Microsoft revives Xbone Voice Control using Alexa & Cortana powered devices

[Master Disaster]

56 minutes ago, Commodus said:

Not strictly speaking.  So many people think smart speakers are actively recording everything you say, but they're not. They're waiting to hear a specific speech pattern and then wake up to listen.  Now, what was just proposed (where they'd do things based on what you say in casual conversation) would require constant listening, and that's a problem.

Let's be real, you actually have no idea what they're listening to, what they're recording or where the recordings are being sent to.

 

I'm not generally a conspiracy nut but the truth is it's every bit as possible that they are recording everything as it is they aren't.

 

I find the idea of a microphone sitting in every room of my house quite disturbing, it's not like I've got anything to hide either, I'd just rather not give anyone the chance to record every moment of my life without me knowing.

[ZacoAttaco]

To be honest, if there were going to reintroduce this feature, having it linked to something like an Amazon Echo would be the way to go. Have it connected to other devices rather than it's own standalone product for me voice functionality becomes that much more useful.

[AngryBeaver]

I have converted a good portion of my house to being a smart home. My thermostat, I have smart plugs for certain devices like lamps, fans, etc. Then I have the smart bulbs with the hue technology. I have 2 echo shows, and either 3 or 4 echo spots. I also have a Fire TV Cube coming Monday. I have pretty much cut the cord, but I am going to use an external antenna plugged in to a HDHomeRun Extended box. It streams life tv to the connected devices on my network and also can transcode to h264 and send it to my NAS for DVR functionality. Then I have your normal, Hulu (no commercials), Netflix, Amazon Prime, etc. Probably need to pick up another 2 firesticks so they can be controlled via alexa as well.

 

My point is... smart homes are going to be the future. I was an early adopter, but they are actually really nice when setup correctly.

[AngryBeaver]

9 hours ago, Master Disaster said:

Let's be real, you actually have no idea what they're listening to, what they're recording or where the recordings are being sent to.

 

I'm not generally a conspiracy nut but the truth is it's every bit as possible that they are recording everything as it is they aren't.

 

I find the idea of a microphone sitting in every room of my house quite disturbing, it's not like I've got anything to hide either, I'd just rather not give anyone the chance to record every moment of my life without me knowing.

Did you know that android devices before the new updates pushed in the last 8-12 months did just that? Apps found they could switch on the microphone on these devices and listen to conversation at random intervals then use keywords to setup ad suggestions. Think bad if you ever had a conversation and then noticed an advertisement pretty close to your conversation. Thankfully this was patched and is why you now have to approved every item that an app wants access to.

[ZacoAttaco]

27 minutes ago, AngryBeaver said:

Did you know that android devices before the new updates pushed in the last 8-12 months did just that? Apps found they could switch on the microphone on these devices and listen to conversation at random intervals then use keywords to setup ad suggestions. Think bad if you ever had a conversation and then noticed an advertisement pretty close to your conversation. Thankfully this was patched and is why you now have to approved every item that an app wants access to.

Cheeky stuff man. And this is just what we know at this point, imagine what's really going on behind the scenes that we have no idea about...

[LAwLz]

19 hours ago, GoodBytes said:

But seriously, I think such sci-fi may come a reality, once we can make a really good system that can run on a processor that is inexpensive, and have no need for external servers.

The "need" for external servers will never disappear because it would hinder companies like Microsoft and Google from making money from your personal data.

Just look at Skype as an example. It used to be peer-to-peer, without having to send a bunch of traffic to external servers. We had the processing power, storage and bandwidth do it it. However, once Microsoft bought Skype they started making it so that traffic went to their servers instead.

 

It's not functionality or processing power that determines if we have external servers or not. It's the companies profit margins that decide where our data is stored.

 

 

18 hours ago, Commodus said:

Not strictly speaking.  So many people think smart speakers are actively recording everything you say, but they're not. They're waiting to hear a specific speech pattern and then wake up to listen.  Now, what was just proposed (where they'd do things based on what you say in casual conversation) would require constant listening, and that's a problem.

