It has been claimed that someday, requests for human-written texts to be edited to seem like they were written by generative artificial intelligence (AI) will turn into the norm. The present issue is that generative AI texts may be deceived into considering they were written by humans, nevertheless it is explained that that is entirely possible if generative AI texts increase in number and turn into mainstream.
On the twenty seventh (local time), Forbes introduced an article by Dr. Lance Elliott titled, “Find out how to Write as if You are Attempting to Idiot Others into Pondering You are a Generative AI.”
Dr. Elliot is an AI consultant and startup founder who has a big subscriber base for his columns. In this text, he points out several phenomena related to the writing of ‘Ghostwriters’, i.e. generative AI.
First, he identified the attitudes of some teachers and tools that claim to have the opportunity to simply detect AI-written texts. The so-called ‘GPT detector’ is understood to not have a high detection rate, and even humans are barely capable of surpass it.
It is because students are usually not only asking easy generative AI to put in writing, but are also tricking teachers by asking them to “generate text that appears prefer it was written by a human.”
A lot of the writings written by generative AI are identified for his or her unique writing style and often used words and sentences. Nevertheless, the issue is solved when the user presents the writings she or he normally writes and asks, “Please edit the writings written by generative AI to my style,” he said.
At the moment, the hot button is to specify a particular persona and supply related materials or links, saying, “Write it as if OOO wrote it.” Most often, they reported that they didn’t realize that it was an AI-written article.
What’s more, the larger problem is that generative AI-written articles are generally rated higher than human-written articles.
Research results related to this have appeared several times up to now. A month ago, researchers on the University of Reading within the UK announced that 33 virtual students who used ChatGPT to create an internet psychology bachelor’s degree exam received 83% higher scores than real students.
Also, since most individuals think that they are usually not good at writing, they’ll receive help from ChatGPT and others, and even request that their very own writing be edited to make it seem like it was written by a generative AI.
Specifically, it’s predicted that this phenomenon will turn into more severe as AI-generated texts increase in number and turn into established as a type of ‘model answer’.
This problem has also been identified in a recent paper. Researchers on the University of Tübingen in Germany announced the outcomes of a study showing that non-English speaking researchers who’re weak in English receive a whole lot of help from ChatGPT and other tools when publishing papers, and this changes the general pattern of writing papers.
So while the demand to “edit my writing as if it were written by a generative AI” could seem absurd, it could turn into a significant trend within the not-so-distant future.
As well as, Dr. Elliott also introduced words and tones that often appear within the generated AI output text.
The five representative adjectives are commendable, progressive, meticulous, intricate, and notable, and the representative adverbs are meticulously, reportedly, lucidly, innovatively, and aptly.
He also advised that if you must seem like AI, you must avoid using curse words. As a rule, AI cannot use curse words due to guardrails.
Reporter Im Dae-jun ydj@aitimes.com
