"The ChatGPT-style AI chatbot ERNIE Bot, or Wnxin Yiyan in Chinese, was completely developed by ourselves," said Wang Haifeng, chief technology officer of the Chinese online search giant Baidu, on a China Media Group's TV program on Saturday.

The company had released its 1.0 version in 2019. It has now been developed into a large language model that is characterized by industry-level knowledge enhancement, including natural language, visual and biological computing models, Wang said this a month after the bot was released.

There were a certain innovative technologies that have been proven in large-scale industrial applications, some of which have been patented or published on papers, or made available to the public, said Wang.

During the TV program, Wang showed the bot's abilities in literary creation, reading, drawing, business writing, math and understanding language.

"It has learnt trillions of words, so it can understand and generate text according to the needs of users," said Wang. He added that the chatbot can generate video, but for the time being, they haven't adopted the capability.

Underlying technologies of the chatbot

The key technologies used in the chatbot are similar to other large language models, Wang said, but there were three unique technologies used in this one, namely enhancement in knowledge, search and conversation.

These technologies would help increase the bot's capacity for strategic thinking, innovative exploration, searching with the country's largest search engine. They also help memorize and understand dialogue and then to respond, according to Wang.

Concerns over chatbots like ERNIE

When asked if AI chatbots could replace some jobs, Wang said the answer is yes, but it won't lead to a high unemployment.

"Chatbots can replace some of the jobs that need to be done by human beings, but at the same time it will create more jobs and give people more opportunities to do creative work," Wang added.

ChatGPT and similar AI have also sparked global concern on education. Wang said he believed education should keep pace with the times, and it should give full play to human creativity and inspire human inspiration while AI grows more and more powerful.

In security issues posed by the new technology, the CTO said any product that has a large customer will face the problem and all sectors of society are needed in solving these issues.

From a developer's perspective, Wang said they need to abide by relevant laws, regulations and ethics, and set up supervision and management mechanisms to solve safety issues from all aspects.

He added that the company's chatbot has built five levels of defense from data collection and processing, model training, to final application to prevent security risks. Moreover, the company has set up a data management committee since 2021 to improve relevant policies and rules.