Two Years of Voice-Based Assistants, Echo and Home

>What I need is natural natural language processing to become so good that it just works.

That's supernatural language processing. You're asking the system to recognize a thought you have in your mind that you can't effectively communicate with words.

The loud keyboard - makes sense. But the company name? My first thought was IBM or Cherry.

I think it can be improved, but I'm never going to expect voice recognition to fill in the blanks for something I can't describe with my voice.

Heck screw the helping you complete your thought for you, if it just handled umms and uhhs and pauses without going "I didn't get that" I would be 100% happier

There's also a legal/moral question of whether it's ok to capture the speech of children to build better models for this kind of thing.

I use it to turn my lights on and off. It works great for this, I don't need to leave my chair/couch/bed and I can turn many lights on quicker than any other method I'm aware of.

I use it to play music, usually through Spotify. I use Amazon music occasionally when I'm looking for a song (usually while cooking or the like) but on the whole it's inferior.

I ask for the weather in two or three cities while I'm tying my shoes, so I can decide if I need a jacket.

I use it for timers when I'm already using my oven timer. I appreciate being able to ask how much time is left while I'm not in the kitchen and thus can't see the oven.

I use it for unit conversions (how many oz in a quart? How many tablespoons in half a cup?).

The most use I get from it is when my hands are busy, my phone is either: in my pocket, or in another room plugged in charging. Or dead (this happens less frequently since I bought a new phone recently). I also find that context matters. When I use my phone's voice assistant I'm usually trying to do something on my phone. With my Echo I'm just trying to do something. I have a lot more success with the Echo.

I bought a few dots when they went on sale this fall, but I've only set up one. It works surprisingly well with the echo. They rarely get confused. Sometimes the device much farther away responds instead.

You can set up a passcode for ordering, or disable it entirely -- I've never used it and probably never will. I already carry a microphone around with me all day and work in an office surrounded by literally hundreds of microphones, or a few dozen in my house (my Dropcam is always on and has a mic. My TV remote has a mic for some reason I hope to never discover. My laptop, my iPad, and various other devices have microphones too).

Look, I didn't see a use case for it either. But once I got on, it infiltrated my home and now feels like an extension of it. I also didn't see the point of smart phones, tablets or IoT devices in general, but now I have all of them.

It's clear this sort of interaction is the future.

Under your definition, your phone is a 24/7 microphone connected to either Google or Apple, your PC a 24/7 microphone connected to Microsoft and literally every other vendor that has a service running.

The privacy concern is valid, but its also valid for literally every other piece of electronics. At some point you have to trust that it's only recording during the short period after you say "OK Google" and if it were doing something more nefarious someone would figure that out and it'd be huge and damaging news.

These are basically first-generation products. The idea is to improve upon their problems, find where they are useful and where they are not, and step closer to having a real virtual assistant.

What would a voice assistant have to do to convince you otherwise?

My son-in-law was over for Christmas, and he didn't know that I had an Echo dot, so to show it off I said, "Alexa, tell me a joke". It told us a lame joke. My SIL had to whip out his android phone and say, "Hey Google, tell me a joke!" and it told us a lame, funny joke. Score one for Google.

Then he said, "Google, tell me a Christmas joke" and it told us a pretty poor Christmas joke. I then asked Alexa for a Christmas joke, and it told us a funny Christmas joke!

So, clearly, in by far the most import imaginable use case, they're pretty much tied :-)

My Nexus 5X has asked me a few times if I'd like to unlock it with my voice, so it seems like they've got it working to some extent.

I think that beyond the security concerns, it will be really useful when you share the assistant, so that it can have the proper context for recommendations, calendar info, etc.

It's way more than playing music by speech. It's looking up stuff, controlling your house, checking restaurant hours, all sorts of functions that would otherwise be fulfilled by a smartphone with less input effort.

Give one of these devices a try, you might be surprised by how much you adapt to them.

I think it's Actions on Google (but I've never tried it): https://developers.google.com/actions/

It looks like there's a "user friendly" version that's api.ai, and then there's an SDK, and the SDK seems to be node.js based. I'm not clear from reading it if you can do a private version just for your device. I'll have to kick the tires one of these days - thank you!

You can delete all of your voice recordings from Alexa using the companion app.

The Home did a better job of interpreting requests, mostly. We found that we had to contort ourselves more to get the Echo to do some things. "Play the Nutcracker" -- no go. Even though it was in our library. "Play tchaikovsky's nutcracker suite" would work, IIRC. The home handled both. Informational requests also worked a bit better on the Home. ("What is a..."). We weren't using any extra skills that we'd miss, so it was a very easy transition.

I wish the Home had the equivalent of the Dot. I wish the Echo had the Home's inference capability.