DeepSeek AI: The Chinese Chatbot That Shook the Tech World

On January 20, 2025, the Chinese research lab behind DeepSeek AI launched DeepSeek R1, a chatbot that has outperformed some of the most advanced AI models in the world. It has surpassed OpenAI’s ChatGPT o1, Meta’s Llama, and Google’s Gemini Advanced in benchmarks related to math and reasoning. The shockwaves of this launch were felt across the global tech industry, impacting AI development and even financial markets.

This article explores DeepSeek’s capabilities, the technology behind it, and its implications for the AI industry.

What Makes DeepSeek AI Stand Out?

1. Superior Performance at a Fraction of the Cost

DeepSeek AI boasts a level of efficiency unmatched by its competitors. While OpenAI charges $200 per month for access to its ChatGPT o1 Pro model, DeepSeek is completely free for users. Additionally, the training cost of DeepSeek R1 was only $5.6 million significantly lower than the billions of dollars spent by companies like OpenAI, Google, and Meta on their AI models.

2. A Challenge to OpenAI’s Monopoly

When OpenAI’s Sam Altman was asked if foundational AI models could be developed with a limited budget, he responded dismissively, saying it would be “hopeless” to compete with OpenAI. DeepSeek has proved him wrong. Within a week of its launch, DeepSeek became the most downloaded app in the U.S., surpassing ChatGPT on both the App Store and Google Play Store.

3. Impact on Financial Markets

The release of DeepSeek AI had a massive impact on the U.S. tech industry, wiping out $1 trillion from the U.S. tech index. NVIDIA, the world’s most valuable company at the time, saw its valuation drop from $3.5 trillion to $2.9 trillion in a single day a loss of $589 billion, the largest in corporate history.

Who Created DeepSeek AI?

DeepSeek AI is the brainchild of Liang Wenfeng, a 40-year-old Chinese entrepreneur. Wenfeng, known for his low public profile, founded a hedge fund called High Flyer in 2015, using AI and mathematics for investments. In 2019, he established High Flyer AI to research artificial intelligence algorithms. By May 2023, he funneled his hedge fund earnings into the development of an AI model, motivated purely by scientific curiosity rather than profit.

Rather than hiring engineers, Wenfeng assembled a team of PhD students from China’s top universities, focusing on rigorous research and problem-solving.

How DeepSeek AI Works

1. Chain of Thought Reasoning: DeepSeek AI employs the Chain of Thought model, similar to OpenAI’s ChatGPT o1. This model mimics human reasoning by evaluating multiple counter-questions before finalizing an answer. Unlike traditional chatbots that provide instant responses, Chain of Thought AI models take time to “think” and analyze different angles before presenting an answer.
2. Mixture of Experts Method: Unlike OpenAI’s ChatGPT o1, which functions as a single model for all queries, DeepSeek AI uses a “Mixture of Experts” approach. This means it has specialized sub-models for different fields of engineering, medicine, and law. When a user poses a question, only the relevant sub-model is activated. This approach enhances efficiency and reduces computational costs.

Limitations of DeepSeek AI

1. Censorship Issues: A major downside of DeepSeek AI is its censorship policies. The model avoids politically sensitive topics related to China. Questions about events like the 1989 Tiananmen Square protests or criticisms of Chinese leadership receive a default “Sorry, I am not sure how to approach this question yet.” However, the AI is more open when discussing criticisms of Western leaders such as Joe Biden or Donald Trump.
2. Longer Response Time: Compared to ChatGPT o1, which takes around 31.15 seconds to respond, DeepSeek AI takes an average of 71.22 seconds. This delay is due to its increasing popularity, causing server overload and slower processing speeds.

Open-Source Model: A Game-Changer

One of DeepSeek’s most significant advantages is that it is open-source. Anyone can download its model and modify it to fit their needs. American companies like Perplexity AI and Microsoft have already integrated DeepSeek R1 into their platforms, removing censorship restrictions and optimizing its functionality.

This open-source approach contrasts sharply with OpenAI, which, despite its name, has shifted towards a closed-source model. Elon Musk has frequently criticized OpenAI, calling it “ClosedAI” for its lack of transparency.

AI Wars: The Battle for Technological Supremacy

The launch of DeepSeek AI marks the beginning of a new era in AI competition. The U.S. government has imposed export controls on NVIDIA’s H100 chips to prevent China from developing advanced AI models. Despite these restrictions, DeepSeek has managed to innovate using older computer chips, demonstrating that software advancements can compensate for hardware limitations.

As the AI landscape evolves, companies worldwide will have to adapt to new challenges posed by powerful, cost-effective, and open-source AI models like DeepSeek.

Conclusion

DeepSeek AI has shattered the notion that only well-funded Western tech giants can develop world-class AI models. With its open-source availability, superior reasoning capabilities, and cost efficiency, it presents a significant challenge to established players like OpenAI and Google. However, issues like censorship and response time need to be addressed for it to gain wider acceptance.

As AI competition intensifies, one thing is clear the AI revolution is just beginning, and DeepSeek has set the stage for a new global AI arms race.