Look briefly at the following transcript of a recorded telephone conversation between someone wanting to book a table at a restaurant and the restaurant.
Restaurant: Good evening.
Caller: Hello?
Restaurant: Hello.
Caller: Hi, um, I’d like to reserve a table for Friday the third.
Restaurant: OK, hold on one moment.
Caller: Mm hmm.
Restaurant: OK… hold on one second.
Caller: Mm hmm.
Restaurant: So Friday November third. How many people?
Caller: For… two people.
Restaurant: Two people?
Caller: Yeah.
No big deal. Someone calls a restaurant and books a table.
The minor twist in this conversation is that one of the two involved in this conversation is not a person – but a computer assistant. Go back and read it again and try and guess which one is the computer and which one is the human. I won’t give away the answer just yet.
First we had human actors recording hundreds of hours of audio to give programmers a range of words that could be used to generate human-like speech. But when we heard our Satellite Navigation voice telling us to ‘make a U-turn as soon as possible’ it was obvious that it was computer generated. When our smart home devices told us a joke or the weather forecast, again we were under no illusions as to where the voice was being generated from. Alexa added a newscaster voice that could be used when you asked Alexa for the latest news. We still weren’t fooled. We knew it was a nifty technical trick.
When we receive random telemarketing phone calls coming from overseas, some of the people on those calls read from their scripts so well that they sound robotic – but we know there is a very-low paid worker sitting in poor conditions in a highly-populated nation somewhere making those calls and having those conversations.
But, if you are working at a restaurant or a hairdresser and you receive a phone call for a booking, you safely expect it to be from a human. Likewise, if you make a call to your favourite eatery, you would think that a person answers the call.
Not necessarily.
One of the first issues to be addressed is an ethics issue. The example at the beginning of this column is from Google Assistant. A small number of people across the world currently have this feature enabled and it is in use right now. Many people were disappointed with the fact that it seemed disingenuous that there was no acknowledgement that you were speaking with a computer. That has now changed. Future conversations will now be identified as such. So don’t be surprised if you make a call or receive a call and the ‘person’ identifies itself as a bot – and you then continue to have a normal conversation.
As for which one is the bot – the ums and hmms seem to give it away. These sound more like a human than a computer…but therein lies the genius of the natural language. These ticks have been inserted to give the impression that it is a human. In this conversation, the caller was the computer.
Which brings me to my next point. Did I really write this column…
With Christmas so close, I am going to add a ‘gift of the week’ the end of each column through to the big day. This week, my Christmas idea is called CrashSafe. It is a 6 in 1 device that is designed to help you in your car. It is a cigarette lighter adapter that you plug into your 12V or 24V outlet in your vehicle that gives you a USB port. It has a battery contained within so if your car has no power you still have enough power to charge your phone or run some of the lights on the device. It has a bright LED light that will run for 6 hours in addition to a warning red beacon that will run for 10 hours. Perfect if you are broken down on the side of the road. It also has a sharp hardened pin for breaking a window and a safety knife that can be used to cut a seat belt – both handy if you are in an accident and stuck in your car.
Mathew Dickerson