Google has explained in a blog post how it actually manages to enable this level of human-computer interaction with the help of Google Duplex.
Google at the I/O 2018 made a host of announcements for the Android ecosystem, Google apps, and its self-driving car unit Waymo. However, in a first, Google introduced a new feature called Google Duplex as a part of the revamped Google Assistant. It is essentially a service that works in the background without involving the user. Google Duplex is an offline service powered by Google Assistant that make a booking on your behalf at the vendors and shops that have no or little online presence.
Amid the many artificially intelligent apps and features that Google introduced at the I/O, the Google Duplex comes with a more practical bent for the users who do not want the superfluous results from all the high-priced places from far away. The Google Duplex takes up orders from the users, such as making an appointment for a haircut at a salon, and performs them in the background. Google Assistant serves as the concierge between the local restaurant or business and the user.
Google has explained in a blog post how it actually manages to enable this level of human-computer interaction with the help of Google Duplex. The Google Assistant will make the conversational experience for the person taking up the bookings more comfortable and as natural as possible. Google has demonstrated in the blog post that Google Duplex uses a natural voice to make reservations at a restaurant with the help of AI to offer contextual and deeply trained answers.
“Google Duplex technology is built to sound natural, to make the conversation experience comfortable. It’s important to us that users and businesses have a good experience with this service, and transparency is a key part of that,” Yaniv Leviathan, Principal Engineer and Yossi Matias, Vice President, Engineering, Google said in a joint statement.
At the I/O, Google CEO Sundar Pichai demonstrated how Google Assistant is getting more than a human-like voice than it ever had. With the natural pauses and other vocalised pauses like ‘hmm’ and ‘umm’, Google Assistant talks to a salon to book a reservation, best suited for the date and time asked by the user. For the complex situations where there is a little pretext to a query, Google Assistant uses its Machine Learning prowess to offer the best solution to such questions.
For example, Google Assistant will call a restaurant to make a reservation for four people on Wednesday. However, the restaurant receptionist tells Google Assistant that no bookings can be made for less than six persons at a time. So, Google Assistant asks for other options – an upfront table without a booking in this case. There are many examples like this where Google Duplex will harness its recurrent neural network (RNN) to interact, with the help of TensorFlow Extended (TFX).
“To obtain its high precision, we trained Duplex’s RNN on a corpus of anonymized phone conversation data. The network uses the output of Google’s automatic speech recognition (ASR) technology, as well as features from the audio, the history of the conversation, the parameters of the conversation (e.g. the desired service for an appointment, or the current time of day) and more,” said the duo.
The Google Duplex technology will be rolled out as a part of Google Assistant is still in development and an experimental version will be rolled out to the developers and early adopters in coming months.