Chinese flock to try out Baidu’s ChatGPT rival Ernie Bot

Updated on: 01 September 2023

Justinas Vainilavičius
Senior Journalist

The heavily censored Ernie Bot responded to over 33 million queries – or 23,000 user questions per minute – on its first day of launch to the general public in China.

According to Baidu, a Beijing-based internet and artificial intelligence (AI) giant, over one million users have conversed with its chatbot in the first 24 hours since it was made publicly available.

The public launch yesterday (August 31st) saw Baidu’s Hong Kong-listed shares rise by 3%, which is in marked contrast to a 10% drop in March after the chatbot was released to a select number of beta users.

Reflecting further interest in the technology, Ernie Bot also jumped to the top of the Apple’s App Store in China on the day of its general rollout.

Foreign models like OpenAI’s ChatGPT or Google’s Bard are blocked in China. According to Baidu, users worldwide can access Ernie Bot via browser and the chatbot is free to use, but a Chinese number is needed to register.

Two other Chinese companies, Baichuan and Zhipu AI, also launched their chatbots on Thursday after getting approval from authorities.

This comes after China adopted a set of rules to govern AI services, which requires publicly available machine-generated content to adhere to “core socialist values.”

In practice, it means that content generated by chatbots like Ernie Bot is heavily censored and reflects the official Communist party line.

For example, it will refuse to answer – or end conversations altogether – if asked questions on topics such as Uighur concentration camps or Hong Kong protests.

Interestingly, users reported that it would not discuss issues that could portray Baidu in a negative light but would delve on criticism when it comes to competitors like Alibaba, a Chinese ecommerce company that has released its own AI products.

Despite Baidu recently claiming that Ernie Bot was better than ChatGPT in several key areas, some Chinese netizens were also dismayed by the frequent factual mistakes the chatbot makes.