What is DeepSeek? The ChatGPT Rival

DeepSeek is a free AI-powered chatbot that operates similarly to ChatGPT. It’s designed to handle a variety of tasks, though opinions differ on how it stacks up against competitors. Reports suggest it matches the capabilities of OpenAI’s o1 model, especially in areas like mathematics and coding.

Like o1, DeepSeek is classified as a "reasoning" model. This means it generates responses step-by-step, mimicking the way humans think through problems. One of its advantages is that it uses less memory than some of its rivals, which helps keep operational costs down.

However, DeepSeek, like several other Chinese AI models such as Baidu’s Ernie or ByteDance’s Doubao, is programmed to steer clear of politically sensitive topics. For instance, when the BBC inquired about the events at Tiananmen Square on June 4, 1989, DeepSeek refrained from discussing the massacre, stating: “I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses.” This reflects the significant challenge of Chinese government censorship in the realm of international AI development.

Despite these hurdles, DeepSeek’s foundational model seems to have been trained on reliable sources while implementing controls to withhold specific information. The team behind DeepSeek claims they achieved this training at a relatively low cost—around $6 million—compared to the "over $100 million" mentioned by OpenAI’s CEO Sam Altman regarding GPT-4.

The founder of DeepSeek reportedly amassed a collection of Nvidia A100 chips, which have been banned for export to China since September 2022. Some experts suggest that this stockpile, estimated at around 50,000 chips, enabled the development of such a robust AI model by combining these advanced chips with less expensive, simpler alternatives.

On the same day that DeepSeek's AI assistant became the most downloaded free app on Apple's App Store in the US, the company faced significant challenges, including "large-scale malicious attacks." As a result, they temporarily limited new registrations. Additionally, their website experienced outages on Monday.

DeepSeek was founded in December 2023 by Liang Wenfeng, who launched the company's first AI large language model the following year. While details about Liang are limited, he graduated from Zhejiang University with degrees in electronic information engineering and computer science. He is now gaining international attention.

Liang was recently seen at a meeting with China's Premier Li Qiang, highlighting DeepSeek's rising status in the AI sector. Unlike many American AI entrepreneurs who hail from Silicon Valley, Liang also has a financial background. He serves as the CEO of a hedge fund called High-Flyer, which leverages AI to analyze financial data for investment decisions, known as quantitative trading. In 2019, High-Flyer became the first quantitative hedge fund in China to raise over 100 billion yuan (approximately $13 million).

During a speech that year, Liang remarked, "If the US can develop its quantitative trading sector, why not China?" In a rare interview last year, he expressed that China's AI industry "cannot remain a follower forever." He noted that while there is often a perceived one or two-year gap between Chinese and American AI, the real issue lies in originality versus imitation. He emphasized that if this mindset doesn't shift, China will continue to lag behind.

When asked why DeepSeek's model surprised many in Silicon Valley, Liang replied that their astonishment comes from witnessing a Chinese company emerging as an innovator rather than merely a follower, which is the norm for many Chinese firms. TECH FYP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .