ChatGPT Maker Releases New Tool That Detects AI-Generated Text To Limit Abuse
When OpenAI's ChatGPT was released to the public last year, it astounded and pleased users around the world who used the artificial intelligence tool for an array of purposes - to write songs, to answer conflicting questions, and much more.
However, ChatGPT's fame also made concerns surrounding it apparent - especially for students who began to use it to generate articles.
Now, the company that made ChatGPT, OpenAI, has launched a tool that essentially distinguishes between text written by humans and text written by artificial intelligence, including its own ChatGPT and GPT-3 models.
Fighting abuse of AI-generated content
For now, the model isn't highly accurate and has a success rate of 26%. Regardless, OpenAI claims that when paired with similar tools aimed at reducing the potential abuse of such tools, it could yield significant success.
The tool would essentially "mitigate false claims that AI-generated text was written by a human. However, it still has a number of limitations — so it should be used as a complement to other methods of determining the source of text instead of being the primary decision-making tool," an OpenAI spokesperson told TechCrunch.
Also read:?ChatGPT Clears Law School Exam In US, Writes Essays On Taxation And Constitution
Initially, the tool is being made available to get feedback from its users in hopes of making it even more robust in the near-future.
Many universities and schools around the world have proceeded to ban the usage of ChatGPT on their networks and devices. It's important to note here that ChatGPT isn't completely accurate and is susceptible to factual errors.
ChatGPT's latest tool is called OpenAI AI Text Classifier which is a language model that was trained on publicly available text from the internet. Unlike ChatGPT, it can predict if the text was generated by AI - whether it is ChatGPT or some other AI tool.
It appears that OpenAI AI Text Classifier was trained on text from 34 text-generating tools from five different organisations, including OpenAI's ChatGPT.
For now, the OpenAI Text Classifier will work on text that has a minimum of 1,000 characters - 150 to 250 words. It is unable to detect plagiarism and it could get things wrong in languages other than English.
Also read:?Why Meta's Artificial Intelligence Head Thinks ChatGPT Isn't 'Particularly Innovative'
The tool labels text as "very unlikely" to be generated by AI - with a chance of less than 10%, “unlikely” - between a 10% and 45% chance, “unclear if it is” AI-generated with a 45% to 90% chance, “possibly” AI-generated with a 90% to 98% chance ?or “likely” AI-generated with a chance above 98%.
While the tool isn't perfect, it's a welcome step to build tools that are able to limit the use of text-generating tools like OpenAI's ChatGPT.
What do you think about this tool? Let us know in the comments below.?For more in the world of?technology?and?science, keep reading?Indiatimes.com.? ?
