I hear this from a lot of people and it makes no damn sense at all. Do you carry a smart phone? If so, do you realize it has a microphone, speaker, camera and multiple tracking devices in it? Do you take it with you everywhere, including in your house? If you're "off the grid" (and somehow still on SoSH all the time), then good for you for your bold stance. If you're not, then how is it you've avoided having a spy machine in your house?Yet another reason not to have the Amazon spy machine in my house.
Yes and no. Yes, in that there is an active microphone. No, in that that data is not (and we can verify it's not) going anywhere. The reason that Echo users have to pick between four wake words is because the necessary stuff to recognize the word is in hardware on the device. Only after it triggers the wake word is anything sent to Lex (the AWS tooling that powers it).An Echo/Google Home/HomePod has to always be listening in order to be a functional product.
My comment was tongue and cheek. The real reason I don't have one is because I really don't want one.I hear this from a lot of people and it makes no damn sense at all. Do you carry a smart phone? If so, do you realize it has a microphone, speaker, camera and multiple tracking devices in it? Do you take it with you everywhere, including in your house? If you're "off the grid" (and somehow still on SoSH all the time), then good for you for your bold stance. If you're not, then how is it you've avoided having a spy machine in your house?
Today I used Shazam for the first time, and it was so good it took me aback and made me wonder about how to turn off the microphone on my G5 Plus. Checked settings to no avail, tried Googling. Does anyone know how to do that? (I can find out how to turn it off on various apps. Does that mean that unless those apps are being used, the microphone is off? ---i.e., Google isn't listening unless I say OK Google, and Shazam isn't listening unless I tap and ask it to?) I really have no big secrets, but I grew up in an age in which you kept your conversations private, and I'd like to continue to do that.I hear this from a lot of people and it makes no damn sense at all. Do you carry a smart phone? If so, do you realize it has a microphone, speaker, camera and multiple tracking devices in it? Do you take it with you everywhere, including in your house? If you're "off the grid" (and somehow still on SoSH all the time), then good for you for your bold stance. If you're not, then how is it you've avoided having a spy machine in your house?
It's addressed a bit above by @Blacken but Google is always "listening" but never sending anything anywhere until the wake words "Ok, Google" or "Hey, Google" are heard. Furthermore, if you train it, Google will only respond to your voice. All other apps require permissions to use the microphone but if you're like everyone else you just accept whatever permissions the apps ask for without paying much attention. Apps are able to listen for keywords and then react so it's entirely possible for you to install an app that does this without you knowing about it if you just zip past the permissions screens. You can very easily go into your app permissions and see what apps have "Microphone" permissions if you're really worried about it.Today I used Shazam for the first time, and it was so good it took me aback and made me wonder about how to turn off the microphone on my G5 Plus. Checked settings to no avail, tried Googling. Does anyone know how to do that? (I can find out how to turn it off on various apps. Does that mean that unless those apps are being used, the microphone is off? ---i.e., Google isn't listening unless I say OK Google, and Shazam isn't listening unless I tap and ask it to?) I really have no big secrets, but I grew up in an age in which you kept your conversations private, and I'd like to continue to do that.
P.S. My husband was just reading about an app called Mute Mic, but it's no longer available. Google Play has others, but I never know whether it's a good idea to add all these unknowns to one's phone.
I presume that you know that your phone app has a mute button so are you asking about something else?Thanks for the reply. I am not really alarmed by Shazam, but its sophistication (it very quickly identified a fairly obscure piece by Shubert) set me to thinking about the issue of the phone's abilities to "listen."
So I guess from what you are saying that one handles this app by app rather than being able to mute the phone. But it would be useful to be able to mute briefly while talking on the phone to consult with your significant other.
I'm not sure I would even categorize it as ML - here's the original paper. I would say it's just signal processing, they aren't even suggesting something as complex as a k-mean (I guess their hash generation process is sort of similar to a k-mean). For fast searching, they seem to indicate that they are exploiting some inherent properties in the dataset being music, by pairing frequency hashes with delta-t in the sample fragments they get nice precision.Going beyond the worry that people are listening to it, you're mistaking what Shazam is doing. It's basically just a DB of known song and movie audio fingerprints (which sounds complicated but is actually pretty easy). It then uses some rudimentary machine learning algorithms (probably just k-means if I had to guess) to get a ranked list of matches based on the audio fingerprint coming from your phone. Unless you're REALLY worried about the world knowing what songs you're listening to, Shazam isn't much of a risk.
I wonder if it actually just uses local compute now to generate the signatures locally and just do the inference remotely? Lots of ways to skin that cat I presume.I'm not sure I would even categorize it as ML - here's the original paper. I would say it's just signal processing, they aren't even suggesting something as complex as a k-mean (I guess their hash generation process is sort of similar to a k-mean). For fast searching, they seem to indicate that they are exploiting some inherent properties in the dataset being music, by pairing frequency hashes with delta-t in the sample fragments they get nice precision.
Shazam has been around a long time (~2002), I had forgotten that the original implementation was something you actually called, it listened to 30 seconds and then it texted the song info back to you. That meant their fingerprinting had to work thru a very narrow-bandwidth channel (about 3kHz for GSM calls). It might use higher bandwidth now that it's running as an app with full-bandwidth access to the handset microphone but the original one certainly didn't.