The emergence of a relatively unknown Chinese AI app, DeepSeek, has sent a shockwave through financial markets and Silicon Valley with its recent release of cutting-edge AI models and prompted US President Donald Trump to identify it as “a wake-up call” for the US tech industry.
DeepSeek has appeared as the most downloaded free app in the US just a week after its launch period.
DeepSeek claims that its R1 Artificial Intelligence (AI) model, made at a fraction of the cost of its rival AI apps, has raised questions on the fate of the whole industry and caused some of the world’s biggest companies to plunge in value.
The Guardian reports, $1tn wiped off US stocks after Chinese firm unveils AI chatbot.”.
“The Chinese AI startup challenging US big techs, Forbes makes the headline.
ChatGPT, Google-owned Gemini, Grok, Microsoft-led Copilot, and Meta-owned AI LLaMA are all jointly experiencing a barren adventure of envy with Chinese AI, DeepSeek.
The Immersion of the US Nightmare
Liang Wenfeng, a prominent figure in both the hedge fund and AI industries, formerly opened a startup trading business from his home in 2015. Later, he comprehended the acridity of deep machine learning and invested his time and money towards it without any delay from 2018.
He bought more than 10k chips of Nvidia in 2020, during the COVID-19 pandemic. He applied them in his trading business.
Founded in May 2023 by Liang Wenfeng, DeepSeek functions independently but is solely funded by High-Flyer, a quantitative hedge fund also owned by Wenfeng. This classic funding model has allowed DeepSeek to pursue ambitious AI projects without the pressure of external investors, enabling it to be determined in prioritizing long-term research and development.
DeepSeek’s journey commenced with the release of DeepSeek Coder in November 2023, an open-source model designed for coding tasks. This was followed by DeepSeek LLM, a 67B parameter model aimed at competing with other large language models. DeepSeek-V2, launched in May 2024, accumulated significant attention for its strong performance and low cost, triggering a price war in the Chinese AI model market. This inconsistent pricing strategy forced other major Chinese tech giants, such as ByteDance, Tencent, Baidu, and Alibaba, to lower their AI model prices to remain competitive. DeepSeek-V2 was succeeded by DeepSeek-Coder-V2, a more sophisticated model with 236 billion parameters.
Mechanism behind climbing to the crest
How Deepseek outperformed the AI rivals from the US stock market is just ‘Token Call.’. ‘Token Call’ refers to the term or root words an AI can analyze.
ChatGPT trained on a dataset as large as 2 trillion tokens. For Gemini, the number is around 1 trillion; Grok is almost similar to them, where DeepSeek has trained at least 14.8 trillion tokens, almost 4 times larger than US AI tech giants.
To train them with Token Call GPO is a must.US tech groups use the A100 GPO and H100 GPO models of Nvidia, while Chinese rivals use the H800 GPO cluster instead, which might be a clone copy of the Nvidia chip or made using its prototype.
DeepSeek proves American hegemony wrong
America previously claimed that the leadership of AI would probably be bound in their grip and wouldn’t be left any scope to go another. That’s why the US banned selling their AI chip to at least 120 countries, including China. Only 18 countries enjoyed the consumption of the AI chip.
But China has interestingly approved 40 AI models within the last six months as the global AI development race intensifies. All of them are seemingly being prepared to overwhelm the US and other Western collaborators.
In a recent statistic on a benchmark performance of the AI tools, Deepseek topped others in five disciplines (including general knowledge, programming, complex reasoning, coding, and brainstorming) out of six. It is almost out of reach in this discipline, maintaining a far distance with ChatGPT and Grok. In those disciplines, only Gemini is trying to compete in the race. It is only a little bit behind Gemini in the discipline of math solving, scoring 83% by 94.1%.
For a fun test, we’ve tried to delve into it practically and found a crystal image as well.
DeepSeek is quite well as a research assistance as it provides just key reasons regarding the questions and explains the clues only without long elaborations wherease ChatGpt provides elaborative key points.Besides, while presenting citations, ChatGPT is identified as inaccurate in some cases.
Javier Aguirre, an AI researcher at Samsung Medical Center in Seoul, South Korea, specializes in researching AI and wrote in a post on LinkedIn on Tuesday, “I am quite impressed with Deepseek. While coding, we usually try to push AI chatbots to the limit to assess their capabilities in assisting with coding.
Today I had a really puzzling and complex problem. Even ChatGPT o1 was not able to reason enough to solve it. I gave a try to Deepseek, and it solved the problem at once and straight to the point.”
DeepSeek’s Future
DeepSeek’s emergence as a cacophonous force in the AI landscape is worth evaluation. Its innovative techniques, cost-efficient solutions, and optimization strategies have challenged the imperium and forced established players to re-evaluate their approaches. While DeepSeek faces challenges, its pledge to open-source cooperation and efficient AI development has the potential to reshape the future of the industry. As the AI race intensifies, DeepSeek’s journey will be one to experience closely.
