Google introduced its new technology Google Duplex at Google I/O 2018 conference which is shocking and delightful in equal measure.
What is Google Duplex
It is a new technology that takes conversational intelligence and conversational interfaces to altogether another level.
Google Duplex can conduct natural conversations over the phone using Google Assistant. Specific tasks like scheduling appointments over the phone can be done using this technology.
The main purpose of the technology is to ensure a comfortable conversational experience with a machine or computing device.
This includes getting your appointments scheduled without the need to pick up the phone, dial a number and talk to the person on the other end.
Google Duplex – Technology Involved
The core of Duplex involves recurrent neural network (RNN) built using TensorFlow Extended (TFX).
RNN is a class of artificial neural network which find heavy applications in handwriting recognition and speech recognition
TFX is TensorFlow-based machine learning platform by Google.
In order to sound natural to the user on the other end, Google uses a concatenative text to speech (TTS) engine and a synthesis TTS engine.
To train the system on a new domain, Google uses real-time supervised training.
Google Duplex Demo
There are several demos showing Google Duplex in action. We are sharing some of them below from Google’s Blog:
[audio src="https://www.gstatic.com/b-g/KOK4HAMTAPH5Z96154F6GKUM74A3Z1576269077.mp3" /]
What About Complex Asks
Google’s Duplex is capable of handling simple to tough scheduling or appointment queries.
However the system has self monitoring capability where for complex queries a human operator is signaled to complete it.
[audio src="https://www.gstatic.com/b-g/BT5EH08P73O41Q94PTWNMV42DAWU8Z192313240.mp3" /]
Duplex Handling Interruptions
[audio src="https://www.gstatic.com/b-g/OROY9DN8QUHYUN1VED9V1QS0387EOX207713725.mp3" /]
[audio src="https://www.gstatic.com/b-g/YBFPW2YQBZPVP4WONSIUO24KV82NY32653447.mp3" /]
Duplex Calling a Restaurant
[audio src="https://www.gstatic.com/b-g/5717BWTLRKCBB8JUDQ1QUA6HMC26WL238301173.mp3" /]
Duplex Calling for Holiday Hours
[audio src="https://www.gstatic.com/b-g/N6WGB2KL3NZCVUJAGOX3BLZRJ31BRP278409629.mp3" /]
Duplex Calling to Book a Meal
[audio src="https://www.gstatic.com/b-g/1RYL7HFNYMJUOLXWHMB4TOYIDA7YFR191723752.mp3" /]
For many years now, researchers and technocrats across the world have craved for machine and human interaction where the conversation is natural.
For too long now, users have been forced to adapt themselves to the system instead of the system adapting to the user.
Duplex comes in to take care of this. As a user, you don’t have to remember or train yourself to the syntax that the machine understands. Conversations are as natural as it happens on a normal day.
What next? Google has already declared that it will continue working on Duplex to ensure that complex scenarios are handled well by the system. Of course that has to be done if the potential of artificial intelligence is to be fully understood.
The future however is to transition from human-machine interaction to machine-machine interaction developed using different technologies. For example, if today Duplex is having a conversation with a human user; the future will see more interaction where both sides are non-humans/machines.
Given the blockchain-enthusiast that I am, the future will see smart contracts, artificial intelligence and conversational interfaces working in tandem to provide a seamless experience. That is what the later half of the twenty-first century will look like, according to me.
PS: All the above conversations/audio involving Google Duplex have been taken from Google’s Blog linked above.