Getting things done by talking to things
How next generation IoT devices will use natural language as the intelligent interface.
Lately I have been telling anyone who will listen about an exciting, and possibly scary, new development in Artificial Intelligence (AI). A concept called Dynamic Program Generation!
The most well known AI systems today are included in phones and computers (think Siri, Cortana) and IoT products (like Amazon Eco’s Alexa and Google Home). These systems rely on word recognition of pre-defined words or phrases that need to be hard coded by developers. When I ask Siri to tell me a joke, she can because both the command and response has been pre programmed. But if I don’t ask in the expected way our AI products can be frustrating…asking Alexa to turn off the hall light is a challenge for my wife Michelle with Alexa responding “I found several devices matching that name, which one do you want?” instead of understanding the simple instruction.
Despite this I love the potential of AI and natural language recognition in product interfaces. Not needing to touch a screen or a button and products just working naturally is one of the great design promises of the connected future.
So why am I excited and what is Dynamic Program Generation?
Dynamic Program Generation is a term I heard in a Tech Crunch Disrupt presentation by Siri co-founder Dag Kittlaus and it refers to a next generation AI product called ‘Viv’. In contrast to current generation AI systems, Viv writes its own code on the fly.
Using extremely sophisticated natural language understanding Viv takes an ‘intent’ and generates a program with software that writes itself in 10 milliseconds! A weather example used in the demo “will it be warmer than 70 degrees near the Golden Gate bridge after 5pm on the day after tomorrow?” results in Viv figuring out all the details of the enquiry (location, time, weather), connecting to the services required and writing a 44 step program to figure out what was being asked and then provide an answer. This is quite incredible when compared with Siri or Alexa who struggle with single commands.
Taking this technology further Kittaus then demonstrates how he uses Viv to perform a range of skills:
- Send Adam $20 for the drinks last night — uses Venmo to make the payment simply and easily without fiddling with forms and buttons.
- “Send my mom some flowers for her birthday — results in options from ProFlowers and seamlessly puts the order together. No messing around with online or mobile order forms that always seem to take longer than they should.
The demonstrations illustrate how Viv’s combination of features including ‘conversational intent’ (understanding what you say with location, time, task and dialogue context) and a better understanding of you (through learning about your friends, location, age and what you like) makes it possible and easy to get things done by talking to things!
Viv’s universe of capabilities that developers are adding to over time already seems huge and the combination of capabilities and machine learning through every request implies that it will grow much faster than AI like Alexa which this week boasted over 5,000 skills.
What this means for the design of The Internet of Things
From paying friends, to flowers, to hotel bookings and weather commands, Viv is already making it possible and easy to get things done by talking to tablets and phones. For other devices Viv already has a marketplace for all the different kinds of devices that the Internet of Things will enable. For example, home monitoring devices targeted to the non-tech aged care market will allow grandma to talk to her device in her language, not through a screen, and imagine how much easier controlling you television could be if it responded more naturally than Siri.
Voice services available today like Siri and Alexa are pretty cool but with Viv the intelligent interface to everything is closer than we think.
To find out what natural language as the intelligent interface means for your IoT project contact Creator.