Photo by Tonik on Unsplash Introduction

Introduction

Recently, AI company OpenAI released a tool to tell if text was written by a human or not. The tool, called ChatGPT, is designed to be an equivalent of the famous Voight-Kampff test from Blade Runner and it may prove to be a valuable asset for those looking to identify AI-generated text. However, chatGPT’s accuracy may not always be 100%. In this blog post, we’ll take a closer look at OpenAI’s new tool and explore what it can and can’t do.

How does the tool work?

Photo by Joanes Andueza on Unsplash

You can paste text that is at least 1000 characters long into the tool’s text field. After analyzing the input, it will produce one of two different verdicts, stating that the text is not very likely to have been created by an AI or that it is a possibly AI-generated. To test the accuracy of the tool, I asked ChatGPT to create a piece of writing about bird migration patterns, which the detection tool afterwards labelled as “potentially AI-generated.” Meanwhile, It labelled a number of articles published by humans as “extremely unlikely AI-generated.” Therefore, despite the tool’s potential to raise false alarms in either direction, my (small sample size) test results indicate at least a modicum of accuracy. However, OpenAI advises against using this tool alone to assess the veracity of a piece of information because it works best for texts that are at least 1000 words long. After the startup’s ChatGPT tool, which generates AI-written text sometimes mistaken for human writing, was released in November, it came under attack from educators. This artificial intelligence platform could churn out graduate-level business and law school essays in seconds from basic text prompts and present a new way for students to cheat on their homework. The reason being New York schools banned the bot from their WiFi networks and school devices.
In spite of ChatGPT’s appearance on certain SEO-friendly articles of late, it’s really about automation for OpenAI. Earlier this month the company secured a $10 billion investment from Microsoft which plans to incorporate it into their search engine Bing and productivity app Office 365. OpenAI allegedly discussed selling shares at a $29 billion valuation late last year which would make it one of the most valuable US startups.

What are the implications?

Photo by yang miao on Unsplash

So although the tool could raise false flags in either direction my (tiny sample size) test suggests at least a degree of accuracy.
Even so, OpenAI warns against using the tool alone to determine content’s authenticity, and that it is best used for texts with at least 1,000 words. Since the startup released its ChatGPT tool in November, which produces content that sometimes passes for human writing, educators have pressured it to make improvements. Because of the pace at which the system writes essays, it is appealing to some students to use the automated system as an opportunity to cheat on school assignments. As a result, New York public schools have already banned the bot from their WiFi networks and school devices.
Last year, OpenAI discussed selling shares at a valuation of $29 billion, making it one of the most valuable startups in the country.
It is clear that OpenAI’s new ChatGPT tool has many implications both for educators and for businesses. Its accuracy in detecting whether text was written by a human or AI is still yet to be seen, but its potential for creating AI-written content that can pass as human-written makes it an intriguing tool to watch in the future.

To know more about OpenAI’s New AI classifier for indicating AI-written text check official blog from OpenAI with tool inside as well.

This article was originally posted on medium at https://medium.com/inkwater-atlas/openais-new-chatgpt-tool-may-help-you-tell-if-text-was-written-by-a-human-or-ai-a20e1906b76b