Baidu, a Chinese search engine company, asserts that its newly upgraded app, Ernie, surpasses the Microsoft-supported OpenAI chatbot in various aspects.
In a formal statement released on Tuesday, Baidu claims that the latest iteration of its Ernie AI model, Ernie 3.5, has excelled in terms of comprehensive ability scores compared to ChatGPT. Additionally, Ernie 3.5 has demonstrated superior performance in numerous Chinese-specific capabilities when pitted against GPT-4, as per the Chinese company.
According to CNBC, one test examined in the study relied on standard admission and qualification exams, similar to those necessary for college acceptance or legal certification. Ernie 3.5 outperformed OpenAI’s ChatGPT and GPT 4 in the Chinese language, but fell behind in the English language. Notably, this particular evaluation was conducted by Microsoft.
Moving on, another evaluation aimed at Chinese language proficiency encompassed a vast array of topics, comprising over 13,000 multiple-choice questions across more than 50 subjects. Ernie 3.5 demonstrated superiority over ChatGPT and GPT 4 in this assessment.
Conversely, a third evaluation, encompassing diverse subjects like science and humanities, was developed collaboratively by a consortium of esteemed U.S. universities. Ernie 3.5 lagged behind both ChatGPT and GPT 4 in this particular test.
Baidu’s recent announcement coincides with the global enthusiasm for AI sparked by ChatGPT, a phenomenon that has also reached China, subsequently leading numerous domestic companies to unveil competing products.
The company made waves in the Chinese tech industry by introducing its own AI product, Ernie Bot, to directly compete with ChatGPT. Ernie Bot, an advanced language AI, is based on Baidu’s earlier Ernie 3.0 AI model and has undergone exclusive testing for the past three months.
Following Baidu’s lead, other major Chinese tech giants such as Alibaba Group and Tencent Holdings have also unveiled their own AI models.
The Chinese search engine giant claims that its new model excels in training and inference efficiency, positioning it favorably for quicker and more cost-effective advancements in the future. Moreover, it has been emphasized that the new model is equipped to support external “plugins,” enhancing its capabilities even further.