govt.chinadaily.com.cn

News

Baidu launches latest large language model

Updated: Oct 18, 2023 By FAN FEIFEI China Daily Print
Share - WeChat
Visitors learn about the latest version of Baidu's large language model at Baidu World 2023 in Beijing on Tuesday. ZHU XINGXIN/CHINA DAILY

Chinese tech heavyweight Baidu Inc on Tuesday unveiled the latest version of its large language model, Ernie 4.0, saying its capabilities are on par with those of US company OpenAI's most advanced GPT-4 model, a move industry experts said will bolster large-scale commercial application of the fast-developing technology.

Ernie 4.0 has achieved a full upgrade with improved performance in understanding, generation, reasoning and memory compared with previous models, Robin Li, co-founder, chairman and CEO of Baidu, said at Baidu World 2023, the company's annual flagship technology conference, in Beijing.

Li showcased the four core artificial intelligence-powered capabilities of Ernie 4.0 at the launch event. He displayed the model's generative skills to produce advertising posters and videos, tested its reasoning abilities with math problems, displayed understanding skills by asking about housing provident fund loan policies and evaluated its memory skills by letting it write a martial arts novel in real time.

"These four core capabilities form the foundation of AI-native applications and have now unleashed unlimited opportunities for new innovations," Li said.

Baidu has also rolled out China's first generative business intelligence product — Baidu GBI — which simplifies data queries and analysis through natural language interactions.

The company has rebuilt many of its most popular applications, such as its search engine, maps and cloud drive services, by using the latest generative AI technologies.

Pan Helin, co-director of the Digital Economy and Financial Innovation Research Center at Zhejiang University's International Business School, said although LLMs have significant potential for applications in fields like culture, retail, healthcare and education, their real value comes from the consumer market, and the ultimate goal of LLMs in industrial applications is to benefit the general public.

Pan said the process of training large AI models necessitates higher requirements for computing capacity and algorithms, while Baidu has accumulated some advantages in comprehensive AI technological capabilities, such as algorithm models, innovation capability and industrial applications.

Chinese AI companies should pool more resources into improving computing power, algorithms and data quality, and step up investment in basic scientific research, to gain a competitive edge in the global AI chatbot race, he added.

Lu Yanxia, research director at market consultancy IDC China, said Chinese tech companies' continuous advancements in AI models will further promote the popularization of LLMs among the public, and bring about fresh business opportunities for domestic AI servers, cloud computing and chip companies.

Copyright©2024 China Daily. All rights reserved.

京ICP备13028878号-6

京公网安备 11010502032503号 京公网安备 11010502032503号