Google, a tech giant has developed new Deep Mind AI (Artificial Intelligence) along with Oxford University. It is recognized as WLAS (Watch Listen Attend and Spell) tool, which can lip-read from a video unedited. Googles AI such helps in hearing impaired people. Google explained how it can be functioned in a recent paper release which can interpret better than humans with much accuracy.
Googles AI can anticipate more words than a human. The team trained this system with the videos from BBC with a dataset of more than 100000 natural sentences 17500 individual words.
GOOGLE’S AI BEATS HUMAN EXPERTS IN LIP READING
Human lip reading professional having more than 10 years of relevant experience was given the same set of videos given to Googles AI. After comparing both the human and the system, Google’s AI have more than half of the words deciphered when compared to human who have less than quarter of the words guessed by the AI.
According to the report of New Scientist Google’s AI decoded 46.8% and human traced 12.4% of their words. This is a clear substantial difference made by WLAS. WLAS also have been trained with the way humans spell to make it more human friendly.
The system also has being trained to learn and speak multiple languages, interlingua, used by different kind of people around the world. Researchers have also said that it is trained well to speak a whole sentence instead of using phrases or word by word to express what it had seen anyone talking.
Also Microsoft recently has attained some AI milestones. Microsoft said that the AI was capable to recognize conversational speech better than humans who can make it professionally.
Earlier Oxford released related paper work which can perform lip work called lipset. Lipset has more level of accuracy when compared to human professional related to similar work. It processed with 93.4% accuracy compared to human with 52.3% accuracy which is far more.
This Googles AI can help virtual assistants like siri if they are connected to the users camera. And such can help its owner by having a look on his/her lips rather than texting it. It will be more helpful even if it is used in noisy and disturbing environments.