Ehm... How do you think it can detect a specific "speech pattern" like "Alexa" if they aren't always listening?

Smart speakers and other things which are voice activated have microphones which are constantly on and analyzing what you say. The difference might be (I don't have one so I can't check for sure) that once the always-listen function detects a keyword like "Alexa", it starts transmitting what it hears to Amazon. So it's always listening, but not always uploading.

 

 

17 hours ago, Master Disaster said:

I'm not generally a conspiracy nut but the truth is it's every bit as possible that they are recording everything as it is they aren't.

Wouldn't be too hard to monitor the network activity and figure it out.

I mean, it could be recording locally too for uploading later, but a teardown would reveal if it has the memory built in to do that. Then again, speech requires very little storage space. Phone call quality is like less than 1KB per second. Here is a 109 second voice clip that's less than 90KB.

So 20 minutes of recording would be 1 megabyte.

 

Not saying that they are recording everything being said, but to activate it at random times and recording a few minutes would be extremely hard to detect since the amount of data needed is so small. It could easily be disguised as keep alive or some type of telemetry data.

 

 

8 hours ago, AngryBeaver said:

Did you know that android devices before the new updates pushed in the last 8-12 months did just that? Apps found they could switch on the microphone on these devices and listen to conversation at random intervals then use keywords to setup ad suggestions. Think bad if you ever had a conversation and then noticed an advertisement pretty close to your conversation. Thankfully this was patched and is why you now have to approved every item that an app wants access to.

I think you are misremembering the story.

The granular permissions were introduced in Android 6.0 over 3 years ago. It was not a response to apps using your microphone to listen in.

The story you're thinking of was that some apps asked for permission to use your microphone, and once that was granted it would run in the background and use the microphone to listen for TV ads and shows. You still had to give them permission to use your microphone. It's just that people didn't expect the apps they gave permission to, to actually run in the background and listen.

[AngryBeaver]

2 hours ago, LAwLz said:

The "need" for external servers will never disappear because it would hinder companies like Microsoft and Google from making money from your personal data.

Just look at Skype as an example. It used to be peer-to-peer, without having to send a bunch of traffic to external servers. We had the processing power, storage and bandwidth do it it. However, once Microsoft bought Skype they started making it so that traffic went to their servers instead.

 

It's not functionality or processing power that determines if we have external servers or not. It's the companies profit margins that decide where our data is stored.

 

 

Ehm... How do you think it can detect a specific "speech pattern" like "Alexa" if they aren't always listening?

Smart speakers and other things which are voice activated have microphones which are constantly on and analyzing what you say. The difference might be (I don't have one so I can't check for sure) that once the always-listen function detects a keyword like "Alexa", it starts transmitting what it hears to Amazon. So it's always listening, but not always uploading.

 

 

Wouldn't be too hard to monitor the network activity and figure it out.

I mean, it could be recording locally too for uploading later, but a teardown would reveal if it has the memory built in to do that. Then again, speech requires very little storage space. Phone call quality is like less than 1KB per second. Here is a 109 second voice clip that's less than 90KB.

So 20 minutes of recording would be 1 megabyte.

 

Not saying that they are recording everything being said, but to activate it at random times and recording a few minutes would be extremely hard to detect since the amount of data needed is so small. It could easily be disguised as keep alive or some type of telemetry data.

 

 

I think you are misremembering the story.

The granular permissions were introduced in Android 6.0 over 3 years ago. It was not a response to apps using your microphone to listen in.

The story you're thinking of was that some apps asked for permission to use your microphone, and once that was granted it would run in the background and use the microphone to listen for TV ads and shows. You still had to give them permission to use your microphone. It's just that people didn't expect the apps they gave permission to, to actually run in the background and listen.

I will need to find the article on it. It is a very similar usage, but it was more recent and this was using the microphone for 10-15 second intervals to look for keywords/phrases in conversations. Also they did add a permission system back then, but only in the last year did they really tighten up the permission system. I will dig in to this a little more when I get a chance, don't want to give false information, but I am pretty certain this is a more recent thing.

