You're a technologically savvy person who keeps up with privacy news. At issue here is the precedent set: more and more devices by default transmit private information to third party servers without the knowledge of the users. Further, people who do not intend to be users (i.e., guests in a home with a SmartTV) suddenly are having their private information scooped up and stored without their consent. It requires a level of vigilance that not every person will have. Do I need to ask my host if they have a SmartTV with this function?

It used to be the case that a service would politely ask you whether or not they could collect data about you, and would provide the same service in both cases.

This option (to minimise data collection/retention) ought to be enshrined in law.

If you pay (or not) for a service, you should be able to get the service without the expectation of your personal and private data being harvested.

In this case the issue is that it is not possible for them to provide the service without potentially collecting private data:

The voice recognition is done on remote servers, and they have no means of preventing you from saying something private when it is recording.

The TV could store a "wake up" word locally. The TV doesn't send any audio anywhere until you speak the wakeup word, at which point in beeps and a light flashes and then all audio is sent away to the cloud for processing.

This is exactly how Amazon Echo does it:

Amazon Echo uses on-device keyword spotting to detect the wake word. When Amazon Echo detects the wake word, Amazon Echo streams audio to the Cloud, including a fraction of a second of audio before the wake word.

http://www.amazon.com/gp/help/customer/display.html?nodeId=2...

> including a fraction of a second of audio before the wake word.

Interesting that they are able to transmit audio-data that occurred prior to the wake word being said (in essence to transmit the wake word). Looks like Amazon is keen to redefine "collected" just like other groups are...

Why is that interesting? It's