Revealed! AI is the "culprit" of Internet junk content production? | Digital Literacy

Review expert: Liu Xiangquan, Director of the Robotics Engineering Department of Beijing Information Science and Technology University

Let me ask you a question first: Have you ever used AI-assisted tools?

Image source | Jiemian News

In this "AI era", the threshold for writing a decent article is already very low. With the help of AI, we may not be able to match the writing style of real writers, but we only need a framework of an idea, and AI can expand it into a complete article.

Image source | Wenxinyiyan

I don’t know if you have noticed that many search results in search engines have been occupied by AI-generated content, and there is no shortage of “junk content” full of errors and omissions.

How can we stand out from the siege of these "junk content" generated by AI?

Why AI started to produce "junk content"

What is "spam"?

Imagine a scenario: when your homework or paper is facing a deadline, and your work progress is really "terrible", so you have to let AI generate a paragraph of text based on your description to "fill in the gaps". This is a kind of "junk content".

If you have used AI to generate content for you, you will find that only by constantly polishing your description of the work content can AI generate a satisfactory manuscript.

A study published this month counted the frequency of words used in peer reviews by scientists researching AI at several prominent scientific conferences.

The word “meticulous” was more than 34 times more frequent than the year before, “praiseworthy” about 10 times more frequent, and “complex” 11 times more frequent. These phrases are favorites of modern large language models such as ChatGPT.

Image source | Zhihu

In other words, researchers at a large number of AI conferences are handing over their peer review work to AI, or at least writing it with considerable help from AI, and this phenomenon becomes more pronounced the closer the deadline gets.

We used to say that deadline is the primary productive force, but now it is probably the “primary productive force” of AI.

In addition to those forced to use AI due to deadlines, some self-media practitioners who love "traffic is king" are using AI to edit low-cost and high-efficiency fake content. These contents have the same characteristics:

1. The title is amazing, attracts traffic, and has high clicks and comments;

2. Content formatting: each article is different, but very similar;

3. Guide comments to vent emotions, but ignore the search for objective facts.

With the "help" of AI, meaningless and false content is flooding the Internet. The time cost of finding the information you really need from these "junk content" is constantly increasing.

Is all this “junk content” AI’s fault?

AI can become more and more intelligent. On the one hand , it relies on artificial intelligence trainers to help AI become more "smart" through language model training. On the other hand , it also relies on AI to obtain information from the Internet through large model data for "self-study".

Trainers help AI continuously learn and improve its content generation capabilities by inputting large amounts of text, pictures and other content. This has also caused spam makers who rely on producing "junk content" to gain traffic and make profits to also focus on using AI to help them cultivate "content farms."

Content farms usually refer to those that use software to create a large amount of junk content to get user clicks, thereby improving their rankings in search engines and becoming easier to search. The core of content farms is to "publish various writing topics" and use a set of algorithms invented by themselves to evaluate two things: first, what are the current hot topics on the Internet; second, what keywords are more likely to bring high advertising revenue.

Image source | Tencent Cloud Developer Community

In fact, "junk content" has been polluting the Internet for a long time, but now AI helps these people produce this content more easily. All you need to do is come up with a sensational headline that attracts traffic, and then input it into the AI ​​tool, and use the "hallucination" effect of the large language model (that is, it can generate content that is actually untrue around the information entered by the user), "junk content" can be produced quickly, and AI has become an "accomplice" in producing "junk content".

How to reasonably use AI in the content production process?

What we should understand is that AI is not only capable of producing “junk content”; through proper learning and training, AI can even write poetry.

Liu Cixin, a well-known science fiction writer, invented an "electronic poet" software in 1989. This software can produce modern new poems at a speed of 200 lines per second and produce poems like the following:

"The asteroids are called! / Around the solid, there are only gelatinous rice fields / No, I don't want to fly!! / I miss / / Trigonometry is watched! / Around Andromeda, there are only living giant rivers / No, I don't want to devour myself!! / I precipitate / / The dragonfly is pinched! / Around the Orient Express, there are only squeaking bows and arrows / No, I don't want to smoke!! / I talk / / The confinement room is warned! / Around the sword, there is only the squeaking time / No, I don't want to sleepwalk!! / I rot."

AI can help the evildoers produce "junk content", but it can also be our good partner in writing and learning.

1. If you want AI to improve, you have to improve yourself first

Sometimes, we always feel that AI is not as smart as it is advertised, which may be due to problems with the prompt words.

The “3W1H” model is a good way to structure prompts, which means “who, what, why, how”: define the role and goal you want AI to play, clarify the requirements and details, explain the background and purpose of the requirements, and give the desired results and red lines.

Before using AI, train yourself to give prompt words. Perhaps cooperation with AI can achieve twice the result with half the effort.

2. Let it become your "inspiration partner"

A journey of a thousand miles begins with a single step. The hardest step in writing a good article is often the first step: finding a good topic.

If you feel that you are lacking in inspiration and cannot find the right person to discuss with you to broaden your horizons, then you might as well talk to AI. Through "100 Questions about the Industry", hot spot analysis, topic extension, etc., let it help you capture the moment of inspiration. Then you will have an "inspiration partner" online at any time.

3. Don’t know how to come up with a title? Ask the AI

How much can a good title add to your article? You may never know until you come across a good title.

If you are bad at coming up with titles, you might as well let AI help you and set several different title style requirements for it. Choose one that you like best and that fits the topic of the article. Then the situation where "writing an article for an hour, half of which is spent on coming up with a title" will no longer happen.

AI has produced a lot of "junk content", but the source is the abuse of AI technology by humans. The correct use of AI can greatly help produce higher-quality content. By the way, the title of this article was generated by AI. What do you think of AI's performance? Can you make AI generate a better title? Tell us in the backstage private message.