AI is betting on college entrance examination essay questions: Science, metaphysics or hype?

For college entrance examination students, the happiest thing is that they will meet all the questions they encounter, and they have already memorized the answers to the questions they don’t know... However, can such a good thing really happen?

Just yesterday, a Weibo blogger retweeted a video posted a week ago, saying, "I used AI to predict the topics of the 2022 college entrance examination essays, and I won the right bet on 7 college entrance examination papers from all over the country." Among them, there are 4 sets of essay questions."

AI predicts college entrance examination essay questions. Is it science or metaphysics?

Sina Technology got in touch with the above-mentioned blogger Fat Hu. As a senior algorithm expert in AI systems, he said that the crawler produced by the team captured the comments published by official media such as People's Daily and People's Daily Online within a year, and established An AI system predicted the seven most likely proposition directions.

However, some netizens questioned that the direction of this kind of AI prediction is macroscopic. "Reliable Chinese teachers will draw directions, and they are all macroscopic. In this way, everyone can be called the king of prediction."

How does AI predict essay questions?

If you want to understand the technical principles of AI predicting college entrance examination essay questions, you must first understand the nature of the college entrance examination essay questions.

In the video announcing the AI ​​prediction results of the 2022 college entrance examination essay questions released on June 1, the blogger introduced the technical principles of this AI system-the college entrance examination essay seems to be an exam, but in fact it is Guided by values, outlook on life and awareness of family and country, this rule determines the score of a college entrance examination essay.

Therefore, "The essence of the college entrance examination is an ideological and political task, which reflects consciousness." And this is the fundamental principle of predicting the college entrance examination composition.

In essence, the idea of ​​​​thesis proposition is the same as that of the Spring Festival Gala sketch review, and there is essentially no difference. The proposition team, who is well versed in this, will also screen out a group of highly important and narrow topics from international events, national policies, meeting documents, guidance and official reports in the past year, and finally select the previous ones. College entrance examination essay questions that have never appeared in the college entrance examination and are most suitable for this year's tasks.

AI is an important tool that is best at mining patterns from large-scale data. The more data it acquires and the more reasonable the distribution, the more accurately it can predict the final results.

Based on the above thinking, Panghu and his team specially wrote a set of crawlers. This set of crawlers captured articles, journals, newspapers and important documents of high-level meetings published by the People's Daily, People's Daily Online, CCTV News and other official media within a year.

Then through the pre-training model and topic modeling algorithm in natural language processing, we established an AI system that can understand Chinese news and articles, and let the AI ​​system learn the college entrance examination essays of the past few years. Question ideas, and finally predicted the seven most likely question directions in the 2022 college entrance examination.

In the prediction announcement released on June 1, the seven major proposition directions announced by Fat Tiger and his team are, "difficulties and hope, hard work and self-improvement, exploration and dreams, innovation and development, struggle and value, fairness and moral education, beauty and environment." Behind the core words of each proposition, there are also predictive models and "model representations" that can be used for expansion and extension.

Take the prediction direction of "Difficulties and Hopes" as an example. In the prediction results, the AI ​​system mainly extracts the predicted topics and core words based on the model representation. In the process of students understanding the proposition, the predicted topics can As the main direction of understanding the writing content of the composition, the introduction provided by the model representation can be embedded into the article as a case reference.

The core words "difficulty and hope" at the top can be used as the main proposition reference direction of the composition title.

How accurate is the prediction?

The result of the AI ​​system's bet has come out. So, is the predicted result accurate?

According to Sina Technology Statistics, 7 sets of the 2022 National College Entrance Examination Paper A, National Paper B, National New College Entrance Examination Paper II, National New College Entrance Examination Paper I, Beijing Paper, Tianjin Paper and Zhejiang Paper were officially announced on June 7 The examination paper found that among the seven sets of college entrance examination essay questions, they were indeed consistent to a certain extent with the propositions previously announced by the Fat Tiger team.

For example, in Volume II of the 2022 National College Entrance Examination, the composition is required to be based on the theme of "Choice·Creation·Future", and also mentions "scientists, scientific research, Beidou navigation system" and " New methods, new thinking, new knowledge” and other words.

In the proposition direction given by the AI ​​system with the theme of "exploration and dreaming", scientific terms such as Beidou and chips have become representations of the prompt model, and the theme direction is also close to the proposition of the college entrance examination essay.

In addition to the New College Entrance Examination Paper I and the National Paper A, the proposition directions predicted by AI also hit relatively close subject directions. But can the highly general proposition directions derived by this AI system from a statistical perspective really help candidates prepare for the exam? At present, it seems that the outside world has mixed evaluations of this result.

Some netizens exclaimed: "Awesome! Can you predict the CET-4 and CET-6 exams?"

Another group of netizens questioned, "A ridiculously funny AI fortune teller can do anything with a more reliable Chinese teacher."

As this pledge Fat Hu, the developer of the AI ​​system who got 4 essay questions right, also objectively pointed out, “This year, there were more propositions combined with hot topics in the college entrance examination essays in various places, which allowed the AI ​​system to bet on 4 propositions correctly. However, the questions in previous years are generally not straightforward. The questions are based on hot spots, but they are profiled and seen from the small. This time there are many hits and it is very special."

In the future, can AI guide candidates to replace teachers?

Although some focus on college entrance examination composition, AI prediction of college entrance examination topic selection has not yet reached the point of directly guiding students to prepare for the college entrance examination or replacing teachers.

Regarding the fact that the direction of propositions given by AI is relatively macro and not specific, Pang Hu also further pointed out, “The AI ​​system’s prediction of college entrance examination essays is more to tell the outside world that AI can already do many things, but in previous college entrance examinations Composition is more of a test of students’ knowledge reserve and ability to use language flexibly. The direction of AI prediction can only be used as an auxiliary reference.”

In addition to AI prediction of college entrance examination composition, as the threshold for using AI technology decreases With the continuous popularization of technology and technology, in the fields of medical care, agriculture, industry, and corporate commerce, it is becoming more and more widespread to promote business process optimization and business forecasting based on AI technology. Are these predictions and judgments based on AI science or metaphysics?

Pang Hu believes that “in fields such as data classification and prediction tasks, the prediction accuracy of some AIs has far exceeded that of ordinary people.” Intelligent voice technology and face recognition technology have been widely integrated into public life. Today, AI has long become a frequent visitor in our lives, and the prediction of results based on AI and big data statistics is no longer a metaphysics.

The above is the detailed content of AI is betting on college entrance examination essay questions: Science, metaphysics or hype?. For more information, please follow other related articles on the PHP Chinese website!