LogInYou can't find Tiananmen Square or Chinese politicians here.
A new AI-powered text-to-image model ERNIE-ViLG has been released by Baidu to compete with DALL-E, Mudjourney, and Stable Diffusion. What makes it stand out is its ability to understand prompts in Chinese, as well as generate anime art and capture Chinese culture better than other tools.
However, its users found out that ERNIE-ViLG blocks certain words connected to politics, like Tiananmen Square, names of well-known political leaders, things like "revolution" and "climb walls", and the name of Baidu's CEO.
According to MIT Technology Review, words like "democracy" and "government" are allowed while prompts that combine them with other words, like "democracy Middle East" or "British government," are blocked. "Tiananmen Square in Beijing also can’t be found in ERNIE-ViLG, likely because of its association with the Tiananmen Massacre, references to which are heavily censored in China."
It seems like censorship is getting to AI art as well. Moderation in general is not something new for such tools, for example, DALL-E doesn't allow using public figures in prompts. However, some believe there is a difference between blocking sensitive and political information.
"It’s different when an AI judges what it cannot generate based on commonly agreed-upon moral standards and when a government, as a third party, comes in and says you can’t do this because it harms the country or the national government," said political cartoonist Badiucao.
Despite the censorship, ERNIE-ViLG is an important player in the AI business, and we're looking forward to seeing what it can offer.
Try it yourself here and don't forget to join our Reddit page and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.
Preview image by ひびきの三丁目
Keep reading
You may find these articles interesting