China’s DeepSeek R1 Safe AI model built against sensitive political topics; cost of training its model is…

China’s Deepkseek R1 Safe Model

Deepseek, the largest language of China, is now suggesting the Chinese government’s policies. His new upgrade is fully compatible with the central government. According to Chinese organizers, Reuters reported that Deepseek’s new variant R-1-safe was developed to prevent discussion on politically sensitive or controversial issues.

Reuters said Deepseek variant was developed together as a safety -oriented version of the artificial intelligence model by Chinese technology giant Huawei. Huawei said that it was “about 100% successful” in the prevention of politically sensitive issues. The new model is designed because Chinese regulators looking for artificial intelligence models of the country are designed because they are made before the public announcement according to the “socialist values ​​of China and follow the strict controls about speech.

WeChat said in a statement, Huawei, Deepseek’s open -source model R1 to educate the large language model of 1,000 Ascend AI chip, he said. For this reason, the safe version was created not only by Deepseek, but also by Huawei, who works with researchers from the University of Zhejiang.

Companies designed the model to equip them with strict guards without affecting their performance. Huawei said that with this design, the new model and speed of the new model are attenuation at only percentage. However, it became strongly resistant to “toxic and harmful speech, political sensitive content and illegal activities”.

How expensive was the training of Deepseek R1?

After the launch, Reuters reported that the Deepseek R1 surprised its competitors with a cost -effective system and open source nature, as it offers unlimited free use to large users. The US spent 294,000 USD to train the R1, which is much lower than its competitors.