Halucynacje chatbotów a prawda: główne nurty debaty i ich interpretacje
Chatbot Hallucinations vs. Truth: Mainstream Debates and Their Interpretations

Summary/Abstract: Generative artificial intelligence (AI) systems are able to create media content by applying machine learning to large amounts of training data. This new data may include text (e.g. Bard by Google, LLaMa by Meta, or ChatGPT by OpenAI) and visuals (e.g. Stable Diffusion or DALL-E by OpenAI) and audio (e.g. VALL-E by Microsoft). The level of advancement of this content may make it indistinguishable from human creativity. However, chatbots are characterized by the so-called hallucinations, which are largely a new type of disinformation. The aim of the research undertaken is to identify the main trends in the debate on the effects of the use of artificial intelligence, with particular emphasis on disinformation involving chatbots in the media environment. The study adopted the research method of a systematic literature review, which, among other things, limits selection bias. The interpretation of the main trends in the debate leads to the conclusion that the disinformation of chatbots in the form of their hallucinations is significant in terms of scale, is optimized and personalized, and has a significant potential to erode social trust.

• Details
• Contents