What is DeepSeek? And why are everyone talking about it?

The DeepSeek, a Chinese artificial intelligence (AI) model, was shot by the Apple Store download, stunning investor, and sinking some high -tech shares.

The latest version was released on January 20 and immediately impressed AI experts, attracting attention in the technology industry and the world.

US President Donald Trump said it was a “morning call” for US companies who had to focus on “competing to win.”

The company’s argument that DeepSeek is very special is that it was built at a paid cost of a model that leads the industry like Openai. This is to use less advanced chips.

As a result, the huge chip production NVIDIA reduced the market value of about $ 600 million (£ 48.2 billion) on Monday, the largest loss of US history.

DeepSeek also raises the question of Washington’s efforts to promote the advantage of technology, given that one of the important restrictions is prohibited from exporting high chips to China.

However, Beijing has doubled, and President Xi Jinping has declared AI as a top priority. For emerging companies like DeepSeek, pivot is important because of traditional high -tech, high -tech, electric vehicles, and AI manufacturing from traditional manufacturing, such as clothing and furniture.

So what do you know about DeepSeek?

AI can sometimes make your computer look like a person.

Machines are usually trained with a huge amount of information, and by recognizing the patterns, you can learn and solve problems using technology.

The final result is software that allows people to talk or predict shopping habits.

In recent years, it is the best technology behind chatbots such as Chatgpt -and DeepSeek. It is also called generated AI.

These programs are learned again from online texts and vast amounts of data, including images, so that new content can be created.

However, these tools can create falsehood and often repeat bias contained in training data.

Millions of people use tools such as Chatgpt to support everyday tasks, such as writing emails, summary of textbooks, answers to questions.

DeepSeek is the name of a free AI chatbot, which looks very similar to Chatgpt, and works.

In other words, it means that it is used in many of the same tasks, but there is room for debate how well it works compared to rivals.

According to a task such as mathematics and coding, it is as powerful as the O1 model of OPENAI released at the end of last year.

Like O1, R1 is a “reasoning” model. These models simulate the same process as human beings and ideas to infer the process, and generate a step -by -step response. Use less memory than rivals to ultimately reduce costs and execute tasks.

Like many other Chinese AI models, Baidu’s Ernie or DOUBAO BYTEDANCE -DEEPSEEK is trained to avoid politically sensitive questions.

When the BBC asked the app on June 4, 1989 what happened in Tiananmen Square, Deepseek did not provide details on the Chinese tabut pick.

“I’m sorry, I can’t answer that question. I’m an AI assistant designed to provide a kind and harmless answer.”

Censorship of the Chinese government is a major issue for AI’s desire internationally. However, the basic model of DeepSeek seems to have been trained through accurate sources, introducing a layer of censorship and refraining from specific information through an additional protection layer.

DeepSeek says that this was able to do this cheaply -the researchers behind it, a training that is a part of “$ 100 million or more” suggested by Openai boss Sam Altman when discussing GPT -4. It claims that it costs $ 6 million (£ 4.8 million).

It is reported that DeepSeek founder has built a NVIDIA A100 chips store that has been banned from exports to China since September 2022.

Some experts believe that this collection has reached 50,000, which led these tips to build such a powerful AI model in combination with inexpensive and unsuitable chips.

On the same day, DeepSeek’s AI Assistant was the most downloaded free app on the Apple Apple App Store in the United States, and was temporarily restricted by the “large -scale malicious attack”.

I was also hit by the website on Monday.

DeepSeek was established by Liang Wenfeng in December 2023 and released the first major AI language model the following year.

Liang, who graduated from the University of Zhijiang with a degree of electronic information engineering and computer science, is not well known. But he is now finding himself in international spotlights.

He has recently been seen at a meeting hosted by China’s finest LI Qiang, reflecting the rise in Deepseek in the AI ​​industry.

Unlike many American entrepreneurs from Silicon Valley, Liang also has a financial background.

He is a hedge fund CEO called High -Flyer, which is an analysis of financial data using AI and invests in investment. In 2019, HIGHFLYER became the first quantum hedge fund in China to raise 100 billion yuan ($ 13 million).

He said in his speech that year, “Why wouldn’t it be China if the United States could develop its quantitative trade department?”

In a rare interview last year, he said that the Chinese AI sector could not follow it forever.

“In many cases, there is a gap of one or two years between China and the United States, but the real gap is between originality and imitation. If this does not change, China is always a follower. “”

He said why DeepSeek’s model surprised many Silicon Valley, as follows:

The Australian Minister of Science has raised some questions about the security of the app.

“There are many questions that need to be answered in time for quality, consumer preferences, data, and privacy management,” ED HUSIC told ABC.

“I am very careful about it. I need to carefully weigh these types of problems.”

Deepseek’s achievements have overturned the belief that a larger budget and the highest -layered chips are the only way to advance AI. This is the prospect that created uncertainty about the future of high -performance chips.

“DeepSeek proves that the state -of -the -art AI model can be developed with a limited calculation resource,” said Wei Sun, a principal AI analyst in AutOunt Point Research.

“In contrast, Openai, which has been evaluated as $ 157 billion, monitors the ability to maintain the dominant advantage of innovation and justify its large -scale evaluation and expenditure without a significant profit. I’m facing. ”

The company’s low cost has been financial market on January 27, reducing high -tech NASDAQ by more than 3 % with a wide range of sales, including tip manufacturers and data centers around the world.

NVIDIA seems to have reached the worst record because the stock price had plummeted 17 % on Monday and began to recover slowly on Tuesday, so the stock price was about 4 %.

CHIP MAKER was the world’s most valuable company when measured by market capitalization, but fell third after Apple and Microsoft on Monday.

DeepSeek is a personal owned company. In other words, investors cannot purchase shares on major exchanges.

The rise of DeepSeek is a major boost for the Chinese government, which is trying to build independent technology from the western part.

The Communist Party has not yet commented, but the Chinese state -owned media was eager to say that Silicon Valley and Wall Street giants were “sleeping”.

“In China, Deepseek’s progress is celebrated as a proof of the country’s growth technical abilities and independence,” said Marina Zhang, Associate Professor of Sydney Institute of Technology.

“The company’s success is considered to be a verification of China’s innovation 2.0, a new era of homemade technical leadership promoted by young entrepreneurs.”

But she warned that this emotion could lead to “technical isolationism.”

Additional reports by Joan da Silva