Google’s DeepMind Makes Talking Device A Reality

by CXOToday Newsdesk    Sep 12, 2016

Googles DeepMind artificial intelligence makes talking device a reality

What was earlier seen in sci-fi movies, just turned into reality, especially if Google’s DeepMind’s latest claims are to be considered. The company has developed capabilities, which will significantly lower the gap between the quality of speeches generated by computers and that of humans. The computer programme called WaveNet, will actually work on Artificial Intelligence (AI) technology, and will be the future roadmap for talking machines, which were earlier perhaps seen in sci-fi movies. In a blogpost, the scientists at DeepMind mentioned the capabilities of WaveNet, “Allowing people to converse with machines is a long-standing dream of human-computer interaction.”

In the existing line of artificial voice generators, the focus tends to remain on the language itself, instead of the soundwaves, which WaveNet has been mastering. The programme uses a neural network,  a human brain replicating technology, which analyses raw waves from an audio signal, and other different types of audio. To show how it works, DeepMind even published some samples with speeches in English and Mandarin, which as it turns out, seem to be a marked improvement on Google Now, Amazon’s Alexa, and even Apple’s Siri. Then there were samples of music which WaveNet reproduced after studying solo-piano music from YouTube.

However, despite the advent of this technology, Google would not be including the capability into it’s programmes anytime soon. The reason is being attributed to the high degree of computing power required by WaveNet to programme itself, making it impractical to be deployed on devices right now. It currently requires a vast amount of data to train itself, as it uses the existing TTS datasets to function.

DeepMind sits under the umbrella of Google’s parent company Alphabet. They are known to work on AI systems which can be used on games like Space Invaders and Go, but Alphabet has been slow to integrate the technologies developed by DeepMind into it’s products. So far, just a single data center efficiency project had been announced so far, but it was on a global scale. More than the advancements, Alphabet now needs to work in tandem with Google, to ensure that practical utility is well-looked into before making a public launch of the same.