Loved Dr K’s article, however I was planning to write this for a while.
Date: 02-02-2025 at 15:10
If all the claims made by the Chinese AI lab – Deepseek – are true, we have some hope of becoming part of the AI superpower in the next 10 to 15 years. In my brief six-year journey, I have witnessed trends, booms, busts, economic crises, inflation, and more. Most of these phenomena are driven by macro events; however, today’s macro tsunamis are mostly created by a group of AI researchers sitting in protective parts of the world – the USA and China. These macro tsunamis feel like history being written every single day.
Histories are not written in days, months, or years; it takes decades. But considering most headlines and successes feel like overnight achievements, we humans tend to feel anxiety and make drastic decisions.
It feels like India is experiencing the same anxiety after the shockwave created by the Chinese AI lab – Deepseek on January 20th, 2025. Just before DeepSeek’s Foundation Reasoning Model, Deepthink R1, most Indian pundits had concluded that India does not need its own foundational LLM model but took a U-turn and started advocating for one.
This sudden change in perspective is due to the possibilities claimed by Deepseek – only $6 million to train a foundational model. Considering Sam from OpenAI mocked the question about the possibility of a foundational model within $10 million, emotional Indians took that as a challenge.
To me, the current anxiety in India regarding the foundational model feels similar to Sam’s comment, except for one difference: I see a lot happening on the ground and a push by the government. This is both good and bad, which I will explain in the rest of the essay.
First, we need to understand why today’s success of Deepseek is the result of a decade of work and what we Indians should learn from that. Kai-Fu Lee, in his extraordinary book AI Superpowers, predicted the same but five years in advance; the book was published in 2020. He called DeepMind’s AlphaGo beating Go Champion Lee Sedol in March 2016 a “China’s Sputnik Moment” because, apart from the live match being watched by 260 million people in China, it also ignited the nation with AI fever.
The entire Chinese ecosystem – Chinese investors, entrepreneurs, and government officials – brought their concentrated focus to a single industry: Artificial Intelligence. Money for AI startups poured in from venture capitalists, tech juggernauts (corporates and industry leaders), and the Chinese Government. Chinese students enrolled in advanced degree programs and streamed lectures from international researchers on their smartphones. Startup founders began pivoting, reengineering, or simply rebranding their companies to catch the AI wave.
By July 2017, the Chinese central government unveiled an ambitious plan to build artificial intelligence capabilities, setting clear benchmarks for progress by 2020 and 2025, projecting that by 2030, China would become the centre for global innovation in artificial intelligence, leading in theory, technology, and applications.
Therefore, Deepseek’s 2025 model is just the benchmark set by China’s ecosystem in 2017 – a decade of work with a collaborative approach. Does this mean China has better-quality Deep Learning researchers than the USA, or have they innovated foundational algorithms? The answer to both questions seems to be no, as even the DeepThink R3 research paper talks about utilizing OpenAI’s foundational model.
In that case, the real questions we should ask ourselves are: If discovery and super high-quality talent are not prerequisites for the success of creating models like DeepSeek, then what foundations can India leverage to position itself as an AI superpower in the next 10 to 15 years?
History holds the answers to these questions. Electricity was discovered many decades before Thomas Edison harnessed it into a practical application (electricity into light) and made it widely available for home and business applications. Once harnessed, electricity revolutionized dozens of different industries. In the nineteenth century, entrepreneurs soon began applying the electricity breakthrough to cooking food, lighting rooms, and powering industrial equipment. It was a transition from discovery to implementation, creating trillions of dollars of economic value for nations that focused on implementation.
Steve Jobs’ implementation of the Graphical User Interface (GUI) across different use cases is another similar example. The invention of the GUI was not done by Steve Jobs; it was developed by a company named Xerox. The story goes that initially, Steve was reluctant to attend the presentation, but today, we all know the history – every digital interaction happens through the GUI.
Therefore, history is a testament to the fact that value creation through new technology is independent of discovery but highly dependent on implementation. And to implement, elite talent (in the case of AI – elite AI researchers) is not a prerequisite.
However, if we examine the foundational models, two key components that made them functional are compute and data. While computing is a specialized area, data is something that can reduce the need for expertise. Therefore, another transition would be from the age of expertise to the age of data – more data means better AI algorithms.
Thus, a successful LLM model can be created using big data, computing power, and work by strong – though not necessarily elite – AI algorithm engineers.
By keeping the above points in mind, if we draw a direct parallel with the harnessing of electricity, we can identify our strengths and weaknesses and, by playing to our strengths, become an AI superpower in the next 10 to 15 years.
| 4 inputs that made the electrification possible | 4 inputs to harness the power of AI (electricity of the 21st century) |
|---|---|
| Fossil fuels to generate it | Abundant data |
| Entrepreneurs to build new businesses around it | Hungary entrepreneurs |
| Electrical engineers manipulate it | AI scientists |
| Supportive government to develop underlying public infrastructure | AI-friendly policy environment |
Today’s Chinese success in the foundational model is based on minimizing weaknesses – in this case, the lack of elite AI scientists – and amplifying strengths: Data, hungry entrepreneurs, and AI-friendly policy support from the Chinese Central Government.
Read the full article: https://www.sumanjha.com/post/a-potential-and-feasible-path-for-india-to-become-ai-superpower-in-the-next-10-to-15-years