Baidu, the leading search engine provider in China, announced that its latest version of the Ernie AI model, Ernie 3.5, has surpassed the widely popular OpenAI chatbot, ChatGPT, in multiple key metrics. Baidu made this statement on Tuesday, highlighting Ernie 3.5's superior comprehensive ability scores compared to ChatGPT and its outperformance in several Chinese capabilities compared to GPT-4.
To substantiate their claims, Baidu referred to a test conducted by the state newspaper China Science Daily, which utilised datasets including AGIEval and C-Eval. These benchmarks are commonly used to evaluate the performance of AI models. OpenAI has not yet responded to Baidu's announcement.
The introduction of ChatGPT sparked a global AI buzz that has now reached China, leading to a flurry of domestic companies announcing their own competitive products. Baidu was the first major Chinese tech company to unveil an AI product to rival ChatGPT, introducing its language AI Ernie Bot in March. Built on Baidu's previous Ernie 3.0 AI model, Ernie Bot has been undergoing invite-only tests for the past three months.
Watch BT Visual Story How to spot AI-generated images from real ones!
Following Baidu's lead, other prominent Chinese tech firms, including Alibaba Group and Tencent Holdings, have also revealed their respective AI models.
Baidu claims that its new Ernie model boasts improved training and inference efficiency, positioning it for faster and more cost-effective iterations in the future. Additionally, Baidu states that the new model will support external "plugins," which are additional applications that enhance the AI's capabilities for specific scenarios. These plugins enable functionalities like summarising lengthy text and generating more accurate answers.
Notably, ChatGPT had already introduced plugin support in March, further driving the competition between Baidu and OpenAI in the AI landscape.
Also Read
'Buying Netflix at $4 billion would've been better instead of...': Former Yahoo CEO Marissa Mayer
ChatGPT beats top investment funds in stock-picking experiment