Stay up to this point with free updates
Simply register for Chinese economy and finance myFT Digest – delivered straight to your inbox.
The two most vital artificial intelligence ecosystems on this planet are the US and China. Geopolitical tensions make it tempting to view the 2 as opposites. China's AI is described as lagging behind, yet leading, the US. In fact, the 2 countries are pursuing different strategies.
Admittedly, that is partly resulting from necessity. Since US startup OpenAI launched the chatbot ChatGPT in November 2022, there was a feverish global race to develop large language models that power generative AI. China is hampered on this race by two aspects: lack of access to advanced US chips and censorship of data.
Without advanced chips, China's computing power is restricted. Companies like OpenAI are expanding the dimensions of their models. GPT-4, OpenAI's latest flagship model, is alleged to be trained with greater than a trillion parameters. Chinese firms must as a substitute concentrate on efficiency. For example, Hangzhou-based DeepSeek released DeepSeek-V2 this yr, an open-source model for big languages that uses 21 billion energetic parameters.
Chinese firms are also hampered by the info they will access. The introduction of China's Personal Data Protection Law in 2021 introduced strict data protection standards much like Europe's General Data Protection Regulation. In recent years, China's regulators have repeatedly stated that personal and company data should be protected.
Such restrictions mean that Chinese firms have more incentive to develop AI services based on smaller, large-scale language models reasonably than training them across the web. While these lack the raw power of their larger rivals, they’re cheaper to develop and run.
China's technology ecosystem is predicated on such pragmatism. Research that can not be translated right into a product within the foreseeable future, resembling artificial intelligence, has difficulty obtaining funding.
This signifies that while the US ecosystem is ahead in breakthrough innovations, China excels in execution: finding the appropriate product-market solution, scaling, and offering highly reasonably priced applications.
The World Intellectual Property Organization (WIP) recently published a Report on the patent landscape on generative AI, which showed that China was accountable for greater than 38,000 patent applications between 2014 and 2023, in comparison with 6,276 applications within the U.S. in the course of the same period. Tencent, Ping An Insurance, Baidu and the Chinese Academy of Sciences led the pack. Of course, patents don’t equal breakthroughs. Alphabet trails Tencent by way of AI patent titles, but Alphabet's Google DeepMind has given the world AlphaFold, which could revolutionize AI drug discovery.
But the big variety of Chinese patents may lead to more products. One area of Chinese AI applications to observe is electrical vehicles. It took China 26 years to supply its first 10 million electric vehicles and just 17 months to supply the following 10 million. About half of the cars sold in China this yr are expected to be smart cars with tablet bikes. Baidu, which just launched 500 self-driving taxis in Wuhan, is running the world's largest autonomous vehicle experiment. The firms may never produce superhuman AI, but “ok” AI that interacts with humans through its cars will likely be popular in China prior to elsewhere.
China's technology sector is commonly portrayed as a monolithic behemoth, buoyed by state-driven ambitions. In reality, regulation and geopolitical constraints mean it’s ahead in some areas, but struggling to catch up in others. AI is the world's most disruptive technology, and each China and the US can play their part. After all, the largest challenges facing our species – from fighting climate change to curing cancer – are usually not Chinese or American in nature, but human in nature.
