What is Deepsek, China’s AI Startup is sending shockwave through Global Tech? , technology
A small Chinese startup, Deepsek, has sent shockwaves through the global technical field with the release of an artificial intelligence (AI) model, with capabilities of which rival the works of Google and Openai.
The manufacturer of the Deepsek-R1 says its model was developed in the United States using less advanced, and less, computer chips than those employed by technical giants in the United States.
In a research paper released last week, the model’s development team said that they had spent less than $ 6m on computing power to train the model-Multivician-Dollars AI Budget enjoyed by American tech veterans like Oopenai, Alphabet and Meta. a fraction.
One of the most influential technical enterprise capitalists in Silicon Valley, Mark Andresen released the model as a “Sputnic moment of AI”.
A sudden emergence of a small Chinese startup capable of rivaling top players of Silicon Valley has challenged the perceptions about American dominance in AI and feared that Akash-high market evaluation of companies like Nvidia, alphabet and meta Can be separated from reality.
On Monday, NVIDIA, which is with one by one on the production of semi-circulars, has lost about $ 600BN in the market capitalization of power normal AI, after 17 percent of its shares.
US President Donald Trump, who had announced to launch a $ 500 billion AI initiative under the leadership of OpenIA, Texas-based Oracle and Japan’s SoftBank last week, stated that Deepsek was called “Wake-up call” on the need of the US industry Must work in “Laser-focused on competition to win”.
What is Deepsek?
The Deepsek, which is located in Hangzo, was founded by Liang Wenfeng in late 2023, also a serial entrepreneur, who also runs a hedge fund high-flier.
Although very rarely known outside China, Leiang has a comprehensive history of combination of burial technologies and investment.
According to Chinese Media Outlet Sina Finance, in 2013, he co-established Hangzo Jacobi Investment Management, which is an investment firm, who to implement trading strategies to AI, to implement a co-east student of Jaang University. Was employed for
Liang established two more firms in 2015 and 2016 respectively and two more firms concentrated on Computer-Directed Investment-Hannfang Technology Company and Ningbo Huannfang Quantitative Investment Management Partnership Partnership respectively.
In an interview with Chinese media outlet waves in 2023, Liang rejected the suggestion that it was too late to join AI for startups or should be considered prohibitedly expensive.
“Breeding alone is relatively cheap-based on public papers and open-source code, minimum time of training, or even fine tuning, is sufficient. However, research includes extensive use, comparison and demand for higher computational and talent, “Liang said, according to the translation of his comments published by Chintok Substack.
Liang said that his interest in AI was primarily inspired by “curiosity”.
“From a comprehensive point of view, we want to validate some hypotheses. For example, we hypothesize that the essence of human intelligence can be language, and human thoughts can essentially be a linguistic process, ”he said, according to the transcript.
“What do you think ‘thinking’ can actually be your brain weaving language. This shows that AGI like humans can potentially emerge from large language model,” he said, Artificial General Intelligence (AGI) Referring to, a type of AI that tries to mimic the cognitive abilities of the human mind.
Deepsek did not immediately respond to the remarks request.
On Monday, Gregory Zuckerman, a journalist from The Wall Street Journal, said that he had come to know that Liang, whom he had not heard earlier, wrote a preface for the Chinese version of a book, which he he he had written, which he he has hedge fund manager Jim Simmons Wrote about ,
“Simmons left a deep impact, apparently,” Zuckerman wrote in a column, stating how Liang praised his book that “many have opened up unresolved mysteries and we have experiences to learn experiences Brings the treasure “.
“Even my mother did not get much from the book,” Zakarman wrote.
Why has Deepsek taken the world of technology from the storm?
Simply put, the success of the company has raised questions about the AI’s approach by both Silicon Valley and the US government.
US tech firms have been widely considered a significant edge in AI, not at least due to their vast size, which attracts them from all over the world to attract top talents and create data centers and create large amounts of expensive high-high Allows to invest massively in buying in quantity. End chips.
The arrival of Deepsek at the scene has challenged the notion that AI takes billions of dollars to lead.
“Openai was established 10 years ago, has 4,500 employees, and has raised $ 6.6 billion in the capital. The Deepsek was founded less than 2 years ago, with 200 employees, and was developed for less than $ 10 million, “Edam Kobissi, the founder of the market analysis newsletter the Cobby, said on X on Monday. .
