AI Is A Self-Taught Speaker, Groundbreaking Technology

Artificial Intelligence is now the focal point of the efforts of many tech companies, with Google, Amazon, and Microsoft leading the charge. However, it would seem that the Chinese search engine firm Baidu is not content to sit on the sidelines as it recently created an AI that taught itself how to speak. This is virtually unheard of and is a huge breakthrough.

In the paper that was published by the company’ researchers, it would seem that an AI called Deep Voice was able to provide text-to-speech functionality derived only from deep neural networks. This means that it was able to synthesize neural speech on its own.

“The system comprises five major building blocks: a segmentation model for locating phoneme boundaries, a grapheme-to-phoneme conversion model, a phoneme duration prediction model, a fundamental frequency prediction model, and an audio synthesis model,” the paper reads. “By using a neural network for each component, our system is simpler and more flexible than traditional text-to-speech systems, where each component requires laborious feature engineering and extensive domain expertise.”

Basically, Deep Voice is able to bypass one of the rather tedious exercises of having to teach and then tweak the text-to-speech systems of other AIs like Alexa or Siri. This is a huge advantage when trying to create a responsive AI that actually speaks and behaves almost like a real human. Constructed speech is robotic. Dynamic speech, however, is flexible and can even be unpredictable.

More than that, it can also save researcher millions of hours having to record, embed, and then encode speech programs into AI in order to make them somewhat interactive, Futurism reports. This could then potentially free up the researchers to focus on the functionality aspect of the AI rather than wasting time one making sure that the machines can string several words together in a cohesive sentence.