Gesture AI investigates the concept of useless A.I., one that looks probable but in fact useless. The design allows users to control surrounding smart devices with custom gestures.
Looking at the current state of the art in artificial intelligence, I learn that DTW (Dynamic Time Warping) machine learning algorithm can classify temporal signals, and thus detect and recognize time-based inputs, such as motion and gestures.
This experiment reveals users can use their customize gestures (with a watch) to control surrounding devices (quick text rely). Yet, this "expressive and convenient" control only lives on the surface.
Unanticipated issues and outcomes begin to unfold as this interaction starts to play out in further use. Forget, Misinterpret, Self-learning, and Multi-trigger are 4 issues that could occur.
Therefore, it's more important to highlight questions and potential problems in this interaction domain:
How much customization can one allow to prevent forgetting?
When does AI know to listen and distinguish command from noise?
How much should AI learn about one and automate tasks?
How does a narrow AI live within a larger ecosystem of smart devices/AIs?
