Chinese Authorities Call For Internet Companies To Add Bias To AI Algorithms — In Order To 'Promote Mainstream Values'
from the Xi-Jinping's-little-red-subroutine dept
Techdirt has been tracking the worsening online surveillance and censorship situation in China for many years now. The latest move concerns the currently hot area of artificial intelligence (AI). It’s a sector that the Chinese government understands better than most Western governments, and which it has made one of its technology priorities. The authorities in China know that AI in the form of algorithms is increasingly deployed to optimize and customize Web sites. They have realized that this fact gives them an important new lever for controlling the online world. As South China Morning Post reports, the Cyberspace Administration of China has released its draft regulations on “managing the cyberspace ecosystem”, which include the following:
The regulations state that information providers on all manner of platforms — from news and social media sites, to gaming and e-commerce — should strengthen the management of recommendation lists, trending topics, “hot search” lists and push notifications.
“Online information providers that use algorithms to push customised information [to users] should build recommendation systems that promote mainstream values, and establish mechanisms for manual intervention and override,” it said.
“Mainstream values” include resources that promote Xi Jinping’s writings; party policies and socialist core values; information that displays China’s economic and social development; and anything else which helps promote Chinese culture and stability. By contrast, “harmful information” is stuff that is “sexually suggestive, promotes extravagant lifestyles, flaunts wealth or hypes celebrity gossip and scandals.”
As is increasingly the case, China is in the vanguard of digital culture here. The rest of the world is beginning to wake up to the serious threat of bias as AI-powered algorithms are deployed more widely. China has moved beyond that stage and is now actively weaponizing bias to push a government agenda. This is a useful warning to those who see algorithmic decision-making as the solution to hard problems.
For example, it is clear that the only way that the EU Copyright Directive‘s upload filters can be implemented is through automated filters using AI. As China’s latest move makes clear, once those filters are in place on major Internet sites in the EU, it would be easy for governments to require that the software should be tweaked to introduce a little bias — to protect the children, or society, or whatever. Those who are horrified by what the Chinese authorities are proposing would do well to start arguing for safeguards to stop the same path being taken outside that country.