Google Duplex is a huge step towards the evolution of AI and Machine Learning. This piece of technology can lead humans and machine (computer) interaction to a whole new level.
The most impressive and also creepy part of Google duplex is that
” Being a computer, it sounds more like a human.”
Now, let me explain what is Google duplex? how it works? what is it capable of?
In May 2018, At Google I/O developers conference Google’s CEO, Sundar Pichai gave an explanation about Google duplex and experimental demo through assistant by making a live call for booking an appointment to a Hair salon and restaurant.
What is Google Duplex?
It is an AI project by Google, which allows the assistant to make a call and lead a human-like conversation with real-life people and carry out the specific tasks given by the user.
Ex: If you ask google assistant to book an appointment for a restaurant, it makes a call on behalf of the user, speaks to person on call, and books it then reports back to user.
The most impressive part is that the person likely wouldn’t even notice that he/she is talking to a robot because of its accent, voice with words like “um”, “Mm-hmm” and it even pause for a moment, breaks like a real human.
How Google duplex exactly work?
Basic working of the duplex : User asks google assistant to book an appointment, google duplex makes a call behalf of the user, speaks to person on call and books the appointment, then reports back to the user.
Google has worked more on the natural sound and conversation from different people, their understanding, interactions, and timing of speech which was developed into an AI-driven voice assistant.
Several challenges are faced in natural human conversations like background noise, sound quality issues.
During the continuous conversation, the speech might be faster even for a machine to catch up, this is why fast processing is required to be in conversations.
Sentences spoken by a person might be having different meaning in long conversations. This is an example given by aigoogleblog.com regarding this issue.
During a conversation for booking an appointment to a number of people, the context “OK for 4” might be either of these “at 4 PM” or “It’s for 4 people”, these above issues are solved by high precision decision making.
To know more about the technical stuff of it’s working like (RNN, TFX), visit Google AI Blog.
Why Google made it sound like a human, not machine?
Google is more concerned about User experience, usability, wanted to make it as efficient as possible.
If you get a phone call from duplex and it sounded like a mechanized robot you would probably hang up the call, but in another hand, if it speaks as a human, you would be engaged in conversation (there might be a chance you wouldn’t even notice whether it’s a human or machine). But during duplex conversation, it would call and specify that it is google assistant and not a human.
Benefits and limitations of Google Duplex
User – If a business information is not available online or in their website like reservation of seats, booking tickets, etc. What would a consumer do? Obviously, you would call the business; this is where google duplex comes into the picture, you ask the assistant to book an appointment, it automatically calls and enquires then book the appointment and report to you. this helps a lot of people with Language barriers, people who are busy and want to maintain connectivity and manage time.
Business – Most of the business has a booking system /appointment management peoples. These businesses can rely on google duplex to take care of appointments with high precision & in numerous bookings. it can also be used for reminding the customer about the booking gradually through calls rather than messaging or mail, this would save a lot of time.
Currently, it is in the beta state or developing stage, therefore there are few limitations like calls made by the assistant is not transcripted or recorded (verified by user).
Google’s explanation towards cancellation of appointment: If a user books many appointments at several places/restaurants and then decides to cancel it using duplex. It would block the use of reservation facilities for a specific amount of time (this is to prevent spam calls).
Concerns about Google Duplex
Many people feel creepy about this idea of a machine to speak like a human and lead a conversation on behalf of a real human. Some are even concerned about security and privacy issues.
Due to this release in advanced technology, there are many questions arising like Is it secure in both ends, user, and business call? Companies can use it as a spamming call tool to make numerous calls which count annoy users as we were fed up of customer care calls, many times a day. These are some potential impacts through google duplex.
This shows the rapid Evolution in Artificial intelligence and Machine learning.
It is surprising to believe that AI arrived a few years ago and now it has come to such an extent that it can speak like a human and make a call to a live person.
Check out this article
What would happen if AI surpasses human intelligence?
What are your opinions on Google duplex?
Should a machine have access to calls and speak behalf of a human?
I am excited about this technology but worried about future, What happens if AI becomes so intelligent that it makes call/ deals with the conversation without our permission.
I have no issue with this evolution of AI, but it should always have the highest priority towards user safety, privacy, and permission management. Overall it’s a great Innovation.😎
Devices supporting Duplex
Google duplex is supported in android version 5.0 and above having google assistant, for now, it’s rolling out in pixel devices and in some cities. Check out more details from here.