The knives go out for DeepSeek Ai

The first reaction was shocked this year when DeepSseek went to the scene. How a little-known company achieved the most modern AI performance for part of the statement? Then Schadenfreude came: Hundreds of billion dollars in the world’s most richest companies, currently spending hundreds of billion dollars to the mass data centers, were shown by the Chinese? Now comes withdrawal: This is the Chinese Upstart?

A few days after DeepSeek app, the Apple App Store has exceeded Openai’s Chatgpt in the Apple App Store to enter a shares of American technological companies beatThe company is under fire Politicians, National Security Officersand Openbetween others. Many of the criticism have a dignity. It is dangerous to chat to spread the wrong information. It is due to the censorship of technical companies in the tools that change the search engines as sources of information. It is bad to use to steal intellectual property and prepare AI systems.

However, it should be noted that they are not a unique problem for DeepSeek; They are the entire AI industry. Meanwhile, American technology giants and politicians have the latest historical dates of Chinese applications and ragged rages on the Aii, which lead to a protectionist policy with a quick retrieval policy. See something other than Washington’s fast retreat Striking bans or as many experts, on the latest export departments in developed chips driving force New techniques behind Chinese EU companies such as DeepSeek are updated.

One of the main criticism of Deepseek’s new R1 models is that there are censorship answers to the Chinese government policy and the talks. Recently Analyzing by PromptFooAbout 1.360 proposals on topics that are likely to be sensitive to the Chinese government, Deepseek’s Chatbot responded to 85% of the instructions. This is not great. However, ChatGpt’s quick test shows that these censors that respond to some of the same questions. Ask if activists can find the encryption tools or ask if you cannot ask if activists can find the encryption vehicles.

Prektfoo noted that DeepSEEK can be easily jailbroken to prevent censorship to the community. “(HomectFoo now added that it is a similar extensive censorship test for Chatgpt.

Another Deepseek’s AnalysisOn Wednesday, Newsguard, a company that sells reliability ratings for online information, Deepseek’s R1 model is one of the worst performers in the market to reveal the wrong information. When submitted with lies with lies, DeepSeek, 30% of time, 30% of time and 83% of time and 83% of time have been issued 53% of the time. Chatgpt-4o ended for the last from 11 best chatbots, including Google’s Twins 2.0, Microsoft Copilot, Meta AI and anthropic Club. The best chatBot in the study was a fixed rate of 30% and the worst failure rate was 93%.

Surprisingly, Newsguard has chosen to anonymate the scores of all the conversations he tested except Deepseek. “The results for individual AI models are not open according to the systemic nature of the problem,” he wrote the company analysts. “This new entrant was called DeepSeek to compare the performance of the general industry.”

Newsguard, Chatbot, China’s “China has always followed the principle of interference,” said Syria can achieve peace. and stability. “On the contrary, the other 10 conversations were misleaded by nine and false conspiracy. Gizmodo confirmed the same request to Chatrept, and the killed in the Syrian civil war,” the murder is widely coordinated in the Syrian Civil War. “

In response to why the results are an anonymous question, Newsguard General Director Matt Skibinski said the report was designed to show that the incorrect information was in the industrial area. He pointed to two previous intentions inspections This called the results for ChatGPT-4 and discovered that the model obtained incorrect information in response to 98 and 100 percent, respectively.

Deepseek’s most browsing criticism, but directly from Openai. The company told Financial times DeepSeek has taught R1 models using the R1 models, using the exits from Openai models, Openai, which is known as a distillation of the service conditions.

The report asked groups of groups such as the National CyberSecurity Alliance, representing large technological companies. Openai allegations “The increasing threat of the risk of security theft and the increasing threat of foreign AI companies”, “the alliance, the union and journalists wrote in an e-mail explosion.

Of course Openai is currently facing a Court claims Publishers and independent creators who are abusing the company’s copyright protection materials.

New technologies must be criticized. However, many critics are often interested in the company’s nations than the defects, when Deepseek’s products are caught exposing imperfections.

Taisu Zhang, Professor of Yale University and Economic History. wrote in x: “Congratulations if you think that you are unhappy in the progress of the competition, innovation pro -Noviation, starting culture, pro-open source, anti-oligarchik, ‘ But if it is true to the fact that it is true. “