What is Deepseek? And how does AI stay?

Technological stocks fell. Giant companies such as Meta and Nvidia faced a flood of questions about their future. And technology executives went to social networks to proclaim their fears.

And everything was due to a new little -known Chinese artificial intelligence company called Deepseek.

Deepseek caused waves worldwide as one of his achievements, which had created a very powerful model with much less money than many experts in ai believed possible, he raised a series of questions, included if US companies were even competitive in ai. not anymore.

Deepseek is “Sputnik Moment of ai“, Marc Andreessen, a capitalist of technological risk, <a target="_blank" class="css-yywogo" href="https://x.com/pmarca/status/1883640142591853011″ title=”” rel=”noopener noreferrer” target=”_blank”>aware On social networks on Sunday.

How could a company that few people had heard such effect? This is what I should know about Depseek, its technology and its implications.

What is Deepseek?

Deepseek is a new company founded and owned by the trade firm of Chinese High-Flyer shares. Its objective is to build artificial intelligence technologies in the OpenAi or GEMINI of Google chatbot line. By 2021, Depseek had acquired thousands of computer manufacturer's computer chips

In China, the new company is known for catching researchers from young and talented of the best universities, promising high wages and an opportunity to work on avant -garde research projects. Both High-Flyer and Deepseek are led by Liang Wenfeng, a Chinese businessman.

In recent years, Deepseek has launched several large language models, which is the type of technology that supports chatbots such as Chatgpt and Gemini. On January 10, he launched his first free Chatbot application, which was based on a new model called Depseek-V3.

Why did the stock market react now?

When Depseek presented its Deepseek-V3 model the day after Christmas, it coincided with the skills of the best chatbots from US companies such as OpenAi and Google. That would only have been impressive.

But the team behind the new system also revealed a bigger step. In a research article that explains how he built technology, Depseek said he used only a fraction of computer chips in which the leaders of the leaders trusted to train their systems.

The main companies in the world generally train their chatbots with supercomputers that use up to 16,000 chips or more. Deepseek engineers said they only needed around 2,000 Nvidia chips.

Why is that important?

Since the end of 2022, when Openai triggered the rise of ai, the predominant notion had been that the most powerful ai systems could not be built without investing billions of dollars in specialized chips. That would mean that only the largest technological companies, such as Microsoft, Google and Meta, all of which are based in the United States, could be allowed to build leading technologies.

(The New York Times has sued Openai and his partner, Microsoft, claiming the infringement of copyright of the news content related to ai systems. The two technological companies have denied the claims of the demand).

But Deepseek engineers said they only needed around $ 6 million in unprocessed computer energy to train their new system. That was approximately 10 times than what Meta spent building his latest ai technology.

How did Deepseek?

The best ai engineers in the United States say that Depseek's research work established intelligent and impressive forms of building ai technology with less chips.

In summary, Startup engineers demonstrated a more efficient way to analyze data using chips. The systems of the leaders learn their skills identifying patterns in large amounts of data, including text, images and sounds. Deepseek described a way to spread this data analysis in several specialized models, what researchers call a “mixture of experts”, while minimizing the lost time by moving data from one place to another.

Others have used similar methods before, but move information between the models tended to reduce efficiency. Deepseek did this in a way that would allow him to use less computer power.

“It has been very clear that other companies, not only someone like Openai, can build these types of systems,” said Tim Dettmers, a researcher at the Allen Institute for artificial intelligence in Seattle and computer professor at the Carnegie Mellon Who Who University specializes in Build efficient ai systems. “Depseek used methods that anyone can duplicate.”

Deepseek's research work raised questions about whether the big American companies could maintain a significant advantage in ai, many experts believe that ai technology will become a merchandise, and many companies sell the same product.

Is Deepseek technology as good as OpenAi and Google systems?

Deepseek-V3 can answer questions, solve logical problems and write your own computer programs as effectively as anything that is already in the market, according to standard reference tests.

Just before Depseek launched its technology, Operai had presented a new system, called Openai O3, which seemed more powerful than Deepseek-V3. But Openai has not launched this system to the broader public.

Operai O3 was designed to “reason” through problems involving mathematics, science and computer programming. Many experts pointed out that Depseek had not built a reasoning model in this regard, which looks like the future of ai

Then, on January 20, Depseek launched its own reasoning model called Depseek R1, and also impressed experts. That finally sent us to investors and others to a panic at the end of last week and during the weekend when realizing the importance of the new Deepseek technology.

American technological giants are building data centers with specialized ai chips. Does this matter, given what Deepseek has done?

Yes, it still matters.

A large number of ai chips can still help companies in many ways. With more chips, they can execute more experiments as they explore new ways to build in other words, more chips can still give companies a technical and competitive advantage.

More chips will also be needed to operate the new race of “reasoning” models, experts said. These require more computer power when people and companies use them.

Has the United States limited the number of Nvidia chips sold to China?

Yes. To maintain the leadership of the United States in the global career of ai, the Biden administration had established rules that limited the number of powerful chips that could be sold to China and other rivals.

But the impressive performance of the Deepseek model raised questions about the involuntary consequences of the US government's commercial restrictions. The controls have forced researchers in China to be creative with a wide range of tools that are available for free on the Internet.

Some experts continue to argue in favor of US commercial restrictions, saying that they were recently established and that they will have a greater effect on China's skills to create the years as the years go by.

Does Deepseek technology mean that China is now ahead of the United States in ai?

No. The world has not yet seen OPENAI's O3 model, and its performance in standard reference tests was more impressive than anything else in the market. But experts are concerned that China is moving forward in open source ai systems.

What exactly is open source ai?

Like many other companies, Depseek has “opened” its last ai system, which means that it has shared the underlying computer code with other companies and researchers. This allows others to build and distribute their own products using the same technologies.

This is part of the reason why Deepseek and others in China have been able to build competitive systems so quickly and economically.

In the world of ai, the open source met for the first time in 2023 when Meta freely shared a call system. At that time, many assumed that the open source ecosystem would flourish only if companies such as Meta, giant companies with huge data centers full of specialized chips, continue to open their technologies.

But Deepseek and others have shown that this ecosystem can prosper in a way that extends beyond US technological giants.

Why is that important?

Many experts have argued that large US companies should not open their technologies because they could be used to spread misinformation or cause other serious damage. Some US legislators have explored the possibility of preventing or strangling the practice.

But other experts have argued that if regulators suffocate the progress of open source technology in the United States, China will obtain a significant advantage. If the best open source technologies come from China, these experts argue that US researchers and companies will build their systems about these technologies.

In the long run, that could put China in the heart of the research and development of ai, which could further accelerate its effort to build a wide range of ai technologies, including autonomous weapons and other military systems.