This is my general complaint about all of these voice controlled services. I want to be able to set my trigger word. I will not be unpaid brand promotion for Google, Amazon, etc. I can mostly deal with "Alexa" or "Siri" since those are actual names, but I'm not going to go around saying "Ok, Google".

Humans name things. Pets, cars, computers, houses, kids(!). We need to be able to name our AI pets, too.

"Ok Google" is one of the most annoying trigger phrases. I literally have a hard time saying it and have to deliberately slow it down to get it to realize I'm trying to trigger it[1]. This has led to me pretty much abandoning that service..

It's the difference between saying "ok goo-gul" and "ok googl". For whatever reason, the second "g" sound gets dropped or blurred when I say 'google' aloud, so I have to remember to slow down and say it the first way.

"Alexa" and "Siri" are simple, and have no glut of soft sounds stuck in the middle of the word.

Hah, I have the opposite problem, it triggers too easily for me. My roommate's name is Hugo, and we both have Android phones and a Google Home. If I say "Hey, Hugo" you can hear all 3 devices turn on.

I know the "OK, Google" detection is supposed to be linked only to your voice, but I find about 30% of the time other people can trigger it anyways.

That's hilarious. And a perfect example of why trigger words need to be broadly customizable.

Yep, my father tried to use "OK Google" on his new 5X and ended up triggering my phone instead.

Agreed - I listen to podcasts with gmaps running on my phone, and about once a week some random non-google phrase will trigger it.

I will never say OK Google (except when saying I will never say OK Google). Gross

How would you prevent this from being a back door into a person's house? Or, more directly, how do you prevent applications from stealing phrases from each other?

That is, you are basically saying that you want everything in $PATH. This would be like if git had decided that "log" should just do "git log". Certainly could make sense. And I agree that users should be able to allow this.

However, the applications? I'm not as sold. You are basically allowing a situation where the fundamental behavior of the system would change from installing a single skill. And it might not be clear on how or why it changed. (Certainly not to most users.)

I think it would work like the “default browser” or “default mail client” concept. While it’s possible to install conflicting apps, it’s certainly likely that I will only have one logical handler for a certain type of general request and likely that the app I most recently installed is the right choice when there’s overlap. (The Alexa phone app could be used to change that, when the default assumptions are wrong.)

Also, voice commands take significantly longer than typing and there is no real auto-complete. The cost of a wordy voice command is much higher, especially if you stumble at any point and have to say it all again (usually while trying to talk over one of Alexa’s wordy error responses).

It just doesn’t make any sense for commands to sound like marketing material. This is the “Windows 95 Start Menu” thinking where every app is under a “FooBar, Inc.” submenu instead of just getting to the point and showing you the app you want.

There is a lot of "best intentions" here. "Default" applications almost work, but really only exist because we have well agreed upon url schemes for specifying a few things. Mainly "mailto:" and "https:". The rest falls into the hell that is default application for file type. Which is the largest source of malware and other nonsense in consumer computers.

Seriously, let that settle for a minute. "Default" behavior for executable applications is the most preyed upon phishing vulnerability there is. I do not want that introduced to a home automation device. I hope we do better than that.

Perhaps they could let users set a skill alias.

I fully agree that should be possible. Apologies if I didn't make that clear.

This and custom voice training for device names are on the top of my wishlist for the Echo.

For some reason my Echo has a really hard time recognizing some of the rather obvious names I give to my devices, like "LG TV". It'd be great if we could train the voice recognition engine to associate certain pronunciations with a specific device name.

The Alexa app does have a generic voice training selection...

One exemple of a possible answer:

"alexa, add a task saying blablabla" "Ok Bob, the task has beeen created in your favorite todo app. Would you like me to transfer it to another app?"

"Oh, jeez, Alexa, just do it already without asking me stupid follow-on questions!"

One of the reasons I'm less sanguine about voice as an uber-interface. I'm not sure you can square the circle between a rich and capable interface and one that isn't interrogating you like that. Computer screens can pop up arbitrarily large amounts of data on the screen (entire EULAs) to be dismissed almost for free. The serial nature of speech is going to be a big challenge for most people. (There are, of course, those who use it all the time as their interface. Perhaps a few will even read this post. But I'd submit their usage doesn't end up looking or sounding much like the Star Trek ideal.)

I am not completely familiar with Alexa (I use Google Home) but can't you make a recipe with IFTTT to accomplish this? Though it still does not rely directly on the ALexa OS.