China Baidu claims its conversation robot is as efficient as

China: Baidu claims its conversation robot is as efficient as ChatGPT

Chinese Internet giant Baidu unveiled a new version of its conversation robot Ernie Bot in Beijing on Tuesday, whose capabilities are now “as good” as the American ChatGPT, its boss assured.

Ernie, whose primary language is Mandarin Chinese, was launched in August and marked a major step in China’s desire to be a global leader in artificial intelligence by 2030.

The new version of the robot is currently only available as a beta version for developers by invitation.

However, drastic content controls in China prevent answers to questions on topics deemed sensitive by those in power, such as politics or the events at Tiananmen Square.

In the style of the famous Apple conferences, Baidu founder Robin Li took the stage on Tuesday to put his artificial intelligence robot through a puzzle session to solve.

In an auditorium built on the site of a former steel mill at one of the 2022 Winter Olympics venues, Robin Li then asks Ernie to write the outline of a martial arts novel.

The robot carries out the action and gradually creates new characters, in front of a visibly convinced audience that follows the writing in real time on a huge screen.

Ernie’s skills in “comprehension, creation, logic and memory (…) are as good as those of GPT-4”, the language model used to develop ChatGPT, Robin Li has assured.

The AFP could not independently verify these claims.

Baidu was the first company in its country to announce work on a local equivalent of ChatGPT, capable of formulating detailed answers on a variety of topics or writing essays in just a few seconds.

The power of ChatGPT is passionately pursued in China, where the American interface is still blocked without bypass software such as VPN and foreign phone number.

Amid this euphoria for these new tools, China announced in April that it would impose a “security review” on artificial intelligence tools before approving their commercialization.

Content generated by artificial intelligence must “reflect fundamental socialist values ​​and not contain (elements related to) the subversion of state power,” these regulations say.