Imperfectly Perfect AI Chatbots: Limitations of Generative AI, Large Language Models and Large Multimodal Models

Generative AI has transformed the landscape of AI chatbots in a way that they are increasingly becoming human-like in their understanding and working. Generative AI is used for creating new contents, and Large Language Models (LLMs) and Large Multimodal Model (LMMs) are a type of generative AI. Most modern AI chatbots are designed using LLMs and LMMs, which provide them with the power of generative AI. LLMs deal with the text, whereas LMMs deal with more modalities including text, image, audio and video. This continuous enhancement of generative AI models is bringing AI chatbots closer to the human working. Nonetheless, these AI chatbots are not perfect, as they have several limitations, and their users should comprehend before fully relying on these imperfectly perfect AI chatbots. Depending on the type of use of AI chatbots, these limitations may or may not impact them significantly. This paper will organise the limitations of AI chatbots into six main categories: intelligence and understanding related limitations; accuracy and credibility related limitations; ethics and regulations related limitations; accountability, transparency and consistency related limitations; design, coding and training related limitations; and human, machine and vender related limitations. The aim of this paper is to emphasise, organise and analyse numerous limitations of AI chatbots into the proposed categories; accordingly, users should become more cognisant about all these limitations in deciding their suitability or unsuitability for the specific use cases.