"Press the button to send your request to your home factory as it begins to think. The internal computer heats up, generates heat, then releases a gentle puff of warm air, making the pinwheel spin. The stronger the wind, the harder the thinking! Beyond just "a feel," you also understand the cost. The Factory tracks the exact energy used for each prompt and breaks it down into familiar actions-brewing an espresso or doom-scrolling social media. Each prompt comes with its tiny energy bill."
"By bringing computation home and making its energy use transparent, the Hot Air Factory prompts reflection on sustainable computing and conscious AI use. The prototype uses a Jetson Orin Nano to run a series of open-source LLM models from Mistral and Meta. Depending on the energy level selected by the user, it runs either a smaller or a larger model, with the latter requiring more computation and energy to generate responses. The models run natively on the OS using llama.cpp and Ollamafile."
Hot Air Factory is a home device that visualizes and measures the energy cost of AI prompts by heating an internal computer and releasing a puff of warm air that spins a pinwheel. The device tracks exact energy used per prompt and translates consumption into familiar actions such as brewing an espresso or doom-scrolling social media. Users can schedule prompts for night shifts to run when the grid is less stressed and can select smaller or larger models to trade intelligence for energy savings. The prototype runs open-source LLMs from Mistral and Meta on a Jetson Orin Nano using llama.cpp and Ollamafile, and a mobile PWA communicates with the device over Wi‑Fi.
Read at CreativeApplications.Net
Unable to calculate read time
Collection
[
|
...
]