Google broke my house because someone said a bad word
Yes that’s right, Google has broken my Internet of Things house because someone swore. But, to be clear, it’s not because someone in my house swore, it’s because someone somewhere in the World swore.
So how did this come about? Well I use a nifty service called Stringify to connect a bunch of my devices into one unified program (it’s like the better-known If This Then That but works more flexibly and reliably). So, for example, when I tell Google Home that we’re watching TV, my main lights go off, the TV back-light turns on, and another screen that reflects on the TV turns off. And Google tells me it’s done all that by saying ‘Hope you enjoy the show’.
Now that response is set by me in the Stringify program; and it turns out that somewhere in the World someone who is using Stringify created a response that included the word ‘fuck’. This resulted in Google turning off the Stringify app’s access to Google Home. This is how the developers reported their interaction with Google:
Google has replied and told me that they have disabled our skill because “the system identified an abuse evidence for the word ‘(the f-word)'”
So because someone used the ‘f-word’ in the programmed response that Google Home gives, everyone using Stringify has a broken home. Imagine if they shut down all Android phones because someone swore on one…
Now there are several lessons to be drawn from all this. The obvious one is that having an house that relies on external services has an intrinsic vulnerability. Not having the lights automatically go down when we watch TV is not a big deal. Not having the burglar alarm turn on when we leave home, certainly could be. And I would argue that Google has a duty to think through and start recognising how people are using Home, and taking some broader responsibility. If you advertise the service based on doing things like controlling your home, and they do, you have to focus your efforts on that service being reliable – not on primary-school-level safe language.
But the second point is more general. What on Earth is Google doing acting as censor of what people say, or have the technology say, in their own home? This is the same Google that is struggling to find a way to censor hate-speech and videos but apparently is succeeding in forcing Google Home to not say ‘fuck’.
Stringify has implemented a ‘fix’ in that it will now filter the allowed responses to fit with Google’s rules – which will deal with the immediate problem if Google get on with re-authorising Stringify. But that doesn’t address the underlying issues: If we’re going to have Google Home as an integral part of the Internet of Things and Google needs to wake up to it’s responsibilities.