Different approaches in detection of machine-generated texts
Daniel Georgiev
Software Consultant, Accedia
I have been a Software Consultant at Accedia for the past few years with .NET as a primary platform. I have participated in the development of several enterprise-grade solutions in the Fintech industry. Furthermore, I am fascinated by the advancement in the AI field which was the reason to choose it for my master’s degree.
Without a shadow of a doubt machine text generation has become a work of art in the past years. Articles, posts, reviews and other are flooding the internet space. An important question that people have to answer is how to detect if a piece of text is written by a human or a large language model. In this talk I am presenting the challenges of this task by describing how some solutions work under the hood and whether sufficient results can be achieved without training a specific model for the problem. In addition, there is going to be a demonstration that shows the performance with several paragraphs.