Google’s new AI robot to mimic Human speech sound
The new AI robot, dubbed WaveNet, has been built by the same team that was responsible for developing AlphaGo.
Google’s DeepMind team has developed a artificial intelligence (AI) robot which can mimic human speech sound, duping humans to think they are talking to another person.
The new AI robot, dubbed WaveNet, has been built by the same team that was responsible for developing AlphaGo — the robot that defeated some of the world’s best GO players.
Also Read: Now, Google Allows You to Share Paid Apps With Upto 6 People Through Family Library
Artificial speech programmes are hard to create and developers currently use only two methods to methods for the same.
The process involves collecting a large pool of words and speech fragments, spoken by a single person, making sounds and accent difficult to manipulate, according to technology website Engadget.
However, the other process involves forming word’s electronically, depending on how they’re supposed to sound; the easier sound may be easier to tweak, but the sound ends up being robotic.
DeepMinds had to feed the neural network raw audio waveforms recorded from real human beings to create the AI speech program. WaveNet speaks after forming individual sound waves.
People who signed up for the blind test and spoke to WaveNet thought it sounded better and ‘more human’ more original than past programs. The results are based on tests conducted in Mandarin Chinese and English languages.
Also Read: Google to train two million mobile developers in India
Google’s DeepMind team has developed a artificial intelligence (AI) robot which can mimic human speech sound, duping humans to think they are talking to another person.
The new AI robot, dubbed WaveNet, has been built by the same team that was responsible for developing AlphaGo — the robot that defeated some of the world’s best GO players.
Also Read: Now, Google Allows You to Share Paid Apps With Upto 6 People Through Family Library
Artificial speech programmes are hard to create and developers currently use only two methods to methods for the same.
The process involves collecting a large pool of words and speech fragments, spoken by a single person, making sounds and accent difficult to manipulate, according to technology website Engadget.
However, the other process involves forming word’s electronically, depending on how they’re supposed to sound; the easier sound may be easier to tweak, but the sound ends up being robotic.
DeepMinds had to feed the neural network raw audio waveforms recorded from real human beings to create the AI speech program. WaveNet speaks after forming individual sound waves.
People who signed up for the blind test and spoke to WaveNet thought it sounded better and ‘more human’ more original than past programs. The results are based on tests conducted in Mandarin Chinese and English languages.
Also Read: Google to train two million mobile developers in India
Comments
Post a Comment