[Commodus]

21 hours ago, Master Disaster said:

Let's be real, you actually have no idea what they're listening to, what they're recording or where the recordings are being sent to.

 

I'm not generally a conspiracy nut but the truth is it's every bit as possible that they are recording everything as it is they aren't.

 

I find the idea of a microphone sitting in every room of my house quite disturbing, it's not like I've got anything to hide either, I'd just rather not give anyone the chance to record every moment of my life without me knowing.

No, I do know some specifics of what they're doing, and it's this irrational "omg they're spying on everything you do" screaming that prevents people from having a serious discussion about privacy implications.

 

Smart speakers like the Echo and Google Home have dedicated components that only wake up the full listening system when they hear a specific voice pattern -- that is, the hotword (Alexa, OK Google, hey Siri).  Until then, the speaker doesn't know what you're saying and isn't actively recording.

 

What happens after that varies, but Amazon and Google only ever get the snippets from issuing those commands, and that's to build preferences (you can delete all the info if you like).  Apple, meanwhile, purposefully avoids tying requests to anything personally identifiable... so if you want a smart speaker but you're really concerned about privacy, go get a HomePod (and an iPhone, since you'll need that to set it up).

 

Besides, stop and think about the implications of what you're claiming for a moment.  Are Amazon or Google going to process audio data from millions of speakers 24/7, let alone keep it all for even a brief period?  No.  It's not even necessarily the server demands that would be excessive -- it's that you'd end up with data that's almost entirely garbage.  Even if Amazon and Google were mustache-twirling villains (they're not), they wouldn't want to indiscriminately collect data like that.  They're better-served by listening to intentional commands, since that gives them a sense of your preferences.

[Commodus]

3 hours ago, LAwLz said:

Ehm... How do you think it can detect a specific "speech pattern" like "Alexa" if they aren't always listening?

Smart speakers and other things which are voice activated have microphones which are constantly on and analyzing what you say. The difference might be (I don't have one so I can't check for sure) that once the always-listen function detects a keyword like "Alexa", it starts transmitting what it hears to Amazon. So it's always listening, but not always uploading.

You could say it's a matter of semantics.  When people hear "always listening," they mistakenly assume that means "it understands everything I say."  That's not what's happening, though.  Instead, there's a component in the speaker that's only waiting until it hears the voice pattern that represents the keyword.  It doesn't know what you said before that, and it only analyzes things during the time window when it's listening for that command.

[Jito463]

4 minutes ago, Commodus said:

You could say it's a matter of semantics.  When people hear "always listening," they mistakenly assume that means "it understands everything I say."  That's not what's happening, though.  Instead, there's a component in the speaker that's only waiting until it hears the voice pattern that represents the keyword.  It doesn't know what you said before that, and it only analyzes things during the time window when it's listening for that command.

Semantics, it's still listening.  Try using that argument with the couple who had a private (albeit largely irrelevant) conversation recorded and uploaded to a random contact via e-mail by Alexa.  Listening is still listening, regardless of what it's "listening" for.

[SupremeGOAT]

hmmmm... xbox seems to be making strides.... yeah.. real progress...

[jagdtigger]

3 hours ago, Commodus said:

Are Amazon or Google going to process audio data from millions of speakers 24/7, let alone keep it all for even a brief period?  No.

Can they trigger on more than one "hotword"? Yes...........

[Commodus]

4 hours ago, jagdtigger said:

Can they trigger on more than one "hotword"? Yes...........

There's a very narrowly defined set of hotwords.  You'd have to show proof that they're inserting secret hotwords and keeping the active transmission of commands open for longer than the usual several seconds.

 

Please, drop the paranoia and work based on evidence.  Evidence.  Facts.  Reality.  Not speculation or fearmongering.

[jagdtigger]

17 hours ago, Commodus said:

There's a very narrowly defined set of hotwords.

And how can you be sure they telling the truth?

 

17 hours ago, Commodus said:

You'd have to show proof that they're inserting secret hotwords and keeping the active transmission of commands open for longer than the usual several seconds.

