DocLab Shorts 24 Nov at 19:45
A program of short stories that would never have existed without the internet.
A camera, a compass, and a clock narrate images based on location and time using machine intelligence.
Narrated Reality consists of three new interfaces for written language: a camera, a compass, and a clock that respectively narrate images, location, and time using machine intelligence. Each device is an antique embedded with a computer that handles all computation locally without using the internet. By taking a photograph, logging a location, or clocking in, users can generate a brief textual passage related to their environment. The text is created by a library of long short-term memory (LSTM) recurrent neural networks, machine learning algorithms that learn to write letter-by-letter after training on hundreds (or thousands) of books.