Baidu’s ERNIE 4.0 turns one person into ‘an AI marketing team’

The generative AI race is not letting up. This week, Chinese search and web portal giant Baidu announced the release of ERNIE 4.0, its new large language foundation model, and previewed a whole host of new software applications built atop it, including one — Qingduo, a creative platform that aims to rival Canva and Adobe Creative Cloud, both of which have recently also added new AI features.

Taking the stage at the company’s annual Baidu World 2023 conference in Beijing, Baidu CEO Robin Li said Qingduo allows “one person to become an AI marketing team.”

Li said ERNIE 4.0 and the user-facing text chatbot, ERNIE Bot — similar to how rival OpenAI’s ChatGPT is the chatbot built atop the company’s GPT foundation model — was built to perform four core capabilities: understanding, generating, reasoning, and memorizing.

In live demos, Li showcased ERNIE Bot’s prowess in understanding complicated and jumbled human requests and responding to them.

Beyond that, it generated a range of content, such as text, images, and videos, in just a few minutes. It even solved complex geometry problems and adapted to new information in real-time while writing a story. “These four core capabilities form the foundation of AI-native applications,” Li said.

Return to the stage

The announcement of ERNIE 4.0 was the centerpiece of Baidu World 2023, the search giant’s first in-person gathering following after a four-year hiatus and online-only format due to the COVID-19 pandemic.

In press materials provided to VentureBeat and other journalists, Dr. Haifeng Wang, Baidu’s Chief Technology Officer, reported that ERNIE 4.0 had improved its overall performance by nearly 30% since its beta testing in September. The new model is available in an invitation-only beta at the moment.

Infusing AI across the product suite

In a page from Microsoft’s playbook, Baidu further revealed how ERNIE and other AI technology was coming to all its apps.

Baidu Search provides more in-depth information and offers an interactive chat interface for complicated queries.

Like Microsoft Bing Chat and Google Bard, it will “aggregate and summarize information from diverse web sources and present a consolidated answer.”

However, Baidu Search also seeks to go further, displaying “text, image and dynamic graph, ensuring answers that are both vivid and concise.”

Baidu GBI, the fruit of a recent healthcare data acquisition, is aimed at business intelligence, and can rapidly generate custom business analytics. Baidu’s press materials describe GBI as “an all-knowing, fast-reacting assistant in the business world,” that provides fast summaries of information, and “reduces the heavy lifting of having to go through multiple spreadsheets to retrieve and analyze data.”

Infoflow, a workplace app, uses AI to complete office tasks such as scheduling meetings to booking travel with awareness of people’s calendars and availability. It also summarizes meetings and chat conversations with key points.

Baidu Wenku, previously a straightforward document-sharing platform, can now be used to conduct research and provide summaries, as well as generate content including documents and slideshows.

Baidu Maps has been upgraded with an “AI guide will take the initiative to suggest next step of action to users during their trip.”

And not to be left out, Baidu Drive, the company’s cloud file storage solution, is getting its own GenAI assistant, YunYiduo, which Baidu claims is “World’s first cloud drive intelligent assistant.” It sounds impressive, allowing users to interact with text or voice to pull up specific files or even information within them, summarizing multiple files or segments of a video, for example.