Why Can China Build DeepSeek and India Can't? An Indian AI Expert Reveals the Shocking Truth
New Delhi (The Uttam Hindu): An Indian AI expert sheds light on why China managed to develop DeepSeek, while India struggles to build similar AI models, despite having a vast pool of software engineers. The recent release of DeepSeek R1, a Chinese AI model has disrupted the global AI landscape. This achievement highlights that Chinese companies are now capable of developing foundation models that rival those of OpenAI’s GPT-4, Anthropic’s Claude and Google DeepMind’s Gemini. While Silicon Valley is still absorbing the shock, a pressing question is emerging closer to home: Why has India not been able to develop AI models at the same scale as China and the US?
Indian-origin American author Sadanand Dhume recently took to social media to pose the question: "What’s the best explanation for why China can create DeepSeek but India cannot?" A response came from a user named "GDP" on X who works in the artificial intelligence field at Amazon. According to this expert, India’s lack of major AI foundation models stems from the absence of a protected market, which is crucial for the growth of local companies. Unlike India, China’s government treats AI development as a national priority and offers strong support.
The expert pointed out that India needs a robust, government-backed initiative, similar to the US's DARPA program to stand a chance in the global competition. In the realm of AI, the creation of Large Language Models (LLMs) is not as complex as it seems. The expert explained that the foundational science behind LLMs, based on Transformer Decoder architecture has been around since 2017. Although there have been minor advancements such as flash attention and Mixture of Experts (MoE), these improvements are readily accessible through research papers. According to the expert, LLMs are not beyond the capabilities of Indian engineers, even those based in India. However, the core issue lies in the competitive landscape. The expert compared India’s AI industry to the absence of Indian tech giants like Google or Facebook. Without a protected market to support early-stage growth, Indian startups are often outmatched by American service providers who have a cost advantage.
This is not the case in China, where the government shields its tech companies and leadership views AI development as vital to national interests. The expert also emphasized that the release of DeepSeek is just the beginning of China’s advancements in AI. He pointed out that China has several more AI models in the works, including Qwen (from Alibaba), MiniMax, Kimi and DuoBao (from ByteDance). China is said to have over 10 AI labs comparable to OpenAI and another 50 tier-2 labs, positioning it as a formidable competitor in the AI space.
