Talking to Buildings with conversation AI

When I did a search looking for articles and research on talking to buildings it was not the conversation AI, data science, social discord I found. Instead it was more about the abstract language of how a building talks to us through its emotional impact. The atmosphere of a cathedral or sports stadium the home space of colour and light. It was like reading Grand designs with Kevin McCloud. Please don’t get me wrong I love the series, but my search for kindred spirits left me wondering am I alone.

For those not familiar with conversational AI it is the ability to communicate with a machine using voice or text messaging as if it was a person providing a service. Amazon Alexa, Google Assistant, Apple Siri and Microsoft Cortana are all example implementations of conversational AI. My favorite AI has to be Replika a social AI that is designed for well being. Many companies are now using BOT’s as chat assistants to process enquiries, take orders and help choose holidays. There are even some medical diagnostics applications being trailed in health care settings.

These AI’s are not just a chat, the conversation or dialog is augmented with media, photos, images, maps diagrams links to videos and a plethora of other services.

The idea of the “intelligent machine” goes back as far as ancient Greece, it has been the topic of many Sci-fi stories my favorites Isaac Asimov’s Robots to Arthur C Clarkes Hal 9000. In the modern era we are finally at the beginning of a new age which I hope will be a golden one.

Why would you want to talk to a building?

We can see that the Internet of things (IoT) has spawned new affordable services and devices, classed as smart things, plugs, lights, sensors, thermostats all controllable through smart speakers all out of the box. This is a big growing market which demonstrates there is demand that people want to control the environment and query it. So put away any thoughts that this is not a good idea the facts say otherwise.

Conversational AI is a branch of data science and uses machine learning. Machine learning is a process where by you take a data set of what is termed “training data” and process it through a model learning process. The out put is a taught model which acts like a function. You give it a piece of new data and it returns an analysis. This analysis, in conversational models, consists of scoring the Intent and extracting any entities. For example if I say:-

What is the temperature in the living room?

The model would return a score for the intent, which I would expect to be the “TemperatureIntent” and extract and room entity called “living room”. With this you only need a method for returning the latest temperature for the living room.

The temperature in the living room is 21 degrees taken from the button device 30 minutes ago.

this is how my AI returns the query

This question and answer is typical of the out of the box or linking the smart assistant to the smart device but what is if you say:-

What was the temperature in the living room last week?

The intent is now an historical record of temperature “TemperatureHistoryIntent” and this is because we mentioned some reference to dates. “Last week” the room entity is still “living room” . The method for this should be work out the day date last week, 7 days ago, find the maximum and minimum temperature and return both values and say

The temperatures in the living room last week, on Wednesday, was a minimum of 16 degrees and 21 degrees.

It is possible to alternate a number of responses just as we would articulate makes it less like a robot, the humanizing element.

In order to perform this query you need to be logging the data and this is where out of the box stops and software programming begins.

If your going to log the data then what about all the other information you can record and call up.

When is the warranty up on the dishwasher?

How do I was use the washing machine?

Can I put this bake bean tin in the green bin?

When is the boiler due for a service?

How many times did we use the dryer last week?

I can perform these types of queries because I am storing detailed asset data, thanks to the COBie schema this is pretty a much easier task.

Asking questions is of what is already there is only part of where this is all going. When you use a bot to order a takeaway or choice a holiday they use a pattern of fulfillment and method of handling the dialog in a waterfall. In buildings we might want to say:-

I bought a new TV lets record it as an asset.

Can you look for a local gardener.

Keep an eye of the energy usage and let me know if we are using more than last year.

These are where you start to off load queries to other services, which support queries in their api’s. I do this for the weather forecast.

We will quickly move to a stage where like Netflix or Amazon prime we get suggestions.

Hi based on the pattern of use bedroom three is not being used would you like me to lower the temperature in that room?

The weather over the last week has promoted grass growth did you cut the lawn this week?

It’s been windy this week and it maybe a good time to clear up the leaves.

It’s going to be windy tomorrow and I noticed the parasol in the garden is up.

This functionality is using notification just like you see on some of the social media apps.

You might be wondering do we need all these things. Peoples lead busy lives and with an aging population we will all need that none judgmental assistants and encouragement.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s