DeepSeek Enters Global AI Race

In 2011, Marc Andresen said, “Software is eating the world.” This came true when m-cap of Tesla overtook the next 10 car companies combined. In 2017, Jensen Huang, CEO of Nvidia predicted, “Software is eating the world, but AI is going to eat software.” ChatGPT reached 100 million users in two months. AI euphoria in the US suddenly seems to have turned to panic.

On 27 Jan, DeepSeek, a cost-effective Chinese AI-assistant overtook ChatGPT in terms of downloads from the US iOS app store threatening the US leadership in AI. It sent shockwaves through global markets, leading to 17% fall or $593 billion loss for Nvidia, the largest single day m-cap loss for a Wall Street stock in history. Nasdaq slumped >3%. Fears over costly US endeavours in AI tech against comparable Chinese alternatives – available on open source and made at a fraction of the cost – spooked markets.

China’s DeepSeek is outpacing Big Tech’s AI models in technical versatility and affordability. Founded in 2023 by Liang Wenfeng in Hangzhou, it reportedly took 2 months and $6 million to build, using Nvidia’s less-advanced H800 chips. As per media reports, their model seems to have outperformed Meta’s Llama, Open AI’s GPT-4 and Athropic’s Calude Sonnet 3.5.

DeepSeek is an open-source AI model, with its code publicly available and technical details explained in research papers. Some experts consider it a significant breakthrough in AI development. While OpenAI and Anthropic will continue developing their own proprietary models, open-source and commercial AI models may co-exist. An expert noted the focus has shifted towards efficiency in AI development, moving away from simply using more computing power. DeepSeek uses a “distillation” technique, employs fewer GPUs and has much lower training costs compared to the US Big Tech companies.

DeepSeek has its limitations. It appears to have pre-existing restrictions on topics considered sensitive by the Chinese government. The company reported a large-scale cyber-attack. “DeepSeek’s sudden rise has questioned the effectiveness of Washington’s efforts aimed at curbing China’s access to high-end tech over national security concerns,” says CNBC.

AI development suddenly seems to have become accessible. Now not only the BigTech, with massive investments, but others too can also play in AI. Instead of a billion-dollar data center, you could get good results with few GPUs. It will be ideal for start-ups and developers as it significantly lowers the cost of development.

The AI industry is changing rapidly. Previously dominated by large tech companies with substantial resources, the field may now see significant contributions from smaller start-ups like DeepSeek. This shift could lead to more widespread, cost-effective AI development in countries like China and India.

Disclaimer: The article has reference to expert opinions and open sources including Reuters, CNBC, Economist and X.

Know About Our Services