Baidu’s Hong Kong shares rebound as users test ChatGPT-like Ernie bot

By Eduardo Baptista and Jason Xue

BEIJING/SHANGHAI (Reuters) – Hong Kong-listed shares in China’s Baidu rebounded 15.7% on Friday as users told of their experiences with the Ernie bot, recouping losses from a day earlier prompted by the chatbot’s launch, which failed to impress.

On Thursday, viewers of the Chinese search engine giant’s presentation panned the company’s lack of a live demonstration and a public launch, comparing it unfavourably to the capabilities and free-to-use launch of U.S. research lab OpenAI’s ChatGPT in November.

But some of the few users who received invite codes to try Ernie later in the day began to post and livestream tests of the Chinese chatbot and do a side-by-side comparison with U.S. chatbots such as Microsoft’s Bing chatbot, which is powered by ChatGPT technology.

One user described on the Weibo social media platform how Ernie had answered a question about the status of Hong Kong philosopher Zhang Jinqing accurately.

“The response was O.K.”, the user said.

“There is a definite gap between Ernie bot and Bing, but it is not insanely big, in certain questions (Ernie) even performed better than Bing,” according to a Weibo post by tech blogger Chapingjun, who has more than 2.4 million followers.

Analysts also said the initial disappointment the market felt from the launch was tempered by the realisation that the Chinese search engine giant was still best placed to build China’s strongest rival to ChatGPT.

“The market’s demand for the industrial application of general large-scale models is rapidly stimulated, and (Baidu) is expected to rely on historical accumulation and first-mover advantages to quickly acquire users and data,” CITIC Securities said in a research note on Friday.

More than 75,000 corporate users have applied for a trial of an Ernie API developed by Baidu Cloud, the Chinese company said in a video published on its official WeChat account on Friday.

(Reporting by Eduardo Baptista and Jason Xue; Additional reporting by Brenda Goh; Editing by Muralikumar Anantharaman and Gerry Doyle)