Recently, Baidu announced that the Ernie 3.5 AI model “beat” OpenAI’s ChatGPT and GPT 4 through a number of tests.
Recently, according to a report by the disinformation watchdog NewsGuard, ChatGPT gives more inaccurate information in Chinese than in English. According to TechCrunch, when users ask ChatGPT to reply in English, it mainly takes all available English data as well as with Chinese.
Earlier, in March, Baidu introduced Ernie Bot, a chatbot based on artificial intelligence model, Ernie of the Chinese search giant. It can be seen that, in the context of the fierce technology race between the US and China, Baidu’s investment and development of Ai is receiving more and more attention.
According to Reuters, Baidu cited test results conducted by China Science Daily using the AGIEval and C-Eval datasets – two standards used to evaluate the performance of AI models. Specifically, in the Chinese test with 13,000 multiple-choice questions with 50 different topics, Ernie 3.5 showed superiority over 2 competitors from OpenAI. However, when it came to taking the third test developed by a group of American universities with questions about the social sciences, Ernie 3.5 was completely left behind by ChatGPT and GPT 4.
In addition, in a June test conducted by Xinhua News Agency, Baidu’s Ernie Bot took the lead among the so-called “Chinese ChatGPT” apps. In it, Baidu has pushed to integrate Ernie Bot in other products such as cloud services or announced a venture capital fund worth $ 139 million to invest in Chinese artificial intelligence startups.