You are on page 1of 13

1

Jens Martensson
Jens Martensson 2
3

Google Duplex
• Google Duplex, a new technology for conducting natural
conversations to carry out “real world” tasks over the phone.
• At the core of Duplex is a recurrent neural network (RNN)
designed to cope with these challenges, built
using TensorFlow Extended (TFX). To obtain its high precision
• Natural conversations

Jens Martensson
History

• Google Assistant was unveiled during Google's developer conference on May 18, 2016, As a part of the Google smart
home speaker and app ALLO.
• Designed to be a conversational and two-way experience.
• December 2016, Google launched "Actions on Google", a developer platform for the Google Assistant.
• In March 2017, Google added new tools for developing on Actions on Google to support the creation of games for the
Google Assistant. And Made available to android and IOS in May 2017.
• In April 2017, a software development kit (SDK) was released, allowing third-party developers to build their own
hardware that can run the Google Assistant
• In May 2018, Google revealed Duplex, an extension of the Google Assistant that allows it to carry out natural
conversations by mimicking human voice

Jens Martensson 4
Jens Martensson
Jens Martensson
Layer 3: System Model
How Google Duplex works with other application.

Jens Martensson 7
Layer 2: Software Application

• Big Data
• Tacotron and WaveNets are able to generate speech which mimics any human voice and which sounds more natural than
the best existing Text-to-Speech systems, reducing the gap with human performance by over 50%.
• TensorFlow Extended (TFX) is a TensorFlow-based general-purpose machine learning platform implemented at Google

Jens Martensson 8
Layer 1: Algorithm

• Automatic Speech Recognition (ASR)


• Text to Speech (TTS)

Jens Martensson 9
Layer 1: Algorithm
At the heart of Google Duplex: Recurrent Neural Network

which h(t) is the hidden state at timestamp t, ∅ is the


activation function (either Tanh or Sigmoid), W is the weight
matrix for input to hidden layer at time stamp t, X(t) is the
input at time stamp t, U is the weight matrix for hidden layer
at time t-1 to hidden layer at time t, and h(t-1) is the hidden
state at timestamp t

Jens Martensson 10
Google Duplex Triggering Evolution

• IT also started helping address accessibility and language barriers, e.g., allowing hearing-impaired users, or users who
don’t speak the local language, to carry out tasks over the phone.
• Technology advances will ultimately contribute to a meaningful improvement in people’s experience in day-to-day
interactions with computers.
• It can be asynchronous, so you can make the request and then go offline while Google Duplex gets on with the job: it will
report back when you’re online again. That’s useful in areas of patchy connectivity, or if you’re just really, really busy.

Jens Martensson 11
Views on Google Duplex

• Performance vs Preference:
• Ethical Issue
• Invasion of Privacy
• Uncanny Valley:
• In May 2018, ethical concerns marred its announcement
• Duplex is currently in development and have a limited release
in late 2018 with Google Pixel users. During the limited
release, Pixel phone users in Atlanta, New York, Phoenix, and
San Francisco will only be able to use Duplex to make
restaurant reservations.

• Ref:
• https://www.itworldcanada.com/article/google-duplex-broke-the-
turing-test-and-leaps-across-the-uncanny-valley/405200
• https://medium.com/@canvas8/brand-insights-google-duplex-ai-
uncanny-valley-82edb3e499d2

Jens Martensson 12
Thank You
Minh & Sai
At Fleming College

You might also like