The artificial intelligence of Google could eventually learn to talk

DeepMind, a division of Google engaged in the field of artificial intelligence has just announced the latest achievement in research. The company has just managed to improve the ability to speak like a human computer. Google acquired a company in 2014 that claimed to have developed a computer program that can talk called WaveNet.


“Allowing people to communicate with machines is an old dream of human interaction with computers,” said DeepMind in his blog page.

Unlike artificial sounds that exist today, WaveNet focuses on the opposite sound waves created by the language itself.

It uses a neural network computer – technology that seeks to imitate the human brain – to analyze waveforms ‘raw’ from the audio signal and the diverse sounds, including types other sounds like music.

As a result, the sound output WaveNet better than other computer sound like Google Now, Siri from Apple, and Alexa of the Amazon.

Although WaveNet voice capabilities better than other artificial voice, the computer still needs a lot of data before it is actually used for practical needs. Therefore, this technology is not going to be integrated with Google products in the near future.

For information, DeepMind known as a developer of artificial intelligence that is specifically designed for playing games such as Space Invanders and Go.

In fact, AlphaGo designed for playing Go was recently defeated world champion Go South Korean, Lee Sedol.

The success was called a milestone in the study of artificial intelligence. Because, Go is a game that has a lot of unexpected movement rather than a chess are known to have a certain pattern.