Artificial Intelligence comes home.
Google Duplex is an extension of Google Assistant that connects users to businesses as unveiled by Sundar Pichai during Google I/O conference early May 2018. It can make phone calls to real humans. Many of us have been using the Google assistant for some time now and are excited by the freedom and power it brings along.
Deep recurring neural network that makes use of Wavenet natural speech synthesis technology including lip sounds and breathing allow these realistic conversations.
Google Duplex has been launched as a simple ‘booking system’ that books appointments on your behalf with real people. The conversations could be straight forward or complex as demonstrated in the conference.
It not only understands context and nuances but also responds like a human with pauses and filler words – “mm-mms” and “aha-s”.
Sometimes these words give time to computers for processing and at other times they just mimic real humans. It not only nails accents as well, but also handles interruptions, elaborates and responds to audio issues. Deceitful? Well maybe. Some people are raising concerns on ethical grounds but that’s a discussion for another day.
Does it pass the Turing test?
The Turing test, developed by Alan Turing in 1950, is a test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Turing proposed that a human evaluator would judge natural language conversations between a human and a machine designed to generate human-like responses. The Turing test was originally meant only for text conversations, but Google Duplex has taken it to another level passing it for voice.
In the domain of making appointments, it passes the Turing test.
The scope as stated by Google is currently limited to bookings or enquiring about opening hours and cannot hold real conversations. But we feel it is the tip of the iceberg.
There’s more… Here’s why.
Google DeepMind’s AlphaGo beat Lee Sedol the world master of the ancient and complex Chinese board game “Go” entirely by playing against itself. It achieved this feat by a self-learning technique known as ‘reinforcement learning’ with no tactical guidance and studying from moves of human experts.
So, Google, in principle, has the know how to further develop Duplex to a much more complex level cutting across accents, contexts and languages and is far more capable than portrayed; so, where does this leave the telemarketers and call centre industry?
In conclusion, it is an incredible technical achievement and super cool seamless implementation of AI in the real world. It is really exciting because we are at the brink of a brand-new era where AI, neural networks, machine learning, pattern recognition merge, opening doors to possibilities beyond imagination.
Before we leave, consider this scenario…
Wife: OK Google. Can you tell me where my husband was last night?
Google: mmm… Sure. I reserved a table for two for him at Michelin star restaurant Le Bernardin in New York. He also purchased a diamond bracelet from Tiffany and paid by MasterCard.
His GPS positioning also confirms that he was there till 1 am. I also heard him invite his friend, Clara, for a nightcap.
Shall I book an appoint with your family lawyer?
OK!! you can take a sigh of relief for now and prepare for the future as it may take a while for commercial roll-out for wide range of users. It has concerns and pitfalls, but one cannot ignore the fact the AI is going to be a part of our day-to-day lives soon, real soon.
by Priti Gupta