Implementation of an IoT Cocktail Machine Using ChatGPT API and ConvAnalyser in the Metaverse

Drinking alone or at home has increased especially among people in their 20s and 30s since the coronavirus disease pandemic started, which influenced the consumption patterns of alcoholic beverages. To help them make cocktails without recipes at home, many cocktail machines have been developed that create them according to predefined recipes after users manually pushed menu buttons. However, they cannot proactively accommodate user preferences or requests. Therefore, this paper proposes an Internet of Things (IoT) cocktail machine system that can recommend suitable cocktails to users based on conversations with the user and a large-scale language model, ChatGPT, in the metaverse. The proposed system consists of metaverse, ChatGPT, IoT server, and IoT cocktail machine components. The metaverse serves as a user interface, implemented by using the Unity engine to create a chat space, and the ChatGPT is generated from the prompt dataset that maps specific words to corresponding emotions (joy, sad, anger, and boredom) and colors (yellow, blue, red, and purple) given by the Plutchik’s wheel of emotions. During user’s conversations with it, the ChatGPT estimates the emotion and color related to specific words the user said and then recommends the cocktail with the color. If a user orders the recommended one or the other one he wants to drink, the ConvAnalyser in the metaverse sends the cocktail information to the IoT server. Finally, the IoT cocktail machine will create the ordered cocktail from this information. Experimental results verified that the proposed system properly recommended colorful cocktails to users based on their emotions and correctly created them.