Its pretty easy for them to hide the undisclosed hotwords since they are using closed source SW. Plus they can silently upload offline stuff while you use them.

 

17 hours ago, Commodus said:

Please, drop the paranoia and work based on evidence.  Evidence.  Facts.  Reality.  Not speculation or fearmongering.

Just like when MS implemented their spyware in win10 and no-one believed any of it until it blew up? Id rather not wait for that. Its better to stop this in its tracks....

[Commodus]

29 minutes ago, jagdtigger said:

And how can you be sure they telling the truth?

 

Its pretty easy for them to hide the undisclosed hotwords since they are using closed source SW. Plus they can silently upload offline stuff while you use them.

 

Just like when MS implemented their spyware in win10 and no-one believed any of it until it blew up? Id rather not wait for that. Its better to stop this in its tracks....

I can be reasonably certain because they don't have much incentive to record absolutely everything you say, and it would be really easy to spot this activity (just monitor network activity and look for spikes when you aren't triggering a known hotword).  As I wrote earlier,  always-on recording would produce a vast amount of garbage that wouldn't be useful.

 

Ever heard the claim "extraordinary claims require extraordinary evidence?"  That applies here.  There is no substantiated evidence to suggest there are hidden hotwords or always-on recording; there is plenty of evidence to suggest smart speakers are operating as the companies claim.  If you're still convinced that they're always spying on people, it's up to you to prove they are.  We can't just operate on hunches, myths and paranoia.

[EarthWormJM2]

On 9/7/2018 at 9:37 AM, Master Disaster said:

I've just had a thought, this will lead to serious trolling.

 

You know how Alexa has issues because it responds to anything saying it's name?

 

How many games are going to be cut short by friends shouting "Alexa turn the Xbox off" lol.

 

On 9/7/2018 at 9:38 AM, VegetableStu said:

"HAHAHAHA EAT SHIT JAKE"

"ALEXA, TURN OFF XBOX"

"HEY WHAT THE F--" *5 players have left the game*

Ahhh, welcome to 21st century trolling, where the trolls can be even lazier at pissing everyone off...

 

That will also really suck for anyone named "Alex" 

 

Me: "Alex, watch out for those snipers!"

Alexa: "watch sniper..." *switches to xbox store and pulls up the movie sniper to buy*

Me:  ...   -.-

[LAwLz]

1 hour ago, Commodus said:

I can be reasonably certain because they don't have much incentive to record absolutely everything you say

They have a massive incentive for it. Access to what you're saying when you aren't at your computer/phone is one of the few blind spots left for companies.

I would bet my left leg that several companies would be thrilled to be able to listen to everything being said in households. I mean, we already have proof that companies have been abusing the microphone permissions for this exact purpose. Companies have already been caught using the microphones in Android phones to listen in and track when and what people watch on TV.

 

1 hour ago, Commodus said:

and it would be really easy to spot this activity (just monitor network activity and look for spikes when you aren't triggering a known hotword)

If I were designing a smart speaker to be a spying device, I would equip it with a little bit of memory, maybe disguised as the storage holding the software, and then continuously record to that space in a circular buffer. Maybe even do something clever like only write to the buffer after 1 minute of continuous talk has happened (maybe a separate circular buffer for that, so you don't lose the first minute of conversation). That would prevent the buffer just getting completely emptied during long periods of silence (like night).

 

As I demonstrated earlier, you can record intelligible speech at 0.83KBps.

So 20 minutes of speech would take up roughly 1MB (slightly less actually).

 

If they were to send maybe 10 minutes of recorded audio inside each search, it would just be ~500KB extra data transmitted per search. Do that a few times a day and you can rack up maybe an hour of secretly recorded speech while simultaneously just looking like normal operations.

Would someone really be able to tell if their searches included an extra 500KB of data even if they were monitoring it? I mean, if they did it from the start it would be impossible to know how much of the data the hotworded recording took up.

