The well-known chatbot, ChatGPT has been outperformed by Baidu’s most recent version of the Ernie AI model, Ernie 3.5, based on a number of important measures, said the company. This announcement was released by Baidu on Tuesday, noting the fact that Ernie 3.5 outperformed GPT-4 in several Chinese skills and had better overall ability scores than ChatGPT.
Baidu used a test by the official newspaper China Science Daily, which used datasets like AGIEval and C-Eval, to support its assertions. These benchmarks are frequently used to assess how well AI models perform. OpenAI has not reacted to Baidu’s statement yet.
A global AI boom that was ignited by the release of ChatGPT has now spread to China, prompting a rush of domestic companies to announce their own competitive technologies. With the release of its language AI Ernie Bot in March, Baidu became the first significant Chinese tech company to introduce an AI product to compete with ChatGPT.
Ernie 3.5 has been conducting invitation-only tests for the past three months. It is based on Baidu’s prior Ernie 3.0 AI model. Other well-known Chinese IT companies, such as Alibaba Group and Tencent Holdings, have followed Baidu’s lead and introduced their own AI models.
With better training and inference efficiency, according to Baidu, the new Ernie model is positioned for future iterations that will be quicker and more affordable. Furthermore, according to Baidu, the new model would accept external “plugins,” or extra apps, which improve the AI’s skills for particular scenarios. These plugins make it possible to perform functions like summing long texts and producing more precise results.
Notably, ChatGPT had already added support for plugins in March, intensifying the rivalry between Baidu and OpenAI in the AI industry.