When ChatGPT and Midjourney hit the world in 2022, it ceased to be the same. As of now, it is hard to say whether this change was for the better or not. Nevertheless, the world dictates its conditions to us, and we must adapt to them. Wherever we go, the pressing question of “Is it human or AI?” follows us. Nowadays, every text, picture, and even video is under suspicion. Did a real human create it? Is it a human generated text? Is it a real photo?
The idea of developing ChatGPT is commendable. It is designed to simplify text creation, reduce time spent on writing, and stop worrying about grammatical errors. However, given the fact that it and other AI text-generation tools are free and generally available, the question of detecting AI-generated text has arisen.
Tools for detecting artificially created text began to appear one by one. Their mission is clear, and so is the idea behind their invention. However, one of their most pressing problems is that they often give a false positive result. If before students had to worry about Turnitin checking their essays, now artificial intelligence detectors have caused them alarm. They ruthlessly deliver a verdict, whether it’s a human or AI, and their decision is not always correct. Let’s figure out what AI detectors look for in texts and what they rely on when analyzing the writing. But for that, let’s check out if you can spot AI content.
TEST. ARE THESE 5 SENTENCES ARE AI-WRITTEN?
“Did AI Make This?” Or, How You Can Tell the Difference
To provide you with a thorough answer to the question, “Are AI tools involved in generating the text under suspicion?” let’s dive deeper into how they work.
In simple terms, artificial intelligence tools are essentially tools that have a remarkable ability to learn and mimic people. Their work is based on complex, specifically trained algorithms designed to teach them to think and write like human beings. To achieve this, they train on a huge array of different data available on the internet. Engineers and technicians make sure that these instruments produce text as human-like as possible. That is, AI has knowledge of grammar and stylistics and tries to give the text the tone of voice that a person would want.
Of course, as previously mentioned, they seem to be a “win-win” option, and what else can you dream of? However, there are troubles in paradise. Despite the fact that AI is skillful in mimicking human intelligence artificially (we couldn’t resist this subtle joke), it still doesn’t quite succeed. At least for now, hello, “Terminator 2: Judgment Day.”
Those who are in the know declare that the main features of artificially generated text are:
- Textual inconsistencies: can you tell the difference? Experts consider inconsistencies in the text to be the first sign of a not a real human generated text. By this, they mean sudden shifts in writing style, tone, or vocabulary within a piece. Where humans stick to the same writing style and tone, artificial intelligence fails to do the job, they say. Inconsistencies within the text also include the wording. That is, if slang and significant informalities suddenly pop up in the middle of a pile of technical terminology and strict academic style, it is most likely that the paper has been generated by artificial intelligence. In this case, the question of “Is it human or AI?” will have an answer — most probably AI.
- Flawless grammar and spelling. We don’t really agree with this point. However, it’s our duty to keep you informed. Some researchers believe that a text with perfect grammar and no spelling mistakes may also raise suspicions that it has been generated by artificial intelligence. Beware of correct grammar and spelling, students! They are now causing concern.
- Unusual use of language: “Is it human or AI?” On the other hand, there are opinions that contradict the above. According to these beliefs, gross grammatical errors, such as incorrect use of tenses, in particular, are red flags for artificial intelligence detectors. Thus, reread your writing carefully so these tools don’t put on your paper the shameful label, “It’s an AI-generated text.” Also, if you have trouble finding synonyms and expressing your thoughts smoothly, we have bad news for you. According to prevailing beliefs, this is a sign of AI-developed text.
- Lack of context or coherence. This point of view seems to us to be quite justified. ChatGPT is trained on data that appeared on the internet until 2021. Therefore, it is natural that it and similar tools can’t generate solid, trustworthy information about events or phenomena that occurred later. Also, not a real human generated text often contains vague and unspecific wording. Such texts don’t brim with facts, as a rule.
Human vs AI: The Battle Is in Full Swing
Artificial intelligence should be treated with respect. This is what numerous books and movies teach us. Our generation is lucky (or not so lucky) to be in the middle of the AI vs human battle. Only time tells whether we will win this one. For now, the experts and general population can only speculate.
However, it is already clear that the development of artificial intelligence is moving by leaps and bounds. It is true that there are different opinions on how to determine whether a text is human generated or written by specifically trained algorithms. However, even now, it is very difficult to say for sure because AI is constantly learning and improving. Detectors do not always give accurate results and can be disputed. Give people the freedom to make mistakes, and, at the same time, don’t judge them for correct grammar! (This is a remark for developers of AI detectors.)
Stanford University researchers ran an experiment. It showed that people can tell whether a text is written by a human or artificial intelligence with a 50% probability. If you want to try your hand at identifying whether a text is real or AI generated, please take the quiz above. Good luck!