ChatGPT’s new tool lets you differ between human or AI content
The AI Text Classifier was introduced by OpenAI to predict content from AI-generated and human-written texts.
artificial intelligence
Highlights
- OpenAI launched new plagiarism detector tool, called ‘AI Text Classifier’
- The new tool is launched to analyse plagiarism and reduce text abuse
- ChatGPT maker called AI classifier to be under feedback process and thus not fully efficient to be used
The rising use of this generative AI tool, ChatGPT, has led to several cases of copyright infringement. The growing cases have increased concerns among writers, illustrators, and academics. In the wake of such cases, some activists have raised this issue on Twitter, by highlighting the significance of patent protection and plagiarism.
In a blog post published on Wednesday, OpenAI announced the release of a tool called an 'AI Text Classifier' to discern between text produced by artificial intelligence and humans, including its own ChatGPT and GPT-3 models. The announcement was made amid the heated-up scenario among schools and students for not making the proper use of ChatGPT.
Is OpenAI's new ‘AI text classifier’ reliable?
According to the company's blog post, "The model currently has a 26 percent success rate and is not very accurate." Nevertheless, OpenAI asserts that it might be very effective when used in tandem with other methods, which could be useful in helping prevent AI text generators from being abused.
The Classifier intends to assist in reducing false claims that text produced by AI was written by a human, reports mentioned. However, ChatGPT’s new tool still holds drawbacks, so it can be incorporated into other methods of analysing the source of text rather than being the primary decision-making tool, as expressed by an OpenAI spokesperson.
The spokesperson further mentioned, "We intend to make this initial classifier offering available to the public for feedback and also to know if such tools are useful enough for the user." In the blog post, the company also stressed the effectiveness of the classifier in evaluating the content generated by human authors.
Training of AI classifier
According to several reports, 34 text-generating systems from five different organisations, including OpenAI itself, were used to train OpenAI's AI Text Classifier.
This content was trained with similar human-written text from Wikipedia, websites retrieved from links posted on Reddit, and a set of human demonstrations collected for a previous OpenAI text-generating system.
COMMENTS 0