Home >Technology peripherals >AI >Zhang Dongwei: Why is the college entrance examination composition level of AI large models not high?

Zhang Dongwei: Why is the college entrance examination composition level of AI large models not high?

WBOY
WBOYforward
2023-06-08 17:22:08878browse

The 2023 college entrance examination will be even more out of the ordinary, because this year's hot topic is not only the difficulty of the college entrance examination essay, but also because there are new "candidates", namely the large AI models of major manufacturers.

The most well-known large AI model in the world is ChatGPT in the United States. It uses an interactive dialogue interface and responds in a manner similar to human communication based on instructions input by the user. It has been astonishing the world since its inception. Currently, ChatGPT has become the fastest-growing application in history.

Through the AI ​​large model, authors can implement citation retrieval and format adjustment, document layout and literature summary, and quickly generate clear, grammatically correct and well-structured texts. In addition, authors can use these tools to question an argument, write code, improve sentences, think about article titles, and even edit and proofread papers.

What scares the education community is that ChatGPT has impacted the existing essay writing system. Schools in many countries have revealed that students’ essays written through ChatGPT have achieved extremely excellent results, and some have even ranked first. Therefore, ChatGPT has been banned from campus networks by several well-known universities such as the United States and the United Kingdom.

Zhang Dongwei: Why is the college entrance examination composition level of AI large models not high?

As we all know, China's cultural environment is very different from that of foreign countries. This is also a "localization problem" often raised by the Internet industry.

So, in the face of the 2023 college entrance examination essay, can the large AI model win?

"Blue Whale Finance" used ChatGPT, Wen Xinyiyan, and Tongyi Qianwen to participate in the college entrance examination essay competition for the 2023 college entrance examination essay on June 7. Judging from the results, the results can only be regarded as barely passing.

Not only college entrance examination compositions, but also compositions written in other scenarios, compositions written by AI large models will be quickly recognized.

This reason is exactly the same as the topic I talked about when I was drinking and communicating at Teacher Liu Ren’s house a few days ago.

Everyone in the industry knows that Mr. Liu Ren was once one of the most well-known journalists in the Chinese IT industry, and he was especially good at portraying character stories.

At that time, we talked about the fundamental difference between "professional authors" and "amateur authors".

Liu Ren said: Amateur writers can write well about things they have personally experienced, while professional writers can write well about things they have not personally experienced.

As for what is "good" and what is "bad", that is whether the article has true feelings injected into it.

Amateur authors can inject true feelings only through their own personal experiences.

Professional authors can also make their works full of realism and emotion by interviewing relevant people, researching materials, using cases, logical reasoning and scene reproduction.

Real feelings are capabilities that large AI models currently cannot achieve.

Therefore, in terms of scientific papers, AI large models have cited classics, strict logic, and accurate wording, which have surpassed most humans.

However, in terms of composition writing, AI large models cannot understand human emotions, let alone current social and political ideologies. It is not easy to achieve "passing".

What's more, in the current era of traffic-first, a good composition must not only be emotional, but also must invent one or two popular and memorable "golden sentences" to be considered outstanding.

So, in the Chinese world, there is still a long way to go before AI large models can replace human authors.

(Readers who want to use ChatGPT can send me a private message)

The above is the detailed content of Zhang Dongwei: Why is the college entrance examination composition level of AI large models not high?. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:sohu.com. If there is any infringement, please contact admin@php.cn delete