(CTN News) – During a hearing before themonth, ChatGPT Senator Mike Rounds pointed out that many prominent AI experts had just called for a six-month pause on giant AI experiments, in response to GPT-4 (the current basis for ChatGPT). Rounds’ conclusion was different.
He said pause is more risky than falling behind our competitors. AI will determine all future great power competition, and now is not the time to stop developing our AI capabilities.”
Apparently Rounds isn’t the only one. Since the meteoric rise of chatbots, this “AI race” frame has become more common. The United States’ top competitor in the “race” is China.
This narrative is wrong. This is wrong not just because China has a poor chance of leapfrogging the United States in the field of generative AI, but because China isn’t interested in leapfrogging the U.S. in the first place.
Let’s start with ChatGPT’s immediate response. Some Chinese companies deployed similar products – but their performance has been disappointing, and their use cases are limited.
However, the Chinese government quickly warned against excessive hype and initiated new regulations that make similar AI systems far more legally fraught.
Like this article? Subscribing is free. Monthly $5.
The Biden administration was restricting high-end computing hardware exports to China even before ChatGPT was announced. In part, this policy may have been motivated by the reliance on advanced computing hardware for cutting-edge AI methods, such as ChatGPT.
However, China hasn’t responded to these controls much, which belies its claim of leading the pack. After floating a big subsidy package in December, China backed off.
A few companies got additional subsidies in March, but the government didn’t pour more money into the sector.
China’s language modeling output has been half-hearted for a while. GPT-3, ChatGPT’s predecessor, sparked a worldwide flurry of language modeling activity, including in China, where new announcements were breathlessly covered by U.S. media.
There isn’t much validation for these 2021 models, and they’re probably overhyped.
But a new, exhaustive compilation of China’s published language models shows that Chinese activity largely died down in 2022, even as U.S. activity accelerated.
This evidence suggests China doesn’t view large language models as the technology of the century. The “AI race” frame overlooks three major reasons why Chinese leaders won’t be as concerned about advances in language modeling as Americans.
Despite China’s insistence that AI is a strategic technology, it has specialized in different fields than the US. China has focused much more on applications of AI and subfields like computer vision than U.S. researchers.
In February, Huawei’s CEO said the company would focus on industrial AI, not chatbots. US analysts can view breakthroughs in language modeling as inherently more significant because of the country’s relative advantage in natural language processing.
Secondly, language models make stuff up. It’s a kink in a new technology here. There are already regulations and arrests in China over the unpredictable and politically fraught comments language models might make.
As long as language models threaten social stability, Chinese leadership will keep them at arm’s length.
The US and China have had very different economic trajectories over the last half century. The percentage of U.S. GDP that comes from business and professional services has increased.
During that time, China’s manufacturing sector overtook the United States’, and it still makes up more than twice the United States’ share of GDP.
It could boost productivity in an economy reliant on professional services, where ChatGPT’s automation potential is highest. ChatGPT might not seem impressive to a country that’s focused on manufacturing.