I guess it had to happen. China has begun cracking down on start-ups offering users running virtual private networks access to OpenAI’s vastly popular ChatGPT generative AI service.
ChatGPT isn’t officially available in China, but local AI chatbot services, ChatGPTRobot and AIGC Chat Robot among them, have been drawing on the service via APIs to produce answers that are translated for local users. Many of those unofficial services have shut down in the past couple of weeks due to “violation of relevant laws and regulations” in an entirely predictable move by Beijing to knock a US-developed service on the head before it catches on in China.
The Great Firewall already censors search results and prevents Western search engines like Google from being accessed within China’s internet realm. Generative AI creates a big headache for Beijing officials tasked with maintaining the country’s censorship machine.
China’s generative AI challenge
To be effective at answering a wide range of questions, these tools need to draw on large language models that are trained using vast amounts of data. But with China excluding some of the biggest internet-based training sources for fear of serving up responses that don’t align with the Communist Party philosophy and view of history, it will be hard for a Chinese equivalent of ChatGPT to match the American version when it comes to quality and relevance. It doesn’t help that many of the open-source tools that have aided the advancement of generative AI are based on the English language, rather than Chinese characters.
On the flip side, drawing on approved information sources, Chinese conversational AI engines will make it easier to entrench the prevailing worldview of the CCP, allowing answers to philosophical questions posed by users to be infused with the moral, political and economic ideology of the party.
ChatGPT is not without censorship – OpenAI is quick to point out that it has put guardrails in place so that the chatbot won’t take a political stance, joke about sensitive subjects or produce a hate-speech-filled tirade.
ERNIE set to debut
Now China’s tech giants are in a race to produce their own generative AI chatbots, with Baidu set to release to integrate its ERNIE Bot into search engine results, devices and digital assistants later this month. Alibaba is testing its own ChatGPT equivalent too.
It will be fascinating to see how generative AI is used in China when it is built into platforms like WeChat and used by hundreds of millions of people every day. But the move against ChatGPT pales in comparison to what the US House of Representatives Foreign Affairs Committee has just voted in favour of – giving President Biden the power to slap an outright ban on TikTok.
The Chinese-developed and owned social media platform has already been wiped from the smartphones of public servants in the US and the company claims to have spent US$1.5 billion on data security initiatives to convince the US Government it is not harvesting data on its 100 million US users on behalf of the Chinese Communist Party.
The Republican hawks on the Foreign Affairs Committee carried the vote, but a ban would have to gain the approval of Democrats in the Senate. It would be unprecedented if a ban was enshrined in law, sending a signal to China and everyone else that the US is willing to intervene in the app economy to remove perceived or real threats to national security.
No TikTok ban here
With tensions between the US and China at an all-time high over China’s balloons being shot down and a leaked Department of Energy report blaming a laboratory leak for causing the Covid-19 pandemic, social media apps and AI are being drawn into the geopolitical tussle.
As for our own government, the GCSB said this week it couldn’t ban TikTok on government devices even if it wanted to – it doesn’t have the legal authority to do so, even if Parliamentary Services has recommended that members of parliament refrain from using the service.
But as we saw with the government’s move to exclude Chinese telecoms giant Huawei from building 5G networks here, we eventually fall into line with our Five Eyes allies on security issues. TikTok is as wildly popular here as it is in the US, which is why its days may well be numbered – in the West at least.