In this post, we will look at a simple test dataset that we use to test people and algorithms on how good they are at identifying AI generated content.
The dataset consists of 20 passages of text. Of these, 50% were written by a human, 40% were written by a GPT based AI model, and 10% were written by a different AI model.
You may think it’s easy to distinguish them, but it’s much harder than you may imagine. AI algorithms have gotten very good at producing text that looks believable — even though in many cases, it’s wrong.
In the quiz below, we have included 10 of those questions for you to try yourself. Can you guess which ones are AI?
Leave a Reply