Google stores voice searches | Hacker News Reader

Google stores voice searches

independent.co.uk

94 pointsapsec11237 comments

Nonsensical scaremongering. I've reviewed my history for several months. I use voice a lot, especially in the car. My history only shows items where I've used Google Now (button press) or used keyboard voice dictation (SMS, Waze, etc. - also button-press). Easy to recognize, you even hear the beep at the start.

I don't have "OK Google" set to "always on" on my phone. Anyone who sees "strange" items may have that setting enabled.

PS Thanks OP, I wasn't aware of this overview page. Insightful. I appreciate Google allows you to delete the samples if so inclined. Where is this page on FB? ;-)

I have always on enabled and reviewed my recordings. Same results as you. Voice searches I initiated and voice keyboard input.

For me it's only showing recordings of voice search I performed. Not a single audio asynchronously recorded in two years. So not really "conversations people have around phones." And I also use Android Wear and Chromebook, so I'm probably more exposed to Google recording stuff.

When I'm performing voice search, it's obvious Google is listening. When I'm using a third-party aid or service, it is obvious they keep records. This is expected, and they let you delete them.

To perform voice search you have to say "ok, google", so phone needs to listen and to send to server every phrase it hears in order to check whether you said "ok, google" or just ordered a beer in a pub. I'm just trying to theorise.

I thought that was client side? Dunno

Possibly, but who knows

It's easy to tell otherwise. If the OK Google detection was server-side, there'd be a constant flow of data trickling from the device to some server (at least when the phone detected sound). No such flow is seen.

Reasons to think it's client side:

* Sending all microphone input to a remote server would be really bad for battery life and data usage.

* It still works even in airplane mode.

* On ChromeOS (which isn't Android, but may share the same code) their privacy whitepaper [1] says:

    If you opt-in to the feature, Chrome OS will listen for
    you to say "Ok Google" and then send the audio of the
    next thing you say, plus a few seconds before, to
    Google. Detection of the phrase "Ok Google" is performed
    locally on your computer, and the audio is only sent to
    Google after it detects "Ok Google"."

[1] https://www.google.com/chrome/browser/privacy/whitepaper.htm...

Sending all microphone input to a remote server would be really bad for battery life and data usage.

It still works even in airplane mode.

I agree that they probably only send speech recorded after the activation words. However, I don't arguments think your arguments hold up. Speech data compresses exceptionally well and generally requires lower sampling rates and/or resolutions. So, it's feasible to compress the speech signal (possibly hardware-assisted), buffer it, and send it to the server at some regular interval or when the user wakes the phone.

Yes, it is possible for them to be sending all recorded sounds (even if the hotwords are not detected) to a remote server. However, as discussed above, there would be no legitimate reason for doing so—it is demonstrably true that the phone is able to recognize the hotwords even when a remote server is not reachable.

If you're only interested in spying, sure. If you want the phone to respond when a user says "ok, google" rather than a minute later it wouldn't work.

Reasonable, thanks. Question is if I said "ok ..." in a noisy environment or my accent is awful or I'm eating an apple, will android try to reqognize the phrase on server or not? If so, phrases similar to "ok, google" may be saved by google.

Like just the other day, I was having a conversation with my friend Snu. I said, "Okay Snoogle, it's time to dispose of the bodies of that couple we killed downtown last night.

Also, keyword detection algorithms are different from general purpose speech transcription algorithms. The general pattern is to write a specialized model that's really good at robustly identifying the key phrase, run that continuously, then trigger recording and full transcription only when the key phrase is heard.

Those keyword identification algorithms can be implemented at the hardware level. I believe this is done for specialized devices like Echo and Kinect, not sure how phones do it.

The 1st generation Moto X had a special minimal chip used just to detect you saying "Ok Google Now" (or a custom phrase) to turn the phone on when it was sleeping. But I think phones that only detect a start word when the screen is on probably don't have custom low-power hardware for it.

> * Sending all microphone input to a remote server would be really bad for battery life and data usage.

Not if you transcribe it first, and upload compressed text.

It may specify it within their patent [1]; "Contextual advertising for video and audio media " .

[1] https://www.google.com/patents/US9087331

Transmitting every piece of audio the phone receives would kill your battery in less than an hour. They absolutely need to do the "ok, google" detection on board and the reason for differential adoption between various phones seems to be that for always on with reasonable battery times they need to do it on a DSP rather than the application processor.

What I found was that (when I still had a Moto X), it would sometimes 'hear' the activation phrase although I didn't utter it and records the next few seconds or so, until it decided that it cannot answer anything. So, in Google Dashboard I saw that it recorded some private conversation, though it was obviously not intentional.

This is clearly for clickbait. Only your voice search commands are recorded just like the history of what you searched using Google or your browser history.

The article speaks in "may"s and "might"s a lot, I can't find one concrete point outside the title, besides of course "after activation by speaking OK Google your audio is sent to Google". It seems to be clickbait indeed.

Could they be? Sure. Are they? We'd need some kind of evidence, not speculation

That's not true. A friend of mine found some random bits of conversation too (he says approximately 90% of the recordings are actual search).

The discussions about Google (et al) tracking show a strikingly similar pattern to discussions about clandestine government surveillance: a fairly even split between people saying "that's not what they're doing" (they only listen to explicitly triggered searches), "it's appalling that they're doing this" (it's a privacy breach) and "we've known they're doing this for years" (how else would they have such good speech recognition).

Personally I think turn voice search on/off checkbox should be much more visible than it is now. Also I experienced at least one time when it was reset by Android update or Google Search app update, so now I have to check it is certainly off after every update.

Has anyone actually saw a single non-search recording of a conversation on that page?

Almost all of mine are recordings of conversations I was having. Most of them sound like my phone was in my pocket and are very muffled, and very few of them have a transcript.

2 of the last 20 recordings are actually questions I asked. 5/20 have transcripts, and the rest are partial recordings of conversations I've had with no transcript.

I never use the voice search, so all of mine were background noise of times I accidentally hit the voice search button.

How is that new? Google is pretty much the only platform (i know of) that barely has issues with crazy swiss german dialects. Bots do not magically learn to understand this stuff.

Spooky, because I'm conscious about privacy & regularly check that my tracking is limited to only the things that I explicitly allow. Yet, somehow Google managed to flip it on for a few weeks over the past 2 years.

As someone else pointed out, 'app updates' or 'adding a new device' resets these settings. If I had one of those "always on" devices, it would have a lot more conversations stored, some that don't even start with the hotword.

If you are really paranoid, we've done a Python script for "sandboxing" the audio use in Google Chrome: https://github.com/nektra/ChromeAudioSandbox it is a sample to show the capabilities of our open source instrumentation lib.

Anyone want to explain why Google requires you to turn on these recordings to use voice search?

I want a setting where they automatically get deleted after processed, which does not appear to exist.

Turn off Voice & Audio Activity,

https://www.google.com/settings/accounthistory/audio

> When Voice & Audio Activity is off, voice searches will be stored using anonymous identifiers. This information won't be saved to your Google Account even if you're signed in.

https://support.google.com/websearch/answer/6030020

Huh. Pretty sure this wasn't available 6 months ago or so when I set up Google Now. You needed to opt in to search and app history to use Now, which included OK Google, or something.

Anyway to delete all the web searches? Seems like there is a way to select a day's search but not all.

Read the article:

To get rid of everything, you can press the “More” button, select “Delete options” and then “Advanced” and click through.

duh.

Reply on news.ycombinator.com