How many points can AI get in the college entrance examination? Take a look at the results of ChatGPT 4 and Wenxin Yiyan

Original source: Hardcore Kanban

Image credit: Generated by Unbounded AI‌

Remember the hot search that was blown up by AI college entrance examination essays not long ago?

Some people think that the writing level of AI has surpassed most people, and some people say that AI can only get 0 points in the test

The results of the college entrance examination began to be released, and we also found the scores of AI...

The two AI contestants this time are ChatGPT-4 from Microsoft and **Wen Xin Yi Yan ** from Baidu.

After they answered the composition questions of the National Paper A, three front-line teachers from different regions simulated the marking of the college entrance examination and scored them.

From the perspective of answering speed, Wenxin Yiyan is even better, wrote 1103 words in 29 seconds; while Chat GPT-4 only wrote 846 words in 1 minute and 42 seconds.

From the point of view of writing, Wen Xinyiyan's composition quotes a lot of famous quotes, gives many examples, and gives three methods of "making time work for me";

Chat GPT-4 first affirmed the convenience of life brought by technology, then talked about how technology makes people slaves, and finally proposed to use technology selectively and arrange time rationally.

Which do you think will score higher?

**The full score is 60 points, and the average score given by the teachers to Chat GPT is 36 points. **

| It can be said that it is an unsatisfactory article. This article too highlights a flaw in the thinking pattern;

| Although relevant, the entire article lacks an effective and credible argument;

| Using too much invalid space to expand the phenomenon of the material itself, instead of creating. Most of them are correct nonsense, and there are too few that are really constructive, operational, and able to hit the pain points.

Look at Wen Xinyiyan’s article again, with an average score of 42. Here’s what the marking teachers said:

| It is the one with the most literary talents and the most detailed arguments, but we must know that it is not particularly good to quote too much;

| Although a lot of these quotations, verses, and many examples are used in it, many of them are examples that many candidates like to use, such as I am repairing cultural relics in the Forbidden City;

| The example is very good, but it does not clearly explain the relationship between people and time;

| It is obvious that I want to write where I want to write, the logic is not strong, and the score will not be high, because his structure is too old, and the whole article is basically driven by evidence rather than logic.

According to the marking standards of the college entrance examination, an excellent college entrance examination composition (category one essay) generally scores 50 points or above.

Although the three teachers from different regions may have overall high or low scores, the overall average result still shows:

**The two AI candidates with fast writing speed can only be regarded as medium level. **

Completing the article according to the algorithm will inherit many of the long-standing shortcomings in our previous college entrance examination composition. Many articles that seem to be good to everyone have gradually no longer meet our current needs for the college entrance examination.

Taking this opportunity, we also quietly conducted a "Turing Test".

In addition to the two AI compositions, a composition written by a real person was also handed over to the marking teachers to see if the teachers could tell the difference.

The opening argument of the real person's composition is "If you blindly rely on technology and become a slave to time, then the development of technology will be meaningless";

On the whole, it says that you can't indulge in technology and enjoy the benefits, and you can't blindly resist the trend of technological development. You must use the "moderate" thinking to use technology and learn to control yourself.

The teachers scored an average of 43 points without knowing that it was a real person's composition, which narrowly beat AI, for the following reasons:

• is the only one I can read, but the problem with this article is that it has misplaced the focus. His entire understanding of the topic itself is a bit biased, so it is actually a subtitle with a partial topic. son of a bitch.
• A gentleman is good at deceiving things, just saying that people should be good at learning and learning with the help of foreign things;
• The structure of the whole article, in the process of raising and analyzing questions, he paid more attention to the relationship between people and technology, and weakened the time element;
• Real actionable arguments are too late to come up with.

In the link of "Distinguish real people's composition", the obvious English-Chinese translation traces in the Chat GPT-4 composition exposed its true identity, and it was first excluded by the teachers.

Wen Xinyiyan's works are more confusing, and the three teachers have all wondered whether this work is from a real person. The reason is that they have also encountered many students who want to get high marks by citing classics and piling up rhetoric in teaching.

After this battle, everyone must have noticed that it is difficult for AI to write closely to the meaning of the topic; humans with stronger divergent thinking will inevitably fall into the trap of "thinking too much" when thinking deeply. **Current AI is still just a icing on the cake for human thinking; **As an important part of talent selection, the college entrance examination is also constantly evolving. Only then is it even better. **