SCALE AI 28 -year -old billionaire CEO issued a warning to this new Chinese new company

Alexandr Wang, a 28 -year -old founder and CEO of $ 13 billion in a $ 13 billion new company, said that about ten years, the United States has maintained its dominant position in China. However, Wang said in an interview with the CNBC on January 23 that as the Chinese startup Deepseek released a “shocking” artificial intelligence model during Christmas, the gap between the two countries was dangerous.

Earlier this week, DeepSeek released the second artificial intelligence model, which can be comparable to the models of top American companies such as OpenAI. In addition to its performance that shocked researchers, the rapid progress of this Chinese startup company has also caused people to question the effectiveness of the export control of artificial intelligence chips. Treatment unit (GPU).

Wang said: “The artificial intelligence competition and the artificial intelligence war between China and the United States are one of the most important issues today.” He recently published the entire version of the advertisement on the Washington Post, urging the Sichuan -Pu government to protect the United States in artificial intelligence Leading in the field. The advertisement recommended a letter from Mr. Wang to the reader. The letter claimed that China’s artificial intelligence model was quickly catching up, and called on the federal government to invest more funds in calculations and data, and launch a energy plan to support artificial intelligence. Prosperity, and other suggestions.

King is one of the youngest billionaires in the world. He was born in Los Alamos, New Mexico, and his father was a weapon physicist working in the famous Los Alamos National Laboratory. After a short period of time for Silicon Valley to work for Financial Technology’s ADDEPAR and Q & A website Quora, Wang started to learn machine learning at MIT at the Massachusetts Institute of Technology, and then dropped out of school and established SCALE AI. Data hard work.

SCALE AI has a valuation of $ 13.8 billion last year, and its customers include the US Department of Defense and Openai. Wang is an old friend of OPENAI CEO Sam Altman. He lived in the same room with him during the COVID-19 major popularity. In 2016, he launched Scale AI in Y Combinator. Y Combinator was a new accelerator and was previously at the helm by Altman.

What is the new version of DeepSeek compared with the cutting -edge model of the United States?

SCALE AI has recently released the “human last test” with the Artificial Intelligence Center. They claim that this is the most stringent benchmark test for the artificial intelligence system so far. Wang said that although there is no model so far, it can reach more than 10% in the test, but DeepSeek’s new reasoning model DeepSeek-R1 has ranked first in ranking. “Their models are actually the best performance, or they are roughly equivalent to the best models in the United States,” he told CNBC.

Due to China’s strict GPU export control, Deepseek claims that it has successfully created a powerful model, but compared with the American system, its computing power is much lower. According to Deepseek researchers, the Deepseek-V3 model released in December last year was trained on about 2,000 AI chips in NVIDIA (NVDA). The Meta (META) LLAMA 3.1 model was trained on 16,000 GPUs.

The king is not convinced. “There are more H100s in the Chinese laboratory than people think,” he said, referring to a NVIDIA GPU that cannot be used legally in China. “My understanding is that Deepseek has about 50,000 H100, apparently they can’t talk about these, because it violates the export control of the United States.”

Wang said that in order to continue to promote its cutting -edge models, the United States must release more computing capabilities and infrastructure. He believes that once artificial intelligence reaches the benchmark of General Artificial Intelligence (AGI), the value of the artificial intelligence market is expected to reach $ 1 trillion. Although artificial intelligence leaders provide various timetables and meanings for GM artificial intelligence, the definition of Wang refers to the system that “basically can be a long -distance worker in the most capable way”. He believes that this achievement will be in It will be realized in the next two to four years.