Hacker News Reader

Comment by JoeCortopassi | Hacker News Reader

Alexa has a fundamental short coming that stops them from being a robust operating system like the writer describes: the cognitive load in memorizing all possible variations of 3rd party apps on their platform. Siri might have greater difficulty with natural language processing than Alexa, but I never have to think much about the word order or phrasing. With Alexa, I'm constantly feeling like I'm playing MadLibs with my eyes closed.

That being said, the openness of their platform is a huge benefit, and the main reason I have one. I for one am very interested in seeing which improves first: Siri's natural language processing, or Alexa's contextual matching of their api.

I know Google Now wins in a lot of areas, and there is a group of people that have no problem using them, but overall sentiment seems to be one of distrust in giving them an always on microphone

View on HN

I made my own app and I can't even remember how to use it. I keep saying the wrong trigger word. It's my one gripe with Alexa is that the app name needs to be a prefixed.

What I want to say:

> Alexa, open the garage door

What I actually need to say

> Alexa, tell Some App Name to open the garage door

> It's my one gripe with Alexa is that the app name needs to be a prefixed.

The thing that annoys me is when this doesn't work.

> "Alexa, play <station name as it appears in Alexa App> on Pandora"

> "I couldn't find <station name as it appears in Alexa App>, here's some crap on Amazon Music"

You know you can set the default music provider in the Alexa app, right? Setting this to Spotify allowed me to stop asking Alexa to "play X on spotify" and just say "play X" without getting bounced to Amazon music.

Yes. It doesn't help. Alexa will understand what was spoken perfectly, it shows as much in the App. Sometimes if finds the station, most of the time it doesn't. Regardless of the default music provider or the qualified "on Pandora", Alexa will fail to find the request and default to searching Prime Music.

Submit this to Amazon and they will fix it.

I try and get it to play Kiss FM UK and says a strange word and then dies. Raised it to Amazon and they are looking into it.

The dream:

    > they will fix it

The reality:

    > they are looking into it

The dream:

    > they are looking into it

The reality:

    > it's written in a ticket somewhere

I submit every card in the App, it hasn't helped yet.

Huh. I have a very high degree of confidence that I used to do exactly that, though it's been a few months since I did so. I just tried now, and it doesn't work, just like you said.

I wonder if they changed something, or my memory has tricked me. (The weird thing is not only that I remember doing it that way, but I have no memories of having any troubles with it, and this is certainly what I would do on my first try)

This is also a big gripe of mine. I created an app that can be used as a fun "neutral" third-party when instructing my toddler to do something.

What I want it to say:

> Alexa, tell %NAME% it's time (for a nap|to go to bed|to eat dinner)

What I have to say:

> Alexa, tell Toddler Boss to tell %NAME% it's time to go to bed.

It's not natural sounding at all.

Exactly! Funny you should say that... one of the first apps I tried to make was to just say "Alexa, tell [my wife's or daughter's name] I love her" and I gave up because unless I name the app with my wife's name as the app name it won't work the way I want.

Russ Hanneman?

This is what he did on HBO's Silicon Valley. It was quite..weird.

He's disrupting parenting!

Could you not create an app called %NAME% which then uses Toddler Boss?

Ha! I didn't even think about that. But it's not ideal for vertically scaling the family :)

Names are cheap... cheaper than kids anyway!

How about speaking directly to your child? ;)

I guess...

I trust to publish such an app on their store (just to go through the process...they were giving away sweatshirts). they rejected me on the grounds of targeting kids. I'm guessing you never tried to publish this?

Nah it was just a fun project to introduce myself to the platform. I'm surprised they did that to you though. I wonder if that policy will change in the future considering there's tons of apps that target kids in their app store.

Is that a intrinsic limitation or just an omission with the implementation?

I have a TP-Link Smart Switch and have their Kasa app integrated with my Echo, and all I need to do is name my switches and say things like "Alexa, [device name] on", "Alexa, turn on [device name]". So it's definitely possible for 3rd party apps to turn devices on and off with Alexa without requiring users to mention the app name.

Maybe you just need to change your app to add your garage door as a device to the Echo and use the provided on/off commands rather than register your own custom commands? I can imagine the device on/off paradigm could be limiting for certain use cases, but not for this one.

I hate this, what would "on" be for a garage door?

I have a google home, and through integration with IFTTT I can say "Okay Google open the garage door" and it will work. It's so much smoother than telling the garage to "turn on", or telling the front door to "turn off".

It's the major reason why i've stopped using and actually sold the echo but i'm getting a second home soon.

I love all of this smart home stuff, but the software developer in me cringes at the thought of some message queue connection getting messed up somewhere and my garage door going up and down 3 hours later as I lay in bed...

I've got fixes for that, mainly that it can't open from the Home when we aren't home, and it announces when the garage door opens by literally "announcing" it through the home's speaker, as well as a notification on my phone. There's also some trickery that you can do with IFTTT's activation time variable and checking how long it's been since it was "asked" to do something vs when it got to my system.

It's not for everyone, but i'm really enjoying the setup. Although if i could I would remove IFTTT from the whole equation. Everything else is running off of a server in my house, so IFTTT is the only part that's really not fully in my control.

But as it's nothing but a "layer" on top of the rest as an "ease of activation" kind of thing, i'm not that worried.

