The latest iteration of its generative artificial intelligence (AI) model, Ernie 4.0, was introduced on Tuesday by Chinese technology giant Baidu (9888. HK), which said it was on par with the GPT-4 model developed by ChatGPT creator OpenAI.
At a gathering in Beijing, CEO Robin Li unveiled Ernie 4.0. He highlighted the model’s memory skills and demonstrated them by creating a real-life martial arts book. He also demonstrated Ernie 4.0, producing films and banners for advertisements.
Analysts didn’t seem impressed. According to Lu Yanxia, an analyst at market research firm IDC, the debut of Ernie 4.0 was not particularly noteworthy compared to the previous version.
In early trade, Baidu’s Hong Kong shares declined 1.32%, lagging a 0.7% increase in the larger Hang Seng Index (.HIS).
“We should see significant improvements once Ernie 4.0 is used hands-on, but concrete upgrades aren’t immediately clear,” Lu stated.
Lu listed Baidu’s integration of generative AI throughout all of its products, including Baidu Drive and Baidu Maps, as additional significant developments from the event.
I showed how, in contrast to the past, users of Baidu Map can now access features by using natural language queries powered by Ernie rather than having to sift through many alternatives.
Amid global enthusiasm for the technology triggered by the release of ChatGPT last year, Baidu, the owner of the largest internet search engine in China, is at the forefront of AI models in China.
Investors were upset only to see pre-recorded demos of the company’s ErnieBot chatbot, which it released in March.
In August, Baidu was one of many companies that received regulatory authorization to make AI products available to the general market. During the ceremony, Baidu’s chief technology officer Wang Haifeng announced that Ernie has accumulated 45 million users since it was made available to the general public.
According to statistics from brokerage CLSA, China already possesses at least 130 large language models (LLMs), accounting for 40% of the world total and trailing only the United States 50%.
A blocklist of sources that cannot be used to train AI models was one of the suggested security restrictions for companies providing services based on the technology that Beijing released last week.