“How are these two companies now contestants?”
In his research paper, the engineers of the Deepsac stated that they used around 2,000 Nvidia H800 chips to train their models, which are less advanced than the most sophisticated chips.
The team said it uses several special models that are working together to enable slow chips to analyze slow chips more efficiently.
For the US government, the arrival of Deepsac at the scene has raised questions about its strategy to try to include China’s AI advances by restricting the export of high -end chips.
The research paper of Deepsek suggests that either the most advanced chips are not necessary to make high performing AI models or the Chinese firms can still source sufficient amounts of chips-or a combination of both.
California -based NVIDIA H800 chips, designed to follow American export controls, was independently exported to China by October 2023, when the then President Joe Biden’s administration of Biden to him in his banned goods list couple.
In his 2023 interview with waves, Lian said his company had stock 10,000 NVidia A100 GPU before being banned for exports. GPU, or graphics processing units, are electronic circuits that are used to speed up graphics and image processing on computing devices.
Tanishq Abraham, former research director at Stability AI, said he was not surprised by the level of China’s progress in AI, who looks at the rollouts of various models by Chinese firms such as Alibaba and Bachuan.
Abraham told Al Jazeera, “While China’s ability to achieve the GPU is banned, China has still managed to perform and squeeze out of what is still there.”
“I think it is a lesson for American companies that there are still a lot of performances that they can squeeze.”
Tara Javidi, co-director of Center for Machine Intelligence, Computing and Safety at California San Diego University, said Deepsek encouraged him about “rapid progress” around the world.
“My only hope is that the focus on this announcement will promote more intellectual interest in this subject, further talent will expand the pool, and, at least but at least, AI research in the US , “Javidi told,” Javidi told.
Meanwhile, the trust of investors in the US tech scene has hit – at least in short -term.
In addition to the dramatic slide of NVIDIA, Google Parents Alphabet and Microsoft on Monday saw a decline of 4.03 percent and 2.14 percent in their stock prices respectively, although Apple and Amazon over exhausted.
Professor Tim Miller specialized at AI at the University of Queensland said, “If the cost number of deepsek is real, now any larger organization in any company can build a lot more and host it.”
“So, in this sense, the game has completely changed because there is a new ‘rule’ that anyone can play.”
Does this mean that China is winning the AI race?
not necessarily.
While tech analysts broadly agree that Dipsek-R1 performs at the same level as a chat-or is even better for some tasks-the region is moving rapidly.
Openai CEO Sam Altman said earlier this month that the company would release its latest arguments AI model, O3 Mini, within weeks after considering the user’s response.
On Monday, Altman admitted that Deepsek-R1 was “impressive”, defending his company’s attention on more computing power.
“We will clearly distribute much better models and it is also valid to be a new competitor! We will draw some release, ”Altman said on X.
“But mostly we are excited to execute on our research roadmap and believe that more calculations are more important than ever to succeed in our mission.”
Rui Ma, the founder of Tech Buzz China, said that the Chinese tech sector was taken till Openi initially released the chat in 2022.
Ma told Al Jazira, “Most of the entrepreneurs had completely remembered the occasion that was represented by the generative AI, and felt very humble.”
“It is clear that they are difficult in work since then. I think this last weekend shows us how seriously they are self-sufficient and took a challenge to ‘catch up’ the silicon Valley. I think Washington should focus on promoting Silicon Valley instead of suppressing China, to maintain its lead to the US. ,
Abraham, former research director of Stability AI, said that perceptions can also be diagnosed with the fact that, unlike Deepsek, companies like OpenIII have not made their most advanced models available to the public independently.
“Deepsek provided its best model to use for free. On the other hand, Openai’s best model is not independent, ”he said.
“So most people who use chat for free are surprised by Deepsek and believe that there is a big jump in capabilities when OpenI has already paid a similar display model for a few months. This pay-walk of Frontier AI model does not actually reduce AI’s progress and abilities. ,
Professor Miller of the University of Queensland said that the advances and recent developments of Deepsak suggest that China is at least “up” with the US in AI.
“I made some prediction at the end of last year to some extent that the next scientific success in AI could come from a small player such as a personal university researcher does not have a lot of computing power – to compete them to compete. Will need to be smart, “he said.
“The clear progress of Deepsek is almost an example of this: due to not having enough computational power to create large models as chat, they had to be smart. Necessity is the mother of invention.”
