- Deepseek, a small Chinese emerging company of artificial intelligence, released a major model on January 20.
- The emerging company caught the competitors in the Silicon Valley and the world of artificial intelligence by surprise.
- Deepseek engineers said they are building lower NVIDIA chips and lower capital.
Deepseek No. 1 achieved the free app -App Store Top Days days after the start of the Chinese start starting, R1.
On January 20, Dibsic, led by Chinese Hedge Fund Director Liang Winfing, issued R1. It caught on the silicon valley and the world of artificial intelligence because LLM can match some of the capacity of the competing models designed by Openai, Google and Meta.
These American companies have taken billions of dollars in obtaining very advanced chips and data to build models that can solve complex problems. However, Deepseek appears to build models that they can perform in similar criteria – in a small part of the cost.
In a paper released in late December, Deepseek researchers estimated that they had built and trained their V3 model for less than $ 6 million using about 2000 Nvidia H800 chips.
American companies, such as Meta, are racing to secure billions of dollars from NVIDIA H100 chips to build their chat stations. H100 chips are the leading graphics processing units in NVIDIA. Because of the US sanctions, China cannot import H100s, and instead import H800s, which has decrease in data transfer, Reuters said.
Deepseek, Meta and Openai spokesperson did not respond to the comment.
An unknown Chinese starter closes the gap with some of the world’s largest technology companies with much lower resources that can undermine American efforts to build a AI trench against global competitors.
After the inauguration of President Donald Trump, Openai announced a joint project with the federal government to spend $ 500 billion on Amnesty International’s infrastructure over the next four years.
Openai’s Chatgpt is currently ranked second on the same Apple chart.
