The not-so-smart IoT house conversion…

Becca (my long-suffering wife): ‘So are we actually going to finally sort out the back of the house?’

Me: ‘Oh..ok it’s probably time to do it…’

Becca: ‘The kitchen’s starting to fall apart, the windows need replacing, and we hardly ever use the dining room’

Me: ‘Guess this is the year…If we’re going to tear it all apart, could we…ermm…rebuild it as a ‘smart house’. You know, all connected together and to the internet, with loads of labour saving devices?’

Or so the conversation roughly went between Becca and I at the start of last year (and clearly I didn’t get it that easy…). Our old seventies house really needed some work and in my CTO head this could be a great excuse to buy some new technology to make all our lives a little ‘easier’ into the mix. Back at work, we’d started to take a really serious look at the impact the ‘Internet of Things’ (IoT) was going to have on the domain name industry and rather than just theorise about the impact of all these connected devices, this was a chance to get some hands on experience with the very technology we were investigating. So we upped our budget a little and decided to build in a collection of internet connected technologies to see how it all worked.

I didn’t want to get into the soldering-iron wielding, raspberry-pi fettling brigade so all the technology had to be off the shelf and commercially available. My only concession to being a not-so-secret nerd was that we might use a few tricks to ensure it all talked together. Over the next 9 months, we added smart lighting courtesy of Philips, two smart heating systems from Nest, Nest cameras and fire sensors, a Roomba internet-connected robot vacuum cleaner and finally placed Amazon’s digital assistant ‘Alexa’ at the heart of proceedings so that we could ‘talk to the house’. Using a bit of custom coding, we connected Alexa to Becca’s E-Commerce business so that we could talk to that too.

img_0763.jpg
Much to our surprise, it all kinda works. Nest seems to keep the house at about the right temperature (and sure looks pretty). The Philips ‘Hue’ light bulbs change colour and dim on command. Alexa’s reasonably happy to answer questions and talk to the various systems and the robot vacuum cleaner scurries out when we’re away and starts to clean the downstairs. In fact when friends come round, I can do a pretty convincing act that we’re a modern day Jetsons family, and they swoon at the clever technology seemingly changing our lives.

But actually it’s not quite as rosy as my act might make out and here’s where we hit the rub with much of this new ‘smart’ technology. Sure it’s pretty clever, but it’s not by any definition ‘smart’. Take the lights – I can say:  ‘Alexa, dim the dining room lights’ and this triggers the Alexa smart home application to dim the lights. 95% of the time it works great. The other 5%, well sometimes it just leaves a random bulb in the middle of the room shining away at full brightness. If we look at the Nest camera system, I get an alert every day, mid-morning when the sun passes over our skylights and creates a pool of light on the dining room floor: ‘Alert – Nest thinks it has spotted movement in you house’. And finally, our robot vacuum cleaner, which I quote: ‘Uses a high-efficiency cleaning pattern and a full suite of sensors to map and adapt to real world clutter and furniture for thorough coverage‘ gets itself stuck in the same two places in our house pretty much every time it goes out.


The real frustration with these fledgling systems and devices is that they are so close to being brilliant and yet by not being 100% reliable, we start to lose faith and trust that they are doing their job properly.

Trying to understand, just how these systems could be made better has led me to spend time looking at a subject that I’m increasingly believing will play a significant role in turning the IoT into something really useful: that is the ability for systems and devices to adapt to their environment and context through a process of learning. At the moment our devices appear clever because they’ve been built to anticipate common scenarios: Ask Alexa what the 4th largest city in France might be, and she’ll come up with a very erudite and factually correct answer. Ask her a broader question such as ‘what stock should I buy?’ and she doesn’t even know where to start. That’s fair enough – she’s yet to understand my circumstances fully, but she makes no attempt to follow up with questions that might help her create more targeted answers. The system is in no way personalised to me and has no learning capability.

When my vacuum cleaner gets stuck the 2nd or 3rd time on the same obstacle, why doesn’t it simply learn that it should take a different path next time. The Nest camera should be asking me whether it’s got the alert right – and constantly getting smarter as it learns the changes within the house.

So my connected-house journey is very much still underway. The technology is there to really change the way we interact with our houses and it does a pretty good job most of the time. But what I’ve learned is that I now need to get a deeper understanding of just how we make machines ‘learn’ and that’s going to take a bit more digging and playing around.

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

Create a website or blog at WordPress.com

Up ↑

%d bloggers like this: