If there is a technology that America needs to bring the “fascinating new era of national success” promised by the president of the USA Donald Trump in his speech during swearing, this is the genetics artificial intelligence (Generic AI). As The Economist notes in its analysis, Generic AI will help increase productivity of the next decade, fueling economic growth. In the best scenario, it will boost humanity through a transformation comparable to the industrial revolution.

The fact that after his swearing -in, Trump announced the “largest infrastructure project in history” shows that he understands the potential of the sector. But so does the rest of the world and most of all the China. While Trump was speaking, a Chinese company released the latest impressive large linguistic model (LLM). Suddenly, America’s lead over China in artificial intelligence seems smaller than ever since it became famous Chatgpt.

The fact that China covered the distance between the two countries is striking because it was far behind and while the US had set a target to slow it down. Joe Biden’s government feared that General AI could secure China’s military superiority. For this reason, the US has limited the exports to China of the best chips to educate artificial intelligence and cut off China’s access to many of the necessary tools for the construction of substitutes. Behind its protective wall, Silicon Valley was imposed on the field. Chinese researchers are intensively studying US research on artificial intelligence, something Americans rarely do respectively.

However, China’s most recent progress overturns the industry and puts US officials in a difficult position. The success of Chinese models, coupled with changes throughout the industry, could overturn the finances of artificial intelligence. The US must be prepared for a world in which Chinese artificial intelligence will be a powerful competitor.

China’s linguistic models are not the best, but they are much cheaper. QWQ, owned by Alibaba, an e -commerce giant, started operating in November and is less than three months behind America’s top models. OR Deepseekwhose creator has sprung up from an investment company, is ranked seventh based on a comparison measure. Obviously, he was trained using 2,000 second -class chips – against 16,000 first -class chips for the Meta model, which Deepseek wins in certain categories. The cost of training an LLM in the US is tens of million dollars and is constantly increasing. Deepseek owner claims to have spent less than $ 6 million.

US businesses can copy Deepseek techniques if they want, because its model is open source. But cheap education will change the industry as the design of the models is evolving. Deepseek’s “Reasoning” model is designed to compete with a state -of -the -art OpenAi proposal. These models speak to themselves before answering a question. This process of “thinking” produces a better answer, but also uses more electricity. As the quality of production increases, the cost increases.

If the fairly good models of artificial intelligence can be trained relatively cheap, then the models will multiply, especially as many countries desperately want to get their own. And a high cost per question can also encourage the creation of more models made for a specific purpose and give effective, specialized answers with fewer questions.

The other consequence of the Chinese achievement is that the US is facing asymmetrical competition. It is now clear that China will innovate to bypass obstacles such as the lack of the best chips, either by increasing efficiency or by compensating for the absence of high quality material with a higher amount. China’s domestic chips are getting better and better, including those designed by Huawei, a technology company that previously achieved the widespread adoption of its telecommunications equipment with a cheap approach.

If China remains close to American competition, it could be the first to make the jump to superintelligence. If that happens, it can gain more than just a simple military advantage. Even if industry remains on its current course, the widespread adoption of Chinese artificial intelligence around the world could give Beijing a huge political influence, at least as worrying as the propaganda threat of Tiktok, a Chinese property ownership of which in the US remains unclear.

What should Trump do? His announcement on infrastructure was a good start. The US must eliminate legal obstacles for the construction of data centers. It should also ensure that hiring foreign engineers is easy and reform defense supplies to encourage the rapid adoption of artificial intelligence.

Some analysts argue that Trump should also abolish the export prohibitions of the chip industry. The Biden government acknowledged that the ban has failed to limit the development of Chinese artificial intelligence. However, this does not mean that nothing has achieved. In the worst case, artificial intelligence could be just as deadly as nuclear weapons. America would never send nuclear weapons components to its opponents, even if they had other ways to obtain them. Chinese artificial intelligence would certainly be even stronger if it now gaining easy access to the best chips.

It is more important to tackle Biden’s plan for the “artificial intelligence rule”, which governs which countries will have access to American technology. This is designed to force other countries to join America’s artificial intelligence ecosystem, but the technological industry has argued that, by introducing bureaucracy, it will bring the opposite effect. With every Chinese progress in the field, this view is gaining more and more weight. If America assumes that its technology is the only choice for countries such as India or Indonesia, it is in danger of being exaggerated itself. Some technological brains promise that the next innovation will put America back far ahead. Perhaps. But it would be dangerous to take a given US lead in artificial intelligence.