These solutions don't mean to be help/assist their users but to help their vendors to collect information about the users.
Due to their limited communication capabilities they're extremely poor at collecting meaningful data - there's simply no point in talking to them as their utility fails with anything but the simplest and most direct commands. There's not much data in "turn kitchen lights on", "set alarm at 10am", "play $album by $artist", "remind me to buy milk" and "convert 4.7 miles to kilometers" - at best, this exposes a few minor behavioral patterns that can be trivially determined even without a dedicated device to talk to.
And, hopefully, outright spying wasn't one of the design goals, even though they can be of dual use. Google, Apple and Amazon may be cooperative but they aren't NSA, after all.
My opinion, it's about use of particular services. Alexa will push you towards Amazon purchases, use of Amazon Music, etc. Similarly, Homepod requires investment into Apple ecosystem. And Google Assistant will push towards all things Google. In my opinion, voice assistants were made to be helpful within their particular ecosystems and thus add more ties to their vendors' services. (And, sure, those services collect data.)
They're clearly designed to make money by keeping you in the Amazon / Google / Apple ecosystem. I don't think any of them make much money from data collection because they don't collect that much data.
I have Google Homes and you can go and look at all the data it collects - for me it is 90% "hey Google play some music" or "hey Google what's the weather" or "hey Google set a timer for 10 minutes".
Useful for improving they speech recognition but not much else.
With it, google can identify you by voice, put a mic in a public infopanel, and gues which commercial has most impact.
If the NSA has access, they can classify public recordings by people at that place, so I assume they are interested.
HN has so many conspiracy theorists because we know what is possible with infotech on an industrial scale, the spooks have the money for it, and snowden basically provided the proof they did it.
There are infopanels (from smaller dodgy companies) with cameras which try to identify viewers behaviours. But actually expecting Google to both identify you and measure response is a bit out there for many reasons. 1. They actually do care about access to data they store - creating a fingerprint database for identification would be beyond what they do. 2. Who actually reacts vocally to an ad in a way that can be captured without environment noise - the assistant can't even set the timer correctly sometimes when I talk to it in perfect conditions. 3. If they did want to do super-shady targeting, your location data crossed with ad location + search/browser history is both better and cheaper than a voice database.
Would it? What if it was structured the way they handle other personal data: customers send in "voice telemetry" from their "info"panels, and Google gives them bulk analytics ("#x different people talked in your venue, with following age/demographic split"), as well as using that telemetry to handle ad attribution and ad targeting in the background. This way - just like with regular web telemetry - neither the advertiser not the "info"panel owner get to see the voiceprint data, but it still gets used to uniquely identify individuals and target ads at them.
As ads get more and more effective, people will be incentivized to put up more and more ads. But there are already way too many ads for my liking.
There are too many ads but it is possible to avoid a great percentage of them if inclined. My interaction with ads is pretty minimal and when presented often enjoyable. Using uBlock Origin I don't see ads on most sites. Youtube premium means no ads there as well. Spotify premium, so no ads for music. Video streaming services like netflix don't have ads.
My only ad interaction left is reddit(which offers premium) and podcasts or youtube video in-content ads where they are narrated by the creators whose content I am happy to support and in some cases I find the way they switch to ads quite funny.
The tailor failed utterly and the "bespoke" version was worse than the old version.
Add to this that they want to keep records on me and it just isn't worth it.
That's the extreme case, more than likely people are a bit unnerved when an unrelated site shows them something they looked up on amazon an hour ago.
Even so I suspect a lot of these complainers just have hidden internet hobbies they just want completely private.
But that is very different from saying that Amazon and Google created them in order to gather data. Amazon was the first to make a voice assistant and their intent was pretty clearly to make money by getting people to buy it and to buy stuff from Amazon through it. To suggest that their true motivation was data collection is pure conspiracy theory.
Also voice identification is a relatively recent (and not terribly reliable) feature so it can't have motivated their creation.
Hell there's no way it's voice identification is good enough for the fantasy scenario you describe. It's only just about good enough to distinguish between 2-3 people.
Who does Google even need to identify my voice? They already know where I am from my phone.
This is what I mean about HN conspiracy theories. It's not just that people here know more about what is possible. None of the conspiracy theories make any sense.
The NSA isn't interested in you period. They just want access to all data so they can hit a specific target.
There's technical capabilities and there's technical interest. Your and there analysis fails at the "interest" part.
Reflexive cynicism is easy, and makes you appear smart to a large number of commenters.
Of course, it doesn't help that often times, the automatic cynicism is entirely accurate, especially when contrasted against empty PR statements.