Baidu’s Ernie Bot Surpasses OpenAI’s ChatGPT in Performance
In an unexpected development in the world of artificial intelligence (AI), Baidu, China’s leading technology company, announced that the latest version of its AI model, Ernie Bot, has surpassed OpenAI’s ChatGPT in several key areas. This comes just three months after the bot’s initial reveal, which at the time, failed to impress the AI community, leading to a significant dip in Baidu’s share price.
More specifically, Ernie 3.5, the latest version of Baidu’s chatbot, has shown superior performance in multiple test sets, outpacing not only ChatGPT 3.5 in general abilities but also outperforming OpenAI’s more advanced GPT-4 large multimodal model in several Chinese-language capabilities.
The tests were conducted using the AGIEval human-centric benchmark, specifically designed to evaluate the general abilities of foundation models in tasks related to human cognition and problem-solving. In addition, the C-Eval multi-level multi-discipline Chinese evaluation suite for foundation models was also employed to gauge Ernie’s competencies.
This swift turnaround in performance can be attributed to Baidu’s improvements in the efficiency of Ernie Bot’s training and inference processes, leading to a faster rate of improvement for the AI model in the future.
The breakthrough is especially noteworthy considering the restrictions placed by the US government on China’s access to advanced semiconductor equipment and products, aimed at limiting the country’s progress in AI and other high-tech sectors. Despite these constraints, Chinese companies like Baidu have managed to keep up in the AI race. The landscape might change in the near future as the Dutch and Japanese governments have decided to join the US in restricting exports to China, adding to the challenges faced by Chinese tech firms.
