DeepSeek—what is it? What differentiated ChatGPT?

"Exploring DeepSeek: How It Stands Out from ChatGPT, Its Rising Popularity, and the Vision of Founder Liang Xuanfeng"

DeepSeek: What is it? What differentiates ChatGPT?

A summary of the reasons behind DeepSeek's popularity and the background of Liang Wenfeng, its author. DeepSeek is what makes ChatGPT unique. A brief review of the reasons behind DeepSeek's popularity and the background of Liang Wenfeng, its creator

DeepSeek R1, its most recent artificial intelligence reasoning model, appeared. It boasts significantly lower cost and claims to be on par with OpenAI o1 published in the United States at the end of last year. It has piqued curiosity all around. DeepSeek is popular in Hong Kong and the Mainland and even tops download lists for the US App Store and Play Store.

Sing Tao News aggregates material from different sources to assist readers grasp what DeepSeek is, what OpenSource is and its advantages and drawbacks, who Liang Wenfeng, the founder of DeepSeek, is, and how DeepSeek will affect the direction of artificial intelligence development, market impact, etc.

DeepSeek is what?

Founded on July 17, 2023, DeepSeek is a Chinese artificial intelligence startup that develops powerful large language models (LLMs) and related technologies.

Starting the open source reasoning model DeepSeek R1 on January 20 this year, DeepSeek exceeded the most recent model, "o1," presented by OpenAI in the United States with its low cost and outstanding efficiency.

OpenSource refers to what?

DeepSeek R1 promised to synchronize the open-source model weights upon release. DeepSeek claimed that this open-source repository (containing model weights) will adopt the standardized and relaxed MIT License everywhere, which is open source and free from constraints, thereby supporting the growth of the open-source community and industrial ecology. It is not required for business applications.

Unlike OpenAI, DeepSeek-R1 follows an "Open Source" open source design, that is, open code, allowing users to use and enhance it at will, according to adjunct professor Mei Liheng of the School of Business at the University of Hong Kong.

Though open source design lets everyone—including governments, companies, or even criminals like those in "KK Park"—use it whenever they want, Mei Liheng pointed out that Open Source also has disadvantages.

What sets DeepSeek apart from ChatGPT?

DeepSeek-R1 claims that its performance is on par with the official OpenAI-o1 version. It points out that DeepSeek-R1 extensively applied reinforcement learning technologies in a post-training stage, significantly improving the reasoning capacity of the model with somewhat limited labeled data. Its performance in tasks including mathematics, programming, and natural language processing matches the official OpenAI o1.

DeepSeek R1 invoices OpenAI at a rate 96% higher than the previous amount, which is also 96% lower.

DeepSeek-R1 API charges are $0.14 (cache hit) / $0.55 (cache miss) per million input tokens and $2.19 per million output tokens. By comparison, the OpenAI o1 API service runs $7.50 (cache hit) / $15 (cache miss) per million input tokens and $60 per million output tokens.

DeepSeek R1 costs $2.19 per 1 million outputs, as calculated by the output API price; OpenAI o1 costs $60, so DeepSeek R1 is 96% less than OpenAI o1.

The man behind DeepSeek, Liang Wenfeng

Liang Wenfeng launched leading quantitative private equity firms DeepSeek and Huanfang. His frontiers of entrepreneurship span finance and artificial intelligence. His DeepSeek organization was only formally founded 1.5 years ago. Huanfang Quantitative Fund manages over 10 billion yuan; in 2021, it once topped 100 billion yuan. Also, he is one of the few large AI models in China with a 10,000-card cluster computing capacity. Many overseas artificial intelligence rivals have also been startled by the company's R&D capability and successes.

Admitted to Zhejiang University, the college admission exam top scorer graduated with a master's degree.

1985 saw Liang Wenfeng born in Wuchuan, Zhanjiang. He scored well on the local college admission test and was admitted to Zhejiang University. He was accepted to Zhejiang University and scored top on the local college admission exam. Artificial intelligence fascinates him. Using technology, he quantified trading for his Master's degree. After graduating, he started "Huanfang Quantitative" and supervised mainland quantitative private equity funds and was intrigued by artificial intelligence. For his Master's degree, he quantified trading using technology. After graduating, he founded "Huanfang Quantitative" and oversaw mainland quantitative private equity funds. Domestic media sources claim that Liang Wenfeng is exceptionally brilliant in mathematics and has been a top student from early life.

Liang Wenfeng once noted that while for programs to make decisions is a science, for individuals, making investing decisions is an art. Liang Wenfeng further pointed out that China's artificial intelligence cannot constantly be in a follower role. "We often argue that Chinese artificial intelligence lags one or two years behind the United States, but the actual difference is innovation vs imitation. Should this not change, China will always be a follower, so some research is unavoidable.