What is Google Duplex?
Google Duplex is a new technology developed by Google for conducting natural conversations to carry out “real world” tasks – such as reserving a table at a restaurant or a scheduling an appointment for a haircut at your favorite hair salon – over the phone.
What is cool about this technology is that it can perform surprisingly natural conversations, allowing people to speak normally, like they would to another person, without having to adapt to a machine.
Below, you can listen to one of the first examples of how this technology works that was presented on the 8th of May at Google I/O 2018.
Behind this simple and flawless conversation, there is a deep training through machine learning, but the result is far beyond any human-machine interaction you have ever experienced through voice.
What do we mean by surprisingly natural?
When people talk to each other, their conversations are extremely complex when compared to the conversations they carry on with machines – i.e. when using voice search or their personal assistant. They often correct themselves in the middle of a sentence, use more words than necessary, or omit parts of the sentence and rely on context instead; they also express a wide range of intents.
Google Duplex’s conversations sound natural thanks to advances in understanding, interacting, timing, and speaking. The system also sounds more natural thanks to the incorporation of speech disfluencies (e.g. hmms and uhs). Even latency has been perfectioned to match people’s expectations.
At the end when you listen to a conversation like the one above, you wouldn’t say that one of the voices is not human.
For users, Google Duplex is making supported tasks easier. Instead of making a phone call, the user simply interacts with the Google Assistant, and the call happens completely in the background without any user involvement.
Even if the first recordings are extremely well performing, Google Duplex is still a prototype, and it has not been launched in the market yet. Before its integration with the Google Assistant, Google promised it to become more transparent, announcing people that they are talking with a machine when it performs calls.