New Advances In AI From Google Acquisition DeepMind

Your next conversation could be with an artificially intelligent robot. Alphabet Inc.’s Google (GOOG) acquisition DeepMind just revealed WaveNet, an AI text-to-speech system that has reduced the gap with human speech patterns by 50% from previously similar systems. According to the company, the same AI model used in WaveNet can be used to synthesize other audio signals, such as music. (See also: 5 Lessons From Google’s Machine Learning Development.) Modeling human speech is a problem because it requires a sampling speed of 16,000 samples per second and contains numerous complexities, such as scale and different patterns reflected in emotions. A blogpost on DeepMind’s site explains the process used by the outfit’s researchers to tackle this challenge. Essentially, they used an autoregressive approach in which future sampling predictions were based on…


Link to Full Article: New Advances In AI From Google Acquisition DeepMind