China’s AI Firms Are Copying US Innovations, Still Several Years Behind

It’s no secret that China idea of technological advancement often boils down to copying whatever the US has come up with. When it comes to artificial intelligence that’s still what is happening. A few weeks ago Wired published an article about a Chinese AI startup that it said was winning the Open AI race:

Meta shook up the race to build more powerful artificial intelligence last July by releasing Llama 2, an AI model similar to the one behind ChatGPT, for anyone to download and use. In November, a little-known startup from Beijing, 01.AI, released its own open source model that outperforms Llama 2 and scores near the top of many leaderboards used to compare the power of AI models.

Sounds impressive but much later in the story we get this.

Not long after the Chinese model was released, some developers noticed that 01.AI’s code had previously included mentions of Meta’s model that were later removed. Richard Lin, 01.AI’s head of open source, later said that the company would revert the changes, and the company has credited Llama 2 for part of the architecture for Yi-34B. Like all leading language models, 01.AI’s is based on the “transformer” architecture first developed by Google researchers in 2017, and the Chinese company derived that component from Llama 2…

The connection with Meta’s Llama 2 is an example of how despite Lee’s confidence in China’s AI expertise it is currently following America’s lead in generative AI. Jeffrey Ding, an assistant professor at George Washington University who studies China’s AI scene, says that although Chinese researchers have released dozens of large language models, the industry as a whole still lags behind the US.

So the Chinese have made some new versions of open source tech from the US but they are still several years behind the leaders. Today the NY Times published an article focused on the irony of this situation.

The Chinese firm, 01.AI, was only eight months old but had deep-pocketed backers and a $1 billion valuation and was founded by a well-known investor and technologist, Kai-Fu Lee. In interviews, Mr. Lee presented his A.I. system as an alternative to options like Meta’s generative A.I. model, called LLaMA.

There was just one twist: Some of the technology in 01.AI’s system came from LLaMA. Mr. Lee’s start-up then built on Meta’s technology, training its system with new data to make it more powerful.

The situation is emblematic of a reality that many in China openly admit. Even as the country races to build generative A.I., Chinese companies are relying almost entirely on underlying systems from the United States.

Not every Chinese AI startup is copying US tech but the story quotes an investment analyst who says the ones Chinese companies have built from scratch “aren’t very good.” China is good at copying but not as good at innovating. And that’s not because Chinese companies aren’t capable or Chinese engineers aren’t smart. Part of the problem, as always, is the Chinese Communist Party.

Chinese companies with the resources to build a generative A.I. model faced a dilemma. If they created a chatbot that said the wrong thing, its makers would pay the price. And no one could be sure what might tumble out of a chatbot’s digital mouth.

“It’s just not possible to get rid of all the problematic ways these systems can express themselves,” said Andrew Ng, who teaches computer science at Stanford and was a former executive at Baidu, the Chinese search giant.

Of course we’ve got our own problems here with AI systems that are tuned to only provide appropriately woke responses. But imagine the risks a Chinese developer would be taking in releasing something like this. What happens if users ask it for information on communism and some of it isn’t flattering? What if they ask for information on Tiananmen Square or other sore spots in Chinese history? Hong Kong? Taiwain? The number of topics which could potentially get the creators into serious trouble in China seems almost limitless and might change from day to day. And so, when Chinese company Baidu rolled out their chatbot it was later revealed the entire presentation had been pre-recorded. There’s just too much at stake to risk winging it and that pressure is holding these companies back and allowing US companies to take the lead.

There are some good comments on this story:

We need to stay ahead of China on the AI front for several reasons. First, they are a totalitarian and surveillance society and will definitely use this technology against us and against their own citizens. Second there are multiple military applications that are very important. Third it is their general MO to steal and copy anything that isn’t nailed down, so we need to protect and safeguard developments that we and our companies have made.

Another one:

AI that has value depends on accurate data – lots of it. But the CCP operates like Orwell’s Ministry of Truth – facts are changed to fit CCP doctrine, inconvenient truths are scrubbed. How can AI be developed using data managed by the MoT? Or will China’s AI also depend on western data? But then the AI product would also be… inconvenient, and censored.

Finally, several people raised the issue of Chinese students who come to the US for an education in tech fields and then return home to launch careers in China based on their US education. One commenter from Boston suggested there was cheating going on.

I am studying at a US university in my final year. It is quite well-known that students from China form large groups on messaging apps with the sole purpose of exchanging answers for homework assignments. Such coordination or collusion (whatever one chooses to term it) undoubtedly helps to boost their collective grades. I feel sorry for other students who take pride in personal initiative and academic honesty, who only have one brain to rely on.

Another curious thing I’ve noticed is that many Chinese students in Economics (my major) tend to target the same few elective courses (of the many available). I did not see any Chinese students, for instance, in the course “Economic Development for Latin America.” Another elective course, “Monetary and Banking Institutions,” was known to have a steep grading curve and quite verbose content. It had zero Chinese students too. Ideally, they will target the same courses offered by the same professor. Many professors only tweak their assignments slightly from semester to semester. Some don’t tweak them at all, even recycling the same questions for midterms and finals. But before some students even enter a course, they have all the notes as well as assignment solutions (courtesy of their peers who brainstormed together) and even copies of exam papers (if their peers were allowed to take them home after submitting the answer sheet).

Is this actually happening? I wish the deans of some of our top universities were doing something to find out.