The strength of the two major powers, China and the United States, in the field of artificial intelligence (AI) has always been a focus of attention for domestic and foreign media. The Financial Times published an article stating that if it were a competition, the race is not about who can "sprint to the finish line" in developing the most powerful AI large models, but rather about who can last longer in this marathon of innovation and application promotion.
From this, the newspaper believes that the Chinese will ultimately laugh last.
The Financial Times analysis states that in terms of the most concerned large models, although China faces a relative lack of chip computing power due to U.S. restrictions, this has instead prompted Chinese companies to make innovations in algorithms and training methods to compensate for the lack of computing power, allowing Chinese AI large models to achieve scores close to those of U.S. large models with top-tier computing power, even under relatively inferior computing conditions.
Moreover, the media also cited analyses from professional institutions indicating that models requiring less computing power can match or even surpass high-computing-power models if they are strengthened in data training.
This point has already been demonstrated in the competition of open-source models. The Financial Times pointed out that Chinese open-source models have already surpassed their U.S. counterparts. Free open-source models themselves can greatly assist in the promotion of Chinese large models in both domestic and international markets. Not to mention the fact that China and the United States currently have completely different trade policies: China is more open to the world and focuses on investing in emerging markets, while the U.S. is not only significantly retreating, but also harming its own allies.
The Financial Times said that China is fully developing its own chips. The government has formulated industrial policies and invested substantial resources in R&D, talent cultivation, and infrastructure construction. Although China's semiconductor industry still lags behind the U.S. in certain aspects, Chinese enterprises have found suitable application scenarios for domestic chips, using these chips for inference rather than for more computing-intensive model training. After all, the gap between Chinese domestic chips and U.S. chips in this area is not very significant.
The newspaper also emphasized the importance of this application scenario for domestic chips, pointing out that the development of AI ultimately needs to be linked to the real economy, to deploy models in the real economy, so as to achieve stable monetization and sustainable development. Therefore, China's chip development strategy and the application scenarios chosen by enterprises can give China long-term and significant advantages.
The Financial Times believes that China's continuous investment in talent cultivation, scientific research, and infrastructure will not only mean that China, which is already leading the U.S. in AI patents, will have more related patents, but also help companies bear a lot of capital pressure and investment risks associated with building data centers.
Furthermore, the Financial Times stated that China's power generation capacity is very strong and has already surpassed the U.S. This means that although China's data center construction started later than the U.S., China can quickly achieve large-scale deployment in this area.
The Financial Times also mentioned China's strong advantages in rare earth resources and processing, and thus pointed out that China now has a complete supply chain covering AI development — this is much more important than focusing solely on the score of a single large model.
Finally, the newspaper concluded that in the AI field, although the U.S. currently leads China in terms of chip advantages at the starting line, China has greater potential. Therefore, the Chinese will ultimately laugh last.
Source: Global Times
Original: toutiao.com/article/7597223126115697158/
Statement: This article represents the views of the author.