­Google I/O 2018 has finished and just like every year, something was announced that has everyone in the industry very excited. This time it was to do with their AI voice assistant. This new experiment of theirs is named Google Duplex.

The demonstration focused on explaining how Google Duplex aims to speed up every day efficiency for its users by taking over the task of calling establishments to make appointments, reservations or ask questions. Many technology companies are embracing this new technology and beginning to bring it into their own applications such as Cisco and their AI voice assistant. It’s easy to see the benefits these kind of developments will bring to our lives.

Using Google Duplex you can simply say “book me a women’s hair appointment between 10 and 12am” and in the background, the AI will call the salon, check availability, and make an appointment if possible. This is exactly what was showed in the demonstration.

At another point in the presentation, Sundar Pichai, Google’s CEO, explained that not only would Google Duplex obviously help busy people organise their day to day life, but it would also be hugely helpful for small businesses. In the holiday seasons, small businesses can sometimes be inundated with calls asking the same questions over and over. By using duplex, these questions could only be asked once and the knowledge learned could be shared with every other google device in the country, saving time and countless unnecessary conversations for small businesses. He remarked that it was essentially a win-win.

In the second demo Pichai showed, the AI was attempting to book a dinner reservation. When the restaurant employee answered the phone it was clear she had a strong accent and spoke in a non-traditional way. However the AI did not faulter and still managed to comprehend what was being said. The method in which it adapted and overcame challenging hurdles are signs that Voice AI is improving at a significant rate.

The crowds reaction to what they were seeing was extremely animated to say the least, but there are many who are not praising this demonstration as a technological feat but instead viewing it a negative light. A recent article from The Verge discusses how its easy to become enamoured with the largescale spectacles that tech companies put on, but often the organisations don’t follow up on what they promised.

This happened last year when Pichai announced a clever tool that could detect unwanted objects from photos and remove them using Google’s powerful machine learning systems. However this tool never made it to market and there hasn’t been any more talk about it ever since. Therefore it may not be a bad idea to also remain sceptical in regards to Google Duplex, as it is a huge task Google have set themselves.

Google have confirmed that on most testing occasions, Duplex successfully completes the task without any intervention from a Google employee. However, in some cases, the AI can become confused and requires human assistance.

This is a statement from Google regarding Duplex:

The Google Duplex system is capable of carrying out sophisticated conversations and it completes the majority of its tasks fully autonomously, without human involvement. The system has a self-monitoring capability, which allows it to recognize the tasks it cannot complete autonomously (e.g., scheduling an unusually complex appointment). In these cases, it signals to a human operator, who can complete the task. To train the system in a new domain, we use real-time supervised training. This is comparable to the training practices of many disciplines, where an instructor supervises a student as they are doing their job, providing guidance as needed, and making sure that the task is performed at the instructor’s level of quality.

In the Duplex system, experienced operators act as the instructors. By monitoring the system as it makes phone calls in a new domain, they can affect the behavior of the system in real time as needed. This continues until the system performs at the desired quality level, at which point the supervision stops and the system can make calls autonomously.