Detection when the text was generated by tools like Cat is a difficult task. Popular artificial intelligence-The detection tools, like GPTZERO, can provide advice to users by telling them when something has been written by a bot and not by a human, but even specialized software is not infallible and can spit false positives.
As a journalist who began to cover AI detection over a year ago, I wanted to organize some of the best Wired articles on the subject to help readers like you better understand this complicated problem.
Have even more questions about identifying chatgpt outings and others Chatbot tools? Register for my Newsletter unlocked from AIAnd contact me directly with everything related to AI to which you would like to answer or want to explore more.
February 2023 by Reece Rogers
In this article, which was written about two months after the launch of Chatgpt, I began to fight against the complexities of the detection of AI text as well as the AI revolution could mean for writers who publish online. Edward Tian, the founder behind GPTZEROTalked me about how its AI detector focuses on factors such as variance in text and chance.
As you read, focus on the section of the text filigree: “A watermark could be able to designate certain models of words to be prohibited for the AI text generator.” Although a promising idea, the researchers with whom I spoke were already skeptical of its potential efficiency.
September 2023 by Christopher Beam
A fantastic piece of the October number of last year, this article gives you an overview of the state of mind of Edward Tian while working to extend the capacities of reach and detection of GPTZERO. The emphasis on the way AI had an impact on school work is crucial.
AI text detection is a level of mind for many educators in class because they note the papers and, potentially, completely give up test duties due to students by secretly using chatbots to finish duties. While some students can use generating AI as a brainstorming tool, others use it to Make entire assignments.
September 2023 by Kate Knibbs
Are companies responsible for reporting products that could be generated by AI? Kate Knibbs has studied how much books generated by copyright have been listed on the sale on AmazonEven if some startups thought that products could be identified with special and deleted software. One of the main debates on AI detection depends on the question of whether the potential of false positives – a text written by humans which is accidentally reported as the work of AI – on the advantages of the labeling of the content generated by algorithm.
August 2023 by Amanda Hoover
By going beyond duties, the text generated by AI appears more in university journals, where it is often prohibited without a Appropriate disclosure. “The articles written by AI could also draw the attention of good work by diluting the basin of scientific literature,” writes Amanda Hoover. A potential strategy for solving this problem is that developers create specialized detection tools that seek AI content in articles evaluated by peers.
October 2023 by Kate Knibbs
When I spoke with researchers for the first time last February of the filigranes for the detection of text of the AI, they were full of hope but prudent of the potential to print the text of AI with specific language models which are undetectable by human readers but obvious for detection software. With hindsight, their apprehension seems well placed.
Just a semester later, Kate Knibbs spoke with several sources that broke the AI filigranes and demonstrating their underlying weakness as a detection strategy. Although it is not guaranteed to fail, the text of the watermark on the AI continues to be difficult to achieve.
April 2024 by Amanda Hoover
A tool that teachers try to use to detect the work in class generated by AI is VirtinPlagiat detection software that has added AI identification capacities. (Turnitin belongs to Advance, the parent company of Condé Nast, which publishes Wired.) Amanda Hoover writes: “Chechitelli says that a majority of service customers have chosen to buy AI detection. But the risks of false positives and prejudice against learners in English have led certain universities to abandon the tools for the moment. »»
AI detectors are more likely to falsely label the written content of someone whose first language is not English As a person of someone who is a native speaker. While developers continue to work to improve AI detection algorithms, the problem of erroneous results remains a central obstacle to overcome.