AI unveils a new tool to detect AI-generated texts //

OpenAI, creator of ChatGPT today released an AI Text Classifier. This could help you determine if the text you are reading was written by AI, or a human.

But there is a catch.

OpenAI claims it has not “examined the effectiveness of its classifier in detecting content written by human authors in collaboration.”

Why do we care? There are already many AI content detectors. These tools can identify AI-generated content and it is safe to assume that search engines can. Consider Google’s helpful Content System.

It may also be smart to use these tools to review any content submitted by freelance writers or agencies if you are publishing content. This is a new age and you don’t want to publish any content that could be flagged as publishing automatically created content. This is against Google’s guidelines. Search Engine Land has already begun to use tools such as this for all contributions.

What’s the AI Text Classifier? OpenAI describes its new tool as a “fine-tuned GPT modeling that predicts how likely a piece is to have been generated by AI from a variety sources such as ChatGPT.”

There are some limitations to the tool, most notably that it may not produce accurate results. OpenAI trained the tool primarily on English content written by adults. It is therefore very likely that false positives will occur. OpenAI also pointed out that AI-generated text can be edited to appear human-generated. Its FAQ section explains:

• When deciding whether an AI-generated document is valid, the results are not sufficient. The model was trained using text written by humans from many sources. It may not represent all types of text.

The classifier’s results Once you have copied your text (Open AI suggests that you should use at least 1,000 characters which is approximately 150-250 words), the AI Text Classifier will determine if the text was created using AI. One of the following labels will appear:

• It is very unlikely.
• Unlikely.
• It is not clear if it is.
• Possibly.
• Likely.

These are the results of my pasted-in text (human written, I swear).

The classifier believed it very unlikely that the article was AI-generated.

You can try the tool out for yourself. OpenAI’s AI text classifier. It’s free.

More coverage. This news was first reported Axios. You can find more coverage at Techmeme.

The post OpenAI launches a new tool to detect AI generated text was first published on Search Engine land.