What is Deepsek? And how is it increasing AI?

Tech stock tumble. Massive companies like Meta and Nvidia faced questions about their future. Tech officials resorted to social media to declare their fear.

And all this was due to a small known Chinese artificial intelligence start-up called Deepsek.

Dipsek on Monday created waves around the world on Monday as one of its achievements – that it created a very powerful AI model with much less money than many AI experts, which was possible – hosting questions Ki, in which American companies were also competitive in AI that AI was also competitive. now and.

Deepsek “AI’s Sputnic moment,” Mark Andresen, is a technical enterprise capitalist, Posted On Sunday on social media.

A company that some people had heard of such an impact?

What is Deepsek?

Deepsek is a start-up that is owned by the Chinese Stock Trading Firm High-Flora. Its goal is to manufacture AI technologies on the lines of the Chatgpt Chatbot of Openai or Google. By 2021, Deepsek had acquired thousands of computer chips from US chipmaker Nvidia, which is a fundamental part of any attempt to create a powerful AI system.

In China, start-ups are known to grab young and talented AI researchers from top universities, promising higher pay and giving an opportunity to work on state-of-the-art research projects. Both high-blue and lampsakes are run by a Chinese entrepreneur Liang Wenfeng.

Over the years, Deepsak has released several big language models, a kind of technology that outlines chatbott such as chatbott and Mithun. On 10 January, it released its first free chatbot app, based on a new model called Dipsek-V3.

Why did the stock market react to this now?

When Deepsek introduced its Dipsek-V3 model after Christmas, it matched the capabilities of the best chatbott from American companies such as Openi and Google. He would have been effective alone.

But the team behind the new system also took a big step forward. In a research paper, stating how it produced technology, Deepsek said that it had used only one fraction of computer chips that major companies of AI companies had trusted to train their system.

The world’s top companies usually train their chatbots with supercomputers that use 16,000 chips or more. Deepsek engineers said they need only 2,000 NVidia chips.

Why is it important?

From the end of 2022, when Openi shut down the AI ​​Boom, the prevailing belief was that the most powerful AI system could not be created without investing billions of dollars in special AI chips. This would mean only the largest technical companies – such as Microsoft, Google and Meta, which are all located in the United States – can manufacture leading technologies.

(The New York Times has sued Openai and its partner, Microsoft, which claims copyright violation of news material related to the AI ​​system. Both technical companies have denied the claims of the suit.)

But Deepsek engineers said that they needed only $ 6 million in raw computing power to train their new system. Meta was about 10 times less than producing its latest AI technology.

How Dipsec made his technology with low AI chips?

Top AI engineers in the United States say that Deepsek’s research paper gave clever and effective ways to manufacture AI technology with low chips.

In short, the engineers of the startup displayed a more efficient way of analyzing data using chips. Learnings learn their skills by indicating patterns in large amounts of data, including leading AI systems text, pictures and sounds. Deepsek described a way of spreading this data analysis in several special AI models – which researchers have called the “mixture of experts” method – while taking data from place to place, reducing the lost time.

Other people have used similar methods before, but the information between models to reduce efficiency is increasing. Deepsek did this in such a way that it allowed it to use low computing power.

“It has become very clear that other companies, not only an OpenII, can manufacture this type of system,” a researcher of Alan Institute for Artificial Intelligence in Seattle and Professor of Computer Sciences at Carnegie Melon University Said. Efficient AI is an expert in creating system. “Deepsek used ways that anyone can duplicate.”

The Deepsek’s research paper raised questions about whether large American companies can maintain a significant lead in AI, many experts believe that AI technology will become a commodity, many companies are selling the same product.

Is the technology of Deepsek as good as Openai and Google’s system?

Deepseek-V3 can answer questions, resolve logic problems and according to standard benchmark tests, can already write anything effectively on the market.

Just before Deepsek released its technology, Openai unveiled a new system, called Openai O3, which looked more powerful than Deepsekeek-V3. But Openai has not released this system to a wide public.

Openai O3 was designed for “causes” through problems related to mathematics, science and computer programming. Many experts said that Deepsek did not create an argument model with these lines, which is seen as the future of AI.

Then on 20 January, Deepsek released his own argument model called Deepsake R1, and also affects experts. It finally sent us investors and others in panic at the end of the last week and in the weekend as they realized the importance of the new technology of Deepsek.

US tech giants are building data centers with special AI chips. Does it still matter, given what Deepsak has done?

Yes, it still matters.

A large number of AI chips can still help companies in many ways. With more chips, they can run more experiments as they in other words detect new ways to manufacture AI, more chips can still give companies a technical and competitive advantage.

Experts stated that “logic” would also require more chips to operate the new breed of AI model. They require more computing power when people and business use them.

Has the United States not limited the number of NVidia chips sold to China?

Yes. To maintain the US lead in the global AI race, the Biden administration implemented the rules that limit the number of powerful chips that could be sold to China and other rivals.

But the impressive performance of the Deepsek model raised questions about the unexpected results of the US government’s trade restrictions. Controls have forced researchers in China to be creative with a wide range of devices that are freely available on the Internet.

Some experts continue to argue in favor of American trade restrictions, saying that they were recently placed and that they will have more impact on China’s abilities to create AI as AI.

Does Deepsek’s technology means that China is now ahead of the United States in AI?

No. The world has not yet seen the O3 model of OpenaiI, and its performance on standard benchmark tests was more impressive than anything in the market. But experts are worried that China is jumping forward on the Open-SOS AI system.

What is really open-source AI?

Like many other companies, Deepsek has “open sour” its latest AI system, meaning that it has shared the underlying computer code with other businesses and researchers. This allows others to manufacture and distribute their own products using similar technologies.

This is part of the reason that other people in Deepsek and China are capable of creating competitive AI systems so quickly and cheaply.

In the AI ​​world, Open Source first collects steam in 2023, when Meta independently shared a system called Lama to the AI ​​system. At the time, many people admitted that the open-source ecosystem only flourishes when companies such as the meta-western firms continue to open their technologies with huge data centers filled with special chips.

But Deepsek and others have shown that this ecosystem can thrive in ways that are beyond American technical giants.

Why is it important?

Many experts have argued that large American companies should not open their technologies as their use can cause disintegration or cause other serious damage. Some US lawmakers have detected the possibility of preventing or throtting practice.

But other experts have argued that China will achieve a significant lead if regulators prevent the progress of open-source technology in the United States. If the best open-source techniques come from China, these experts argue, American researchers and companies will manufacture their systems over those technologies.

In the long run, it can put China to the center of AI Research and Development, which can further intensify its effort to build a wide range of AI technologies, including autonomous weapons and other military systems.