OpenAI’s ChatGPT emerged as a prominent example of AI’s utility for individuals, businesses, and industries. However, it’s not flawless and may sometimes provide inaccurate or incorrect responses, causing controversy and landing people in trouble.
Despite its usefulness, ChatGPT has faced criticism for its occasional inaccuracies and errors, which have even led to trouble for its creators, OpenAI.
In 2023, it wrongly accused an American law professor of misconduct during a school trip, citing a non-existent Washington Post article. The entire incident was fabricated.
Despite its impressive abilities, ChatGPT has alarming flaws. While it can provide useful information, its responses should not always be accepted as truth.
With more people relying on ChatGPT for online tasks or work, it’s likely that much of the information encountered online is generated by AI, including ChatGPT.
The internet is becoming a place where it’s hard to distinguish between human and AI-written content. While advancements like ChatGPT-5 may address this in the future, it’s crucial for people to remain vigilant and recognize when content is generated by ChatGPT.
How can you tell if ChatGPT wrote something?
Since humans trigger ChatGPT’s responses, the detail in its replies depends on the detail of the prompt. Without clear instructions, especially on complex topics, it may provide vague or inaccurate answers.
To someone unfamiliar with the topic, like those lacking background knowledge, this may not be apparent. However, to others, it could be obvious that the text originated from ChatGPT.
These are the main indicators to watch for…
Language Usage and Repetition
ChatGPT is classified as ‘narrow’ AI, meaning it can’t comprehend or mimic human emotions or behavior. It also lacks independent thought, resulting in responses that may lack personality or creative language.
Although it can make errors, ChatGPT is trained to minimize them, often resulting in simple, robotic responses to reduce inaccuracies.
This becomes apparent when requesting tasks like writing a review for a favorite film or product, where it might omit crucial details like actors’ names or product dimensions.
If you notice missing vital information in a review, there’s a chance it was generated by ChatGPT.
Similarly, ChatGPT may exhibit repetitive language patterns, especially in longer passages, despite being trained on extensive language data.
Visualizations
The case involving a law professor illustrates an AI hallucination, where it fabricates information in its response. This remains a significant issue with generative chatbots like ChatGPT.
AI experts advise fact-checking ChatGPT’s responses, especially when dealing with niche topics. There are numerous instances of hallucinations, varying in severity, underscoring the importance of verifying its accuracy with other sources before relying on it.
If you’re knowledgeable about a topic, it’s easier to spot inaccuracies. For instance, a match report on a soccer game you watched would be straightforward to assess for factual errors. However, researching more specialized subjects, like the thermic effect of food, may present challenges in distinguishing between human and AI-generated content.
Mistakes from Copying and Pasting
This is perhaps the most noticeable mistake. People sometimes unintentionally copy and paste ChatGPT’s responses, including its side comments like ‘Sure, here’s a movie review for…’
It’s a clear indication whether what you’re reading was written by ChatGPT or a human, based solely on human error rather than an error made by the AI.
Read the text carefully.
ChatGPT is designed to sound and respond like a human, making it difficult to distinguish if something was written by AI based on just one or two sentences. It becomes clearer when reading the entire article thoroughly, which may reveal certain errors like hallucinations, repetition, general language, or copy-and-paste mistakes.
While humans can edit ChatGPT’s responses to eliminate many of these issues and make the text more human-like, this often requires extensive editing, making it easier to write the text themselves.
Detecting ChatGPT Content
The emergence of ChatGPT and similar AI chatbots has led to the development of various AI content detectors. These tools claim to identify whether a text is written by a human or AI.
Some of the best AI content detectors can pinpoint which parts of the text are human and which are AI-generated. They may also provide a percentage estimate of human versus AI content.
However, these AI content detectors are not flawless. There are instances where they mistakenly identify human writing as AI and vice versa.
Using an AI detector can help identify any AI elements in the text, especially if there are lingering doubts after checking for repetition, hallucinations, and general language.
When dealing with niche subjects you’re not entirely familiar with, it’s essential to fact-check ChatGPT’s responses. It’s better to be cautious than to rely solely on AI-generated information.