Sounding Exactly Like Human
It looks like Google’s DeepMind Artificial Intelligence (AI) never gets tired of taking on the challenge of being a human being. After trouncing the world’s Go grand master with its AlphaGo, it has now created what could be the most accurate speech machine yet.
Known as the WaveNet, it’s a voice synthesis program that can sample real human speech and directly produce an audio based on the same. Even though text to speech programs are progressively becoming important in computing, users have been relying on not-so-good bots and programs such as Amazon’s Alexa, Microsoft’s Cortana and Apple’s Siri. While these programs often rely on repeating actual recordings of human voices, WaveNet can actually produce a completely computer-generated voice.
What makes this entire process interesting is the fact that WaveNet can generate not less than 16,000 samples in a second to create its own audio samples with minimal human involvement. This is a program that’s designed to mimic how some parts of the human brain function and do exactly the same.
While it’s too early to tell the implications that WaveNet might have on us, it will certainly change the way humans interact with machines. The future is here.