Google has been employing some highly unorthodox methods to train its DeepMind artificial intelligence. Today, Google has revealed that its UK-based DeepMind division have collaborated with scientists from the University of Oxford to develop the world’s most advanced lip-reading software. Their method of achieving it? Television shows.
According to published study, DeepMind and the University of Oxford has fed the A.I. thousands of hours of TV footage from the BBC to a neural network, training it to annotate videos based on mouth movement. The result of this test shows that the A.I. is capable of identifying what is being said on screen solely through mouth movement analysis with 46.8% accuracy. By comparison, a professional human lip-reader managed only a 12.4% accuracy rating.
While the technology is still in its early stages, Google is looking to use this part of DeepMind to assist individuals with impaired hearing. Other possibilities includes film annotating as well as communicating to digital assistants solely using lip gestures.
Source: The Next Web
