On Monday, Deepseek captured the start of the Chinese artificial intelligence appleApp Store, get rid of Chatgpt for Deepseek assistant in artificial intelligence. Global technology shares were sold and were at a pace to survey billions of dollars in the market.
Later on Monday, Dibsic said it will temporarily limit user records “due to the widespread malicious attacks” on its services, although current users will be able to log in as usual.
Technology leaders, analysts, investors and developers say the noise-the fear followed by the backward in the constantly changing self-noise cycle-may be justified. Especially in the era of the AI Arms Truc race, where technology and startups are racing to ensure that they are not delayed in the market It is expected that the highest $ 1 trillion In revenues during a decade.
What is a dipcique?
Deepseek was founded in 2023 by Liang Wenfeng, co-founder of the High-Flyer, a quantitative hedge box focusing on artificial intelligence. According to the AI Startup, the AI Research Unit of the hedge fund in April 2023 to focus on large language models and access to artificial general intelligence, or AGI – a branch of artificial intelligence and its competitors say they are following quickly. Deepseek is still fully owned and funded by high models, according to analysts at Jefferies.
Deepseek began to capture Steam earlier this month, when Startup R1 released the thinking model that competes with Openai’s O1. It is open source, which means that any developer of Amnesty International can use it, and it has been spent to the highest application stores and Industry leadersWith the users to be praised by its performance and thinking capabilities.
Like other Chinese Chatbots, it has its own restrictions when asked about certain topics: When asked about some of the Chinese leader Xi Jinping’s policies, for example, it is said in Dibsic The user directs away From similar lines of interrogation.
Another major part of the discussion: Deepseek’s R1 has been built despite the United States Check the chips exports To China three times in three years. Estimates are different from the costs of Deepsek R1, or the number of graphics processing units in which they entered. Jefferies estimated that the modern version was “the cost of training worth only $ 5.6 million Dead‘s llama. “But regardless of the specified numbers, reports agree that the model has been developed on a small part of the cost of competing models by Openai, anthropology, Google And others.
As a result, the artificial intelligence sector is steeped in questions, including whether the number of increasing astronomical financing rounds in the industry is necessary-and if the bubble is about to explode.
shares Nafidia Nearly 17 % fell on Monday due to the closure of the market, with the chips maker Asml Nearly 6 % decreased. Nasdak decreased more than 3 %. Four technical giants – Deadand Microsoftand apple ASML is all set to report this week.
Analysts in Raymond James detailed some of the questions that the artificial intelligence industry afflicted this month, “What are the effects of investment? What do you say about open models for ownership? Work?
Bernstein’s analysts wrote in a note on Monday that “according to many (sometimes hysteries), we saw (during the weekend) the effects of anywhere” this is interesting “as we know it.”
How American companies respond
Some American technology executives climb to respond before clients turned into a potential cheaper offers from Deepseek, where Meta started four of Debsik.War roomsWithin the artificial intelligence department.
Microsoft CEO Satia Nadella books On X that the Deepseek phenomenon was just an example of Jevons’s paradox, and writing, “Since artificial intelligence has become more efficient and accessible, we will see its Skyrocto use, and convert it into a commodity we cannot get enough.” Sam Altman, CEO of Openai, tweeted a quote his lineage to Napoleon, wrote, “The revolution cannot be made or stopped. The only thing that can be done is for one of her children to give it a direction through victories.”
Yan Li be, the chief artificial intelligence scientist, President of Amnesty International, books On LinkedIn that the success of Deepseek indicates a tidal change in the artificial intelligence sector for open source technology.
Lecun has written that Deepseek has benefited from some of the Meta technologies, that is, its Llama models, and that the startup “has reached new ideas and built it at the head of others. Since their work is published and an open source of profit from it.
Alexander Wang, CEO of Scale Ai, told CNBC last week that the last Debceic model is “destroying the land” and that the R1 version is more powerful.
“What we found is that Dibsic … is the best performance, or almost equally with the best American models,” said Wang, adding that the artificial intelligence race between the United States and China is the “Amnesty International War.” Wang provides training data to the main AI operators including Openai, Google and Meta.
Earlier this week, President Donald Trump announced a joint project with Openai, Oracle and Softbank to invest billions of dollars in the USA’s International Infrastructure. The project, Stargate, at the White House by Trump, CEO of Softbank Masayoshi Son, Oracle Larry Ellison founder and CEO of Openai Sam Altman. The main primary technology partners will include Microsoft, NVIDIA and Oracle, as well as ARM for semiconductor. They said they would invest $ 100 billion to start and up to $ 500 billion over the next four years.
Amnesty International is developing
Deepseek’s ingenuity news also comes amid the growing noise about artificial intelligence factors – models that go beyond chat keys to complete the user’s multi -stellar tasks – which are chased by technology giants and startups alike. Meta, Google, Amazon, Microsoft, Openai and Anthropic have expressed their goal of building AICEC AI.
“He can” use computers in the same way we do. “He said he could do tasks with” dozens or even hundreds of steps. “
Openai released a similar tool last week, as it provided an feature called the operator that will automate tasks such as vacation planning, fill models, make restaurant reservations and grocery arrangement.
the MicrosoftIt is described by the start of the novice operation as “an agent who can go to the web to perform tasks for you,” and added that he is a trainer to interact with “buttons, menus and text fields that people use daily” on the web. He can also ask follow -up questions to increase the allocation of the tasks you complete, such as login information for other websites. Users can control the screen at any time.
– Michael Bloom of CNBC contributed to reporting.
