5 Ways to Detect Texts Written by ChatGPT and Other AI Chatbots

TEMPO.CO, Jakarta - The presence of ChatGPT made by OpenAI since November 30, 2022 has given rise to a number of pros and cons. Artificial intelligence (AI) technology is sometimes viewed as only an aid to ease human labor. However, many people are concerned that comparable "chatbots" might replace certain jobs.

Now, what is a chatbot? a chatbot is a computer program that simulates human conversation with an end user. Not all chatbots are equipped with artificial intelligence (AI), but modern chatbots increasingly use conversational AI techniques such as natural language processing (NLP) to understand user questions and automate responses to them.

Several parties have banned the use of ChatGPT in various institutions, including schools. As in Australia, a spokesperson for the Queensland state Department of Education said it had banned the use of ChatGPT for students. This rejection is the result of concerns about making student assignments completely handed over to AI.

So, how do you detect texts written by ChatGPT and other AI chatbots?

How to Detect Texts Written by ChatGPT and AI Chatbots

Data scientist at the Mozilla Foundation, Jesse McCrosky, said that there is no easy method for detecting text written by AI. The tools currently available have limitations. “Detection tools are always imperfect, making them virtually useless for most applications,” McCrosky said.

According to him, it is impossible to create an AI detection tool, because software with an artificial intelligence system can compile text created specifically to avoid this kind of detection. "There may be a race between ChatGPT text detectors and people avoiding tracking tools, but there will never be a situation where the detectors can be completely trusted," he said.

Even so, the Poynter Institute offers some tips for spotting misinformation in ChatGPT posts, namely checking for text patterns and inconsistencies. Additionally, readers can look for signs of errors and check the context of the writing.

Sites to Detect Texts Written by ChatGPT and AI Chatbots

Researcher at the University of New South Wales (UNSW), Md Badiuzzaman, shared platform recommendations for checking texts written by ChatGPT and other AI chatbots. Here's the list:

1. OpenAI API Key

The ChatGPT development company, OpenAI actually provides a tool to detect AI text, namely the OpenAI API Key https://platform.openai.com/ai-text-classifier. However, this tool can only be used to track text with a minimum of 1,000 characters (150-250 words). Apart from that, this tool is still being refined starting July 2023.

2. GPT-2 Demo Detector

Readers who feel they have found writing that indicates AI creation can use the site created by GPT-2 at https://openai-detector.hf.space. Badiuzzaman, who tested the capabilities of this platform, using texts from ChatGPT, said that 94 percent of the texts were fake. However, when he tried to insert a partial paragraph from his paper, GPT-2 said it was 99 percent original.

3. GPTZero

GPT Zero via the site https://app.gptzero.me/app/welcome detects confusion in an article. The ambiguity in question refers to how many possible words are suggested by the automatic algorithm. The reason is, the sentences produced by AI usually consist of similar words and phrases.

4. DetectGPT

The University of Stanford launched an AI writing detection tool called DetectGPT. Internet users can use it by accessing the site https://detectgpt.ericmitcell.ai/. However, this platform is only limited to checking GPT-2 generated text.

5. Writer.com

How to detect ChatGPT results can also be done via the site https://writer.com/ai-content-detector. This page can extract AI-generated text. When Badiuzzaman tried it, Writer.com identified his paper as 99 percent man-made.

MELYNDA DWI PUSPITA

Editor's Choice: Singapore Listed as Country with Strongest Passport, Here's How the Rest of Southeast Asia is Ranked

Click here to get the latest news updates from Tempo on Google News