to heat a candle with the warmth of you
Every machine-generated thought consumes power and produces heat —a silent, residual trace of its activity. While we cannot see energy directly, its byproduct heat offers a way to feel and sense the invisible processes shaping our world.
Addressing the energy consumption of artificial intelligence through the medium of heat, the project directs the heat waste towards a candle. Running a large language model (LLM) on a small raspberry pi, the self-referencing system is prompted with “to heat a candle with the heat of you”. Addressing the LLM with "you" it will start to generate an answer that does not reflect how the system is actually heating the candle.
The resulting heat from the computation slowly melts and erodes the candle in the physical world.