My (now sidelined/back burnered) startup's main product was kind of like an advanced IFTTT that runs on your own hardware. The major difficulty was scaling up development and production without capital, and getting a basically Turing-complete system to be easy to use.

While I would love a "self hosted" IFTTT, the biggest issue to me would be getting the integrations.

IFTTT is the ONLY thing on Google home right now that lets you create fully custom voice actions. It's got a lot of things like this.

There are a few similar services, but at the end of the day, if it's within my capability to talk directly to a service, I'm going to just write my own code to talk to it (or my FOSS HA system will more likely support it very quickly).

IFTTT's power to me is completely in its ability to integrate with otherwise "locked down" services.

> "Okay Google open the garage door"

Isn't this quite a security risk? I'm imagining a burglar standing outside with a megaphone and ordering your house to let them in.

Well considering the fact that it's pretty far away from any windows, and my condo is on the 3rd floor, couldn't they just break the glass window? Or shove the door down? Or more likely use a bump key or even easier fuck with the garage door RF "encryption" which IIRC is laughably weak.

But that aside, it's setup so that my home automation system will only listen to the voice actions if there is someone home. But that was more to prevent IFTTT bugs from opening when I'm not home.

If someone wants to go through all that trouble to break into my house, let them. I have insurance and security cameras. I'm not going to build my house like a vault to prevent someone from stealing my TV any more than I'm going to walk-around in a bullet proof vest to prevent someone from shooting me.

Edit: Sorry if this comes across a little shitty sounding. I get asked this just about every single time I bring this up. I have this response pretty loaded...

I think there's special integration for home automation. I have the same switches. Remember when you first set them up and you had to go to a special "Smart Home" section of the Alexa app? There's a special section there just for that stuff.

I'm not sure if that part is open to developers or not. It looks like it is. There's a "Get More Smart Home Skills" button that shows 59 options, so it's probably not just for the big boys.

You could set it up to say "Alexa, tell my garage to open." Not perfect, but a little better.

As someone who recently got an Echo Dot for Christmas, it seems like some apps do have this capability. For example, I paired my Sensi Thermostat, and I only need to say, "Alexa, set thermostat to 70 degrees".

Smart home functions are built-in, and not a "skill".

> With Alexa, I'm constantly feeling like I'm playing MadLibs with my eyes closed.

Well said!

Used Siri since 2011 and use to love it. Then this summer I used Google Now and just recently bought a Google Home. Google wins hands down and makes using Siri frustrating!

For me Google Now and Alexa are the next huge leaps in computing ... like the iPhone was. I wish Google would create an app store much like Apple's where users try an app out and buy it for a buck if they like it.

I have my google now hooked up to turn my TV, lights and thermostat on. It understands my commands 97% of the time vs. Siri's 85%.

Can you activate Google home without saying Google yet? Okay Google! Is about the most awkward thing to have to say as a wake word as I have heard. It makes me want to throw my phone if I have to say it more than once. Hey Google! Isn't much better.

They are never going to win if you have to babble awkward phrases when you use it. Alexa just rolls off the tongue.

Edit: Amusingly, when I said "OK Google!" "Change your wake word" to my phone just now... It replied with: here is information from Amazon.com. (information about how to change Alexa's wake word to Echo or Amazon).

Ok Google. You have to fix this.

You can say "Hey Google" with the Home. It's a little easier to say.

before I started I am using Siri to type that stop it Siri…

Well I agree… I don't like having to say OK Google turn the light on… Or OK Google… Or everything… It does not feel conversational. But there needs to be a command otherwise the AI speaker will be doing thing every time I hear something… Maybe we can change the name that would be good

Eventually this sort of thing will be triggered with gestures. Big at first, but gradually smaller until one can input surreptitiously.

Gestures from way across the room or from a separate room entirely? To me voice beats that.

I'm not sure why it should matter where one is when one performs a gesture. If I'm sitting on the couch I certainly don't want to have to stand up to adjust the lighting. Everything gets smaller until it eventually disappears, and the future versions of Echo etc. will just be integrated into the ceiling or a "smart" shirt or whatever. One won't have to worry about where they are, unless one wants to have a conversation without FBI or the spouse hearing it.

Echo has an always on microphone as well. The story is that audio gets sent to a temporary buffer on a server somewhere to listen for your trigger word, and I trust them, but it would be so easy to start storing that data if they wanted, (or rather, the NSA wanted).

I'm not normally really intense about privacy (relative to other tech people I know), but all these devices creep me out.

No, the processing for the trigger is done locally, no network communication until the trigger is recognized. That's why you can only have one of a couple of trigger words.

The same is true of most modern android phones with Google Now AFAIK. E.g Moto X was one of the first phones with a low power dedicated ASIC for wakeup-word recognition to further make it sustainable on mobile devices.

Yeah I read somewhere that it was all sent off, but everything else I've looked at says it's all local.

The data is not sent to a server all the time - only when someone is using the device.

Source: tcpdump

Do you have a rough guess as to how compressed the outbound audio is?

I was under the impression that the audio is stored locally while listening for the trigger word. Once the trigger word is recognized, Alexa begins recording and sends the result to be processed.

Reply on news.ycombinator.com