
   
   

5. Building on India’s historical strengths
   
   India is a rich and old civilisation, a civilisation that lives implicitly in its culture. This civilisation has much to offer to the new age driven by AI. This simple statement carries much strength, which I guess is not well recognised. Perhaps this is due to the limited intellectual capital at the intersection of ancient India’s depth and modern technology’s prowess.
   
   The Square should be a place where this intersection is explored and allowed to flower. I highlight three areas to illustrate opportunities at this intersection.
   
   AI Ethics: Technological interventions have always come bundled with deep ethical questions. AI magnifies these ethical choices given the scale of impact and the speed of disruption. Ethics remains an active area of research widely. India, with its rich philosophical history, has much to say about Ethics, and much that is quite different from the western objectivism. For instance, Swami Vivekananda’s modern, concise, and piercingly insightful take on ethics needs to be carefully studied to provide a fresh and foundational view on AI ethics.
   
   AI and the Self: The rise of AI is subtly also a tension between the subjective individual and the objective modelling of that individual. For instance, YouTube today may know my video watching preferences more accurately than my friends or even myself. This is expected to only accelerate as content, interaction data, and digital penetration increases. How does an individual marginalised to accurate AI models live life meaningfully? India’s rich philosophy on subjective experience and its developed technologies of leading contented lives are essential toolkits of life in a post-AI era. The Square must explore and showcase this with scientific rigour and clarity.
   
   Championing the case of the Global South: India’s philosophy has espoused a cosmic oneness whose geo-political manifestation is in thinking of peaceful, collaborative relations with all countries. India’s experience in the Square - in education, broad-basing benefits of AI, maintaining data sovereignty, and innovation - must be generously shared with the global south. In particular, African, ASEAN, and middle-east countries should be the immediate benefactors of the Square. This would establish India not only as a benefactor country, but also build hard support for India’s otherwise less-than-potent soft-power.

---

We must not expect the government to lead this effort. AI is a wealth creator. The 143 Indian billionaires should come together to build the Square as a concrete representation of what India stands for.

The time for incremental and unimaginative thinking is over. Let’s get building.

---

PS. Solving AI for India would require many ideas. Consider reflecting on these and writing me an email with your suggestions for improvement. Also do share this post with others.

---

I write a post roughly once in 6 weeks covering the areas of technology for India and Indian philosophy. If these topics interest you, you may subscribe to this free Substack.