From academic texts to love statements, through emails and fiction works. The Artificial Intelligence (AI) generative The world is revolutionizing And as we know him.
Right now, it is very likely that you cross yourself with texts generated by AI and it is very difficult to differentiate them from those created by the human being. But they exist Some tricks to discern them.
According to the writing expert Naomi Alderman, The key is to observe how the words fit, instead of what they say, the British newspaper collects in a report Daily Mail.
“The words that should work together do not even fight each other (fighting is interesting), but they get rid of your hands like a cheap toy,” Aderman explained. The expert identifies Seven signs:
Too generic content
Table of Contents
Alderman says that one of the most obvious signals of AI is that the content is extremely generic. The expert says that these texts are lacking the personality that a human would be able to create.
In addition, since IA are usually designed to be useful, they tend to produce A friendly and something professional language. Alderman says: “Even when they are asked to be daring, you still have the feeling, from what is written, that they are exaggerating.”
Lack of coherence
Since AI generates word sentences by word, they can often lack the sense of internal coherence. Alderman describes this as a feeling of “The more you look at it, the less it means.”
She calls it the “Word salad feeling” which can sometimes be obtained when reading content generated by AI. If the words are combined strangely or in a seemingly disorderly order, it can be a good sign that they were not written by a human.
Words chosen at random
Even if the words fit correctly, you are in themselves could be incoherent when they are produced by a chatbot. To explain this feeling, Alderman refers to an episode of Friends where Joey uses a synonyms dictionary to rewrite prayer “Put it somewhere where you see it every day.”
However, the resulting phrase, “keep where your eyes land daily,” does not transmit exactly the same meaning. The texts written by AI can have A similar feeling of incoherenceas if the words had been chosen at random.
Although technically the meaning is the same, The tone of words may not be quite correct. Alderman says: “I can feel how spaces open between words and how the meaning falls between them.”
Would anyone say it out loud?
If you are not completely sure if a text was really written by a human, a simple test is to ask him if he would say it out loud. “Try to read it aloud or, better yet, Try to imagine that you say it spontaneously to another person“Alderman says. If you don’t seem something that your partner would naturally say to another person, then there is the possibility that you have not really written it.
Do you have a point of view?
While you read the texts out loud, Alderman also suggests that you pay attention to “point of view” of prayers. Something written by a human will probably sound as if it came from a genuine human perspective, while the text generated by the lack of that sensation.
“When you say those words out loud, You feel as if, at best, you were reading a piece of bureaucracy written by a committee. There is an emotional feeling of … that there is no person at the other extreme, “explains the expert.
An impersonal or distant tone can often be the product of some large linguistic models, especially if so They have been scheduled to behave.
Is writing bad?
Although many people They turn to chatgpt To improve its writing, Alderman points out that the results are usually, simply, bad. This does not necessarily mean that the text would be criticized in a class of creative writing, but rather that lacks that instinctive human touch.
This could be especially clear if your partner has suddenly begun to produce love poetry. While humans have a great instinct for the rhythm when writing, IA still do not have that same understanding of the rhythm. So, although it could simply mean that your partner is a bad poet, a poorly written verse could also be an AI interference signal.
What do others say?
If you doubt a text, you can always ask another AI for him. You can simply copy and paste the message in chatgpt and ask: “Was this written by a LMM?”
The AI can then analyze whether the text has the characteristic distribution of words and score that they usually use Large linguistic models.
There are even programs Artificial Intelligence Verification Specialized Online that offer to scan text for use with artificial intelligence writing tools.
