Chinese Language Understanding Evaluation (CLUE)

2022: Tencent and Alibaba AI systems understand Chinese speech better than people

On November 28, 2022, the results of testing language models of Tencent and Alibaba Group based on artificial intelligence (AI) were released. Smart algorithms are reportedly able to understand Chinese better than people do.

We are talking about assessing the capabilities of AI using the CLUE (Chinese Language Understanding Evaluation) method. Tests have shown that rival Hunyuan models developed by Tencent and AliceMind developed by Alibaba scored more points in a number of tasks for the first time than ordinary citizens. So, the Hunyuan system showed a result of 86.918 points. In second place with a slight lag is AliceMind - 86.685 points. By comparison, people get an average score of 86.678. In the fourth and fifth positions in the ranking are the AI systems of the smartphone manufacturer Oppo and the product delivery company Meituan.

Tencent and Alibaba AI systems understand Chinese speech

Experts say the findings suggest a qualitatively new level of development in AI-based language models. For example, earlier in 2022, the Chinese Internet giant Baidu announced that its bot Du Xiaoxiao wrote an article that scored more points than the work of most applicants who took all-Chinese state entrance exams to universities (gaokao).

On the other hand, the researchers emphasize that most AI models still have a long way to go before they can truly understand the intricacies of languages. In 2021, scientists from the Alabama Polytechnic Institute (USA) and the research division of the American company Adobe discovered that many AI systems that coped with some tasks to understand the language better than people could not give qualitative results when the words in the sentence were rearranged randomly.^[1]

Notes

Шаблон:Remarks