by Jim Lundy
At Google I/O, Google’s annual developer conference, an incredible thing happened. During Google’s main keynote, I watched a Google Assistant demo that will change the world and put pressure on nearly every other provider to step up their game in transitioning from voice search and chatbots to full digital assistants.
This blog overviews the demo and why Google Assistant just changed the world.
Google Assistant Can Talk to Humans and Schedule Appointments
The demo that Sundar showed presented two examples. Both involved having the Google Assistant—named Duplex—call a place of business, talk to a human, and schedule an appointment. The first demo had Google Assistant call a salon and schedule a haircut.
The Google Assistant voice sounds human and given the upgrades to the platform, was able to have a conversation, negotiate a time that worked for the “client,” and confirm the desired service (a haircut). That was impressive, but it gets better.
The second demo involved calling a Chinese restaurant to request a reservation for several people. Along with understanding the accent, Google Assistant was able to determine from the conversation that a reservation for that number of people was not available, but that they could walk in that day. Google Assistant accepted that, thanked the human, and ended the call. All in all, it was very impressive.
Google Duplex Powers the Assistant
Google Duplex is the name for the technology that enabled this natural conversation. Duplex is not for the feint of heart. It is using some of Google’s best technology in Artificial Intelligence including:
- Recurrent Neural Network (RNN)
- TensorFlow Extended: a deep learning algorithm trained via a ton of phone calls
- Google Automatic Speech Recognition (ASR)
- Speech-to-text engines (Tactotron and Wavenet)
Mimicking a human requires a lot of technology to work together, and Google accomplished this. There is more to come here but Google just upped the game and advanced the ball down the field. It is on the cusp of making Google Assistant fully actionable for complex tasks, such as scheduling, and it executed this well.
Pressure on Amazon, IBM, and Microsoft
While all of the Assistants from Amazon, IBM, and Microsoft can talk, none have shown the ability to navigate a conversation to reach an outcome. They are all mainly first-generation, and address single tasks through human direction.
Yesterday, at Google I/O, we saw a glimpse into the future of real, interactive digital assistants. There will be naysayers about the negative aspects of digital assistants, but for people and enterprises looking for faster outcomes, this demo showed how things will evolve. There is a new race and Google is leading it.