Photo by At the intersection of art and technology, innovators are pushing the boundaries of human-machine interaction. One such experiment was showcased at the Lampu festival in Bali, where a novel AI confessional booth was built, inspired by the principles of Catholic confessionals and the capabilities of current Large Language Models (LLMs).
A glimpse into the AI confessional booth
The AI confessional booth allowed anyone to engage in conversation with an artificial intelligence, sparking interesting discussions and interactions. The project was a proof of concept, built to test the feasibility of integrating a voice interface with an LLM. The initial implementation involved a simple pipeline consisting of Speech-to-Text (STT), LLM, and Text-to-Speech (TTS) models.
However, several challenges arose during the first tests of this demo. These included response delays, ambient noise, and the need to mimic live conversation. To tackle these issues, the team had to carefully design the user experience of the booth. This involved deciding how to detect new users, recognize the beginning and end of a user’s speech, and provide feedback during delayed responses.
The architecture of the system consisted of three main components: STT, LLM, and TTS. The hardware used was an Arduino, specifically the ESP32 model with a built-in Wi-Fi module, connected to a laptop running the backend. The backend was responsible for handling the pipeline of STT, LLM, and TTS models.
Despite optimizations, a 3-4 second delay was still present. To mitigate this, the team implemented various UI feedback mechanisms, such as LED indicators, filler words, and a screen in the booth displaying different states with animations.
The results were impressive, with the AI confession room attracting hundreds of attendees over four days, at a cost of just $50 in OpenAI APIs. The experiment demonstrated that it is possible to add an intuitive and efficient voice interface to an LLM, even with limited resources and challenging external conditions.
The AI confessional booth in action
This innovative project highlights the potential of merging art and technology to create unique experiences that push the boundaries of human-AI interaction. As LLMs continue to advance, we can expect to see even more exciting developments in this space.
The intersection of AI and art
In conclusion, the AI confessional booth at the Lampu festival showcases the possibilities of integrating LLMs with voice interfaces, even on a limited budget. The success of this project demonstrates the potential for innovative applications of AI in various fields, from art to technology and beyond.
The future of AI is bright
