If you own an Alexa you've probably been frustrated with it. Maybe you've even been irrationally mad at it. Misheard commands are just one of those nuisances of modern life. But an Alexa can't cause a full blown professional catastrophe, right?
Right?
Perhaps at home, but I had just hooked up our entire video studio to run on Alexa commands. The lights were on, our guest was presenting, and several hundred customers were watching over the live stream. That's when our Alexa, the brain of the room itself, went rogue.
Interrupting live stream as the producer is bad enough, but what I had done was much worse. I had architected the whole room to be a house of cards, ready to tumble at the slightest breath.
That's because Alexa really did run the room. The lights, the timer, the cameras, the switcher, the microphones, even the encoder running the live stream itself could be controlled to some degree with voice commands only. I had worked diligently with engineers from other teams to figure out how to attain the maximum control over every element of the room and offload it to my digital companion.
It felt like magic when we tested it. But now I was watching Alexa interrupt a show that was already running to deliver what may have been the longest description of the Seattle weather that I had ever heard. Evidently something the presenter had said triggered the wake word (still set to "Alexa") and started it off on a surprise meteorology segment of a show that had previously been about smart home devices.
The presenter was a good sport. After the initial surprise he laughed it off and told Alexa to stop mid weather report but I had more explaining to do once it was over. But I still knew right away that this was a good failure, and the best of what could have happened.
I took away two lessons from that day - the first being to always mute smart devices on set of course - and the second was to totally change how I was thinking of studio automation. Alexa was nice, but this failure was probably the least embarrassing compared to what could have happened. What started as an experiment using devices and help we had in-house became a question that I'm still trying to answer: what else can we automate on video sets?
It turns out, a lot more than you would think. Read this next segment for a breakdown of how the room worked (and didn't).
Comments