Amazon Echo to HA Controllers

[wkearney99]

You can just define what you want. 

 
Yes, the same can be said for a C compiler.  However, that doesn't mean the larger market wants to have to do that themselves.
 
Crafting effective NLP-like intents is not as trivial as most folks think.  Otherwise everyone would have done it already.

 

[wkearney99]

I'd say Echo's appeal and acceptance indicates it's "good enough" for most people.

 
Agreed.  There's plenty of room for improvement.  Meanwhile they've gotten their foot-in-the-door with some pretty decent hardware and an infrastructure to support it.  Both of which are a lot more than anyone else has done, to date.

 

[pete_c]

I know this doesn't have any direct relation to the OP but....
 
Watched the Google I/O conference (venue?) today. 
 
The google home presentation / home use advertisement looked very much like the original Amazon Echo advertisement.   
 
Except for ....
 
[youtube]http://youtu.be/UHJ0uOPs4s0[/youtube]
 
They did mention stuff that was fixed from their competition without stating any names.

 

[Dean Roddey]

Yes, the same can be said for a C compiler.  However, that doesn't mean the larger market wants to have to do that themselves.
 
Crafting effective NLP-like intents is not as trivial as most folks think.  Otherwise everyone would have done it already.

 
I thought I made it clear that it would be optional. They just need to *allow* it, not require it. As it stands, with our Echo support, it's not terribly hard to do any of the sorts of things I gave examples of above. They all ultimately boil down to one of five or six patterns that our Echo support handles, and all you have to do is provide the CQC action to respond to those patterns, and they get passed along the important info from the spoken phrase to use.
 
That kind of flexibility is crucial to using the Echo in serious automation. If they don't care about serious automation, then that's their choice of course. But I at least wanted to make it clear that providing support for serious automation does not some huge change on their part, and wouldn't affect the otherwise normal operation of the Echo for anyone else.

 

[potts.mike]

But the thing is, a voice control system shouldn't have any fixed set of commands. It's really just a remote control, and it should be able to send any command or make any query. If you have an automation system. The automation system can be configured to respond to anything you want, in any syntax you want. 

 
Dean, this is not really true in the case of the echo. While it is a remote control for CQC thats just one of many things it can do. I can order an uber, a pizza, ask where my packages are check my calendar for the day, play music from services CQC doesn't support etc... There is no way that any automation system can have "drivers" for all these services, they just aren't mainstream enough. But by providing access to the connected home API you can do a lot (off/on, set, dim etc...). Just because it is designed for a switch, per say, doesn't mean it cant be translated otherwise (turn on espn). I use the smarthings integration everyday because its easy, even easier for my family.

 

[ChrisCicc]

There is no way that any automation system can have "drivers" for all these services, they just aren't mainstream enough.

 
This is exactly why we're opening up the API. We'll try to provide all the core automation-related voice grammars, but anyone can build a plugin to add their own. 

 

[DeLicious]

Then you can do anything you want, such as:
 
Alexa, is today's high temp greater than X
Alexa, how long till sunset
Alexa, what is the ground saturation percentage
Alexa, how many cars are in the garage
Alexa, what are the currently violated zones
Alexa, set the living room to party mode
Alexa,  Bob is currently in vacation mode
Alexa, which rooms have active motion
Alexa, how many inches of rain fell in the last x days
Alexa, have music follow first floor rooms if active
Alexa, have the sprinklers run today
Alexa, set all bedrooms to music playlist x

 
But how do you differentiate these from Alexa's normal set of functionality?  How would it know a query about the high temp is meant to go to CQC instead of its internal weather app?  How would it know to ask CQC about the sunset instead of its built-in knowledge? How would it know to send music requests to CQC instead of playing natively on the Echo?
 
I have to imagine that all of their internal apps take precedence over skills, and thus, if you want to send something specifically to a SmartHome skill, there has to be some indicator to do as such. I highly doubt they'd be willing to turn over their entire voice experience to third parties.
 

I'm curious to see how flexible the Echo's upcoming local API will be. 

 
I haven't heard anything about a local API.  Do you have any sources for this?

 

[Dean Roddey]

Dean, this is not really true in the case of the echo. While it is a remote control for CQC thats just one of many things it can do. I can order an uber, a pizza, ask where my packages are check my calendar for the day, play music from services CQC doesn't support etc... There is no way that any automation system can have "drivers" for all these services, they just aren't mainstream enough. But by providing access to the connected home API you can do a lot (off/on, set, dim etc...). Just because it is designed for a switch, per say, doesn't mean it cant be translated otherwise (turn on espn). I use the smarthings integration everyday because its easy, even easier for my family.

