On June 7 last year, Sam Altman, the CEO of OpenAI, was asked this question when he was in India: Can Indian entrepreneurs compete with OpenAI, or should they even try? His answer, a blunt and definitive "No", was jarring to most of us. Altman followed with an even more fundamental question of his own: was India even a player in this massive technological shift.?
While Altman may have a point, he is wrong in assessing the depth of talent that India has and the potential it holds to contribute to the AI revolution. India is a country with more than 5,000 years of sustained human history, which has witnessed massive wars, famines, occupations and diversity. It holds the civilisational knowledge and accumulated thought processes that can and will play a critical part in building a safe and sustainable future for a man-machine society. India can act as a crucial counterbalance to an AI framework developed in the echo chambers of San Francisco.
Silicon Valley's Flawed AI Playbook
In typical Silicon Valley fashion of moving fast and breaking things, OpenAI has chosen to open up the AI revolution to everyone. ChatGPT was developed by training AI on the vast and free human knowledge available on the internet. We have Elon Musk unleashing his bots developed on Twitter data, and humanoids on the shop floors of Tesla. Mark Zuckerberg is unloading billions on AI development and computing, while releasing large open-source models. Amazon is buying power utilities in gigawatts as the Artificial General Intelligence (AGI) race gets real.
The GPU-guzzling, greed-driven AI boom is already scary, and adding fuel to this are the geopolitical dynamics between the US and China. The world is set to witness an unprecedented "Intelligence explosion". And we as a society probably aren't ready for such a rapid transition shaped purely by the Silicon Valley playbook.
Personal Stake In AI's Future
As a parent of two young kids, I think selfishly. While I have been dabbling with AI for more than 15 years, I want my kids to grow up in a world that is truly open to exploring human potential. It was a rainy afternoon in Bengaluru in 2008 when I came across an article by Ray Curzweil, which said that the sum total of all AI would be more than human intelligence by 2029. The claim shook me; I realised that this would be a game-changing moment for humanity, primarily because we are unsure of what lies beyond that point - a 'Singularity' - and the shape society would take after that. Soon after, I made the crazy decision to resign from my well-paying job at Wipro to start India's first AI chatbot company, 'Vimagino', with three friends and fellow IIT Kanpur alumni. Some remnants of this chatbot demo can still be found on YouTube.
Mind you, this was two years before OpenAI was born (even before the pre-deep-learning era), and it was all done in India. We even went live on two websites, Indiatimeshopping.com and Yatra.com. Unfortunately, we couldn't grow because not many back then practised risk-funding that could back such companies in India. Perhaps we were a little too early (or were we).
I eventually ended up co-founding pi Ventures, an early-stage venture fund, with Manish Singhal. The idea was to back risky AI bets in India, one of which is tracking well today and is on its way to becoming a revenue-making AI unicorn - a rarity in this GenAI era. However, I can see that while these AI companies from India could do well, eventually, for-profit companies would have to optimise functions like revenue and, of course, profit. It rankled me that if for-profit businesses made their way to AGI, the pursuit of profit might lead to AI ignoring humanity. We could end up being just dressed-up animals in a cage, well-fed with a universal basic income, but, in all actuality, complete non-players in the grand scheme of things. That is a very depressing scenario. So, I stepped back from pi and helped co-found ARTPARK (AI & Robotics Technology Park), a not-for-profit organisation whose motto is to make AI work for humanity.
Languages, 'Shruti', And 'Smruti'
ARTPAR, supported by a Rs 230 crore seed grant from the Department of Science & Technology and the government of Karnataka, is a unique effort incubated by the Indian Institute of Science (IISc). The organisation recently released 16,000 hours of spontaneous speech from 80 districts under project Vaani, open-sourced under Bhashini. Such initiatives can help in the development of geographically anchored models in a country like India, which is packed with diverse languages and local dialects that sometimes change in a matter of a few kilometres.
These languages are rivers of knowledge, overflowing with rich human intelligence and enriched by centuries of usage, feedback, and historical chronicling of local events. This longitudinal data of human effort spanning thousands of years, trying to describe the world around, is a living, thriving body of human, annotated knowledge.
There's another unique thing about India: a lot of its collective knowledge exists in "shruti"(in voice) and "smruti"(in memory) as centuries of invaders destroyed its written knowledge and manuscripts. This method even allowed the Upanishads to survive over time. For centuries, this knowledge has been practised as living traditions, weaved into the daily lives of billions. The physicist Ervin Schrödinger, who worked on the nature of reality and is known for his Schrödinger's cat experiment, reportedly carried a translation of the Upanishads and revered its core insights about the nature of apparent duality. Indeed, its unique views on self-knowledge, the nature of reality and consciousness can help chart a new way of thinking about safe AI as we race towards the so-called 'AGI'.
The 3 Stages Of AI Development
There are three stages that mark the learning and development of AI.
- Stage 1 is learning from freely available data on the internet, which resulted in ChatGPT and other such systems powered by LLMs. It has become useful in creating text and solving other problems, though it's still not alarmingly influential.
- Stage 2 is about creating AI agents, that research, co-work with and co-learn from human feedback. It's here that India, with a population of 1.4 billion brought up on a rich diversity of thought and action, can play a big role. We are expected to get to some aspects of AGI at this stage.
- Stage 3 is all about learning and augmenting from the physical and digital world. This is going to be tricky and critical, and we will have to be careful about what AI ends up learning, because it will soon power 40-50% of the workflow in real-life situations. It's this stage that will truly power a man-machine society of 7.5 billion humans working with more than 30 billion intelligent machines. Unlike human beings, biases in AI technology can grow exponentially, and in a much shorter span of time, humbling creators if they are not careful.
India Must Take The Lead
The US, with its powerful internet and social media giants driven by profits, may end up creating a path to AI that puts profit ahead of humanity. China, which is prone to imprinting its central leader's thinking onto algorithms to exact control, may power a future where AI might rule over human beings. Both scenarios are not sustainable. On the other hand, India, with its rich, sustainable, civilisational history, has the potential to shape a future where humans can realise their full potential, together with AI.
Sam Altman might have been wrong about India's potential contribution, but he was spot-on about the inadequate resources we are putting into becoming an AI force to reckon with at a global level.
The India AI mission is a great start, but we need to double down with a sizable private capital pool and AI compute resources that can support our talented technologists and researchers. We need to enable them at warp speed to digitize and mine the rich civilizational knowledge India possesses for building AI.
Otherwise, our kids may not like the 'Matrix' future we may leave for them. Maybe it's time for India to take the 'Red Pill' and pull up its socks.
[Umakant Soni is chairman of AIfoundry, co-founder of ARTPARK (AI & Robotics Technology Park), and a passionate advocate for inclusive development]
Disclaimer: These are the personal opinions of the author