What is DeepSeek?

DeepSeek is a China-invented AI model which represents a paradigm shift in the development of Artificial Intelligence Models. As the Chinese paper which developed DeepSeek said, “Our goal is to explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure reinforcement learning (RL) process”. Traditionally, Large Language Models (LLMs) require supervised fine-tuning with human assistance to develop reasoning. DeepSeek has replaced supervised learning with purely computer-assisted “reinforced learning”

DeepSeek was invented by Liang Wenfeng, CEO of DeepSeek. Based in Hangzhou, Zhejiang, it is owned and funded by the Chinese hedge fund High-Flyer. It is interesting that a hedge fund dealing with stock markets invested in AI research. DeepSeek has been officially acknowledged and supported by the Chinese government. DeepSeek has sent shock waves to the Western AI community. They have been surprised that Chinese researchers could come up with a paradigmatic shift in AI modelling. Even greater surprise was the decision by the Chinese developer to make it open-source, available to anyone. This has begun a race amongst developers to use DeepSeek to develop new models and AI applications.

What does it Imply for the Future of AI?

DeepSeek opens a new vista for AI development. It ‘distills’ large Language Models (LLMs) into smaller models. This saves time and energy. It will speed up the development of AI applications. It has demonstrated that large language models can be trained using computer resources with less computing power and no human intervention. Several big and small companies around the world are likely to use DeepSeek to develop more refined, more advanced models. Thus, the goal of Autonomous AI will be easier to achieve. A lot of investment is likely to go into developing DeepSeek-like models and applications. DeepSeek is adept at mathematical reasoning, code writing and analytical thinking. Models like ChatGPT are more proficient at verbal reasoning compared to DeepSeek. Thus, ChatGPT and DeepSeek can complement each other. DeepSeek’s features have tremendous applications in health, manufacturing, mathematics, code writing and so on. DeepSeek-based AI will also have applications in strategic technologies, including defence and security.

What does it Mean for the US-China Tech Race?

The US officials and AI companies have been taken by shock and awe by the Chinese innovation. President Trump has described DeepSeek as a “wake-up” call. The commentary coming out in US media and academic circles is that the Biden era 2022 export controls prohibiting the export of advanced chips to China have not worked. NVIDIA, the manufacturer of advanced AI chips, apparently circumvented export controls by manufacturing H800 chips, which are the less advanced versions of H100 chips, but still represent high computing power. It was complicit in exporting or smuggling them to China.

The American analysts say that DeepSeek's computing power was based on smuggled American chips. There may be some truth in that. The Chinese sources themselves admit that millions of NVIDIA chips worth USD 6 billion were imported into China during 2022. 2023. However, this does not take away from the Chinese achievement. The China AI strategy declared in 2017 had said that China would develop world-class Large Language Models (LLMs) and AI products by 2025. This is what has happened. In the race for technological dominance, any means will be used.

The AI race between the US and China is hotting up. The Trump administration has set up a “Council of Scientific and Technological Advisers” to come up with forward-looking policies to retain AI American dominance in high technologies, including AI. In China, big companies are setting up of consortiums to develop AI.

The Americans still have an edge in AI because they have access to high-performance computing power and funding. Although China’s DeepSeek was the first to announce its model, several American companies have also been working on similar lines. We can expect American AI giants to come up with their DeepSeek-like models or even better ones. With the DeepSeek model having become open source, smaller developers could also join the race and develop more efficient AI models. Thus, the AI race has just opened with DeepSeek in 2025.

What does it Mean for India?

If DeepSeek is a wake-up call for the US, it is more so for India. It needs to be realised that DeepSeek represents an escalatory moment in the deepening technology race, which will inevitably have strategic consequences for the economy, defence and security.

India risks marginalisation in the AI race unless it puts its act together quickly. It needs to invest in Large Language Models, high-performance computing, and building skills and human resources. What is likely to happen is that with DeepSeek becoming an open-source model, Indian developers will use it to develop applications. But India needs to step up fundamental research in LLMs and Reinforcement Learning.

Where does India Stand?

The government of India launched the India AI Mission in 2024 with Rs 10,300 crore budget. As a part of this mission, India AI Mission, India is building a domestic LLM of its own. It seeks to develop its own GPU in the next three to five years through the National Semiconductor Mission, and a domestic foundational AI platform can be expected in the next 10 months.

The key to AI is high-end computing. The government is procuring 18693 high-end GPUs to bolster the compute capability in the counting. It has selected 10 companies which will supply these chips. Nearly half of the total GPUs will come from Yotta alone, who has committed to offer 9,216 units. These GPUs are being procured by the government under the India AI Mission, which was sanctioned Rs 2,000 crore for 2025-26 (a fifth of the scheme’s total outlay of Rs.10,371.92 crore), as per the Union Budget 2025 proposed by Finance Minister Nirmala Sitharaman on 1 February 2025.

Under the India AI Mission, the government has selected 18 AI-based apps focussed on areas such as agriculture, learning disabilities, and climate change to provide the first round of funding.

Conclusion

India’s AI mission is development-oriented. This is understandable as AI has huge applications in health, agriculture, education, climate change, etc. The government has identified eighteen areas for AI applications. What is required is the focus on defence and security. AI is akin to nuclear technologies. It is dual use. AI will revolutionise warfare. As India implements its AI, Quantum and Cyber and semiconductors missions, the requirements of national security must be built into them right from the beginning. There is a need to synergise the AI and Quantum technology missions. We are fortunate that we generate a lot of data. We should also focus on building our own high-end computing capability without that. We will be only users and not shapers of AI. We need to invest in fundamental research and R&D, besides developing applications.

(The paper is the author’s individual scholastic articulation. The author certifies that the article/paper is original in content, unpublished and it has not been submitted for publication/web upload elsewhere, and that the facts and figures quoted are duly referenced, as needed, and are believed to be correct). (The paper does not necessarily represent the organisational stance... More >>