 
That would be easily enough remedied by allowing one skill to invoke another, or to allow text to be fed to Alexa instead of speech, which would be trivial for them to support since it just involves skipping one (big) step. Then effectively something like CQC could be a 'driver' for any of other other things while still providing a single point of entry for voice control and all of the extra flexibility. Basically we could just pass through commands we don't understand.

 

[Dean Roddey]

But how do you differentiate these from Alexa's normal set of functionality?  How would it know a query about the high temp is meant to go to CQC instead of its internal weather app?  How would it know to ask CQC about the sunset instead of its built-in knowledge? How would it know to send music requests to CQC instead of playing natively on the Echo?

 
That's really the point of all I'm this stuff I've been saying. Right now, you can't. If they would just allow the user to 'lock' a specific skill in as the 'native' skill, then for folks who want the flexibilty of using the Echo as an open ended input mechanism into an automation system could do so without having to use the 'second class citizen' syntax.

 

[ChrisCicc]

I haven't heard anything about a local API.  Do you have any sources for this?

 
Me?

 

[DeLicious]

That would be easily enough remedied by allowing one skill to invoke another, or to allow text to be fed to Alexa instead of speech, which would be trivial for them to support since it just involves skipping one (big) step. Then effectively something like CQC could be a 'driver' for any of other other things while still providing a single point of entry for voice control and all of the extra flexibility. Basically we could just pass through commands we don't understand.

 
I think we probably have different definitions of "trivial".  This is Amazon's baby right now.  Don't you think they have dozens of managers thinking about the experience they want the customer to have with the device? I doubt any one of them is thinking, "you know, we spent years putting this voice experience together, handling the natural language usage appropriately, directing everything to the right place to get answered, so you know what we should do next? we should throw that all to the side and let another program do it".  How could they trust third parties to pass back the commands they don't understand unaltered?  How could they trust third parties to pass back the commands they don't understand at all?  Doesn't that open a giant gaping security hole when someone says "buy a dildo from amazon" and that gets passed to CQC?    Then what happens when another 3rd party abuses it and doesn't pass anything back and tries to handle everything itself?  Then Amazon has completely given up the entire Alexa experience, and I simply don't see that happening.

 

[DeLicious]

Me?

 
yes, sorry that wasn't clear. multiquote stuff...

 

[ChrisCicc]

yes, sorry that wasn't clear. multiquote stuff...

 
The multi quote worked correctly I guess I'm your source for now...

 

[Dean Roddey]

I think we probably have different definitions of "trivial".  This is Amazon's baby right now.  Don't you think they have dozens of managers thinking about the experience they want the customer to have with the device? I doubt any one of them is thinking, "you know, we spent years putting this voice experience together, handling the natural language usage appropriately, directing everything to the right place to get answered, so you know what we should do next? we should throw that all to the side and let another program do it".  How could they trust third parties to pass back the commands they don't understand unaltered?  How could they trust third parties to pass back the commands they don't understand at all?  Doesn't that open a giant gaping security hole when someone says "buy a dildo from amazon" and that gets passed to CQC?    Then what happens when another 3rd party abuses it and doesn't pass anything back and tries to handle everything itself?  Then Amazon has completely given up the entire Alexa experience, and I simply don't see that happening.

 
You are way over-thinking this. It's actually extremely simple. Think about the current sequence of events:
 
Echo sends audio -> server processes audio spits out text -> Text parser parse text and decides if it matches a valid phrase
 
The only thing that would change is the middle process gets left out. That would have nothing to do with security at all. They always would end up with text to process, however it got there. It would change nothing other than how the text got generated. Everything that has to do with what gets done would be driven by the spit out text, so all of the checking and validation of the text would already happen either way. So it wouldn't be a bit different from how it is now, in that you can say anything you want to the Echo. Do you consider that a security issue?
 
And, again, no one HAS to use it. It would just be an option that would allow for a lot more flexibility, and the customer wouldn't know or care that it was going on. 
 
I just don't get how, here and on Reddit, if someone makes a suggestion to improve the Echo for automation purposes, that so many people somehow act like this is an affront to morality, or that they are incapable or error or have already thought about everything that could be thought of and and if they haven't done it it's because it's bad. 

 

[Dean Roddey]

Is there any significant advantage, if you are primarily interested in voice control, to the Kinect FW v2 vs. the original v1? And how much juice does the host need if that's really all you are using it for?