4 ms·
These solutions don't mean to be help/assist their users but to help their vendors to collect information about the users.
by meheszjeno 6y ago
These solutions don't mean to be help/assist their users but to help their vendors to collect information about the users.
- trevyn 6y agoCorrect; incentives are not aligned. Even for Apple.
- briga 6y agoCan't they do both? I personally find Google Assistant very helpful. I'm sure Google finds my personal data helpful. Seems like a fairly reasonable trade-off for me
- imglorp 6y agoI've had the opposite experience: completely useless due to poor integrations. Example: "Send a slack to Mary". Fail. "Read my most recent text message", fail. Even Hangouts is badly integrated, if at all. I think the problem is, as always, a combination of business goals. First, I suspect they don't want to help apps they haven't made a cross marketing deal with, like Slack. They'll install the app but won't work with it. Eg, "Play Tool on Spotify" works fine. Second, Google's internal novelty-chasing merit system doesn't reward a coherent customer experience. You'd think Hangouts (or whatever chat is called this week) would be seamless after 5 years, but it's like they never met.
- drdaeman 6y agoI disagree. It's not about collecting information - at least, not directly. Assistants don't seem to be designed to gather neither any substantial amounts of information, nor any unique information. 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.)
- IshKebab 6y agoWhy does HN have so many conspiracy theorists? 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.
- hyperman1 6y agoThere is 1 more important piece of information: Your voiceprint. 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.
- viraptor 6y ago> put a mic in a public infopanel, and gues which commercial has most impact 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.
- TeMPOraL 6y ago> 1. They actually do care about access to data they store - creating a fingerprint database for identification would be beyond what they do. 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.
- TulliusCicero 6y agoThey're pretty obviously designed to do both.