Introduction

Explain to the learners that this activity will not resolve all their problems and will not make them experts in detecting AI-generated content but  it can show them key details worth paying attention to in the future, so they will not be misled by AI-generated content. When ChatGPT became available for the public many newspapers and blogs started to experiment with the program just to see how well it performs in real life scenarios and if people can detect non-human written content? While the program sometimes commits mistakes its productivity is astonishing and can free up lots of time for writers. But it needs revision. A perfect solution could be that content creators proclaim if their content was partly or entirely written by AI. For now it is just a utopia, thus we have to educate ourselves. AI is already deceiving us; according to Cornell University researchers, about 66% of the time, people perceive fake news articles generated by GPT-2 as credible. There is AI-detecting software but it is not entirely consistent. When greater amount of text is available detection can be easier, typos, commas might be signs that the content is human written, also predictable text with standard length sentences and recurrent wording might be signals that the content is AI-written. The whole problem lies within the speed of the process. The velocity of development in this sector means that every way to spot AI-generated text becomes outdated very quickly.


Show the group the following video (in English, auto-generated captions are available for other languages:
(00:00 - 07:33)


10 min
cookieCookies help us deliver our services. By using our services, you agree to our use of cookies.