Exploring the Challenges and Potential of Generative AI: Insights from an Empirical Study

Abstract

In November 2022, ChatGPT marked a transformative moment in artificial intelligence, leveraging generative AI tools (GAI) and large language models (LLMs) to mimic human language effectively. Their rapidly gained popularity has reshaped a number of areas, presenting both challenges and opportunities. With their widespread adoption, concerns arise regarding the reliability of information generated by these tools, particularly in detecting fake news and addressing hallucinations. This study, conducted within the OpenFact project, involves an experiment with students tasked to generate and critically assess text produced by GAI tools. Results indicate a common occurrence of hallucinations in LLM outputs and a lack of transparency in information sources, posing challenges for practical applications. Furthermore, participants demonstrate the ease of manipulating GAI tools to generate false information, underscoring the risk associated with their widespread use. These findings contribute to understanding the limitations and implications of utilizing GAI tools in information verification processes and highlight the need for further research in this area.

Ähnliche Arbeiten