In the ongoing quest for AI supremacy, China's top search engine provider, Baidu Inc., has claimed that its latest artificial intelligence (AI) model, Ernie 3.5, outpaces the globally recognized Microsoft-backed OpenAI chatbot, ChatGPT, on various crucial performance metrics.
On Tuesday, Baidu declared that the latest version of its AI model, Ernie 3.5, has outdone "ChatGPT in comprehensive ability scores" and bested "GPT-4 in several Chinese capabilities." This announcement was based on an evaluation using AGIEval and C-Eval, two widely recognized benchmarks for gauging the performance of AI models. The test was conducted by China Science Daily, a state-run newspaper.
The claim comes against a backdrop of a worldwide interest in AI sparked by ChatGPT, which has also taken root in China, resulting in a number of local corporations launching their competing products.
Baidu was the pioneer among large Chinese tech companies to launch an AI product to compete with ChatGPT, debuting its language AI, Ernie Bot, in March. Built on the foundation of the older Ernie 3.0 AI model, Ernie Bot has been in invitation-only testing for the past three months.
Several other Chinese tech giants, such as Alibaba Group and Tencent Holdings, have subsequently presented their own AI models.
Baidu also highlighted that its new model offers improved training and inference efficiency. This enhancement positions Ernie 3.5 for quicker and cost-effective future iterations. In addition, Baidu revealed that its new AI model would support external "plugins."
Plugins are ancillary applications that allow Baidu's AI to tackle more specific use-cases, including summarizing extensive text and generating more accurate responses. This move follows ChatGPT, which also launched plugin support in March.
OpenAI has yet to comment on Baidu's claim as of the time of the report.