 

1 hour ago, Commodus said:

As I wrote earlier,  always-on recording would produce a vast amount of garbage that wouldn't be useful.

And also a lot of very useful information.

Again, companies have already been caught doing this, so you can't just brush it off as "companies aren't interested in it".

 

 

1 hour ago, Commodus said:

Ever heard the claim "extraordinary claims require extraordinary evidence?"  That applies here.  There is no substantiated evidence to suggest there are hidden hotwords or always-on recording; there is plenty of evidence to suggest smart speakers are operating as the companies claim.  If you're still convinced that they're always spying on people, it's up to you to prove they are.  We can't just operate on hunches, myths and paranoia.

Where is that evidence?

I am not saying they are secretly recording people at all times, but it is not as impossible/improbable as you seem to think, and it's certainly not paranoia.

We already know companies are interested in listening in on microphones, and we know that it would be fairly trivial to disguise it.

On top of that we don't really have any way to check whether or not they are doing it.

 

To me, the claim that they aren't doing it holds just as much water as the claim that they are doing it. Personally I don't believe they are doing it, but I have absolutely 0 evidence to support that stance, nor do I have any evidence to disprove it.

[Commodus]

5 minutes ago, LAwLz said:

They have a massive incentive for it. Access to what you're saying when you aren't at your computer/phone is one of the few blind spots left for companies.

I would bet my left leg that several companies would be thrilled to be able to listen to everything being said in households. I mean, we already have proof that companies have been abusing the microphone permissions for this exact purpose. Companies have already been caught using the microphones in Android phones to listen in and track when and what people watch on TV.

 

6 minutes ago, LAwLz said:

And also a lot of very useful information.

Again, companies have already been caught doing this, so you can't just brush it off as "companies aren't interested in it".

They don't have an incentive to collect absolutely everything, though, and there's something you're forgetting: that garbage data could nullify the useful info.  Let's say you're talking about buying a new car in GTAV, but you're actually a broke college student who takes the bus everywhere in real life.  How does Google know not to serve you car ads?  Limiting recording to commands is more useful because it signals intent -- you know what the user wants.

 

14 minutes ago, LAwLz said:

Where is that evidence?

I am not saying they are secretly recording people at all times, but it is not as impossible/improbable as you seem to think, and it's certainly not paranoia.

We already know companies are interested in listening in on microphones, and we know that it would be fairly trivial to disguise it.

On top of that we don't really have any way to check whether or not they are doing it.

 

To me, the claim that they aren't doing it holds just as much water as the claim that they are doing it. Personally I don't believe they are doing it, but I have absolutely 0 evidence to support that stance, nor do I have any evidence to disprove it.

Didn't say it was impossible, but unlikely?  Yes.

 

The evidence comes from the way smart speakers are designed to work, the companies' own positions (not to take them at face value, but they have explained the basics of what happens) and the lack of personalized ads or other content based on info contained outside of commands.

 

And while it's not paranoia to say it's theoretically possible, it is paranoia the way someone like jagdtigger is acting, where he assumes the speakers are spying on people and would only be dissuaded if... well, let's be honest, nothing would dissuade him, because it's more about clinging to the fantasy than acting on reality.

[Master Disaster]

On 10/09/2018 at 9:20 PM, Commodus said:

There's a very narrowly defined set of hotwords.  You'd have to show proof that they're inserting secret hotwords and keeping the active transmission of commands open for longer than the usual several seconds.

 

Please, drop the paranoia and work based on evidence.  Evidence.  Facts.  Reality.  Not speculation or fearmongering.

So the multiple occasions where smart speakers have seemingly done random things without any instructions to do so don't count as evidence then?

[Commodus]

2 hours ago, Master Disaster said:

So the multiple occasions where smart speakers have seemingly done random things without any instructions to do so don't count as evidence then?

Not necessarily.  You'd have to show that there was intent to actively record and interpret what people were saying.   It's a bit like those people who attribute a weird event to ghosts: just because something is difficult to explain doesn't mean it's as sinister as you think it is.  Reality is often much more boring than we'd like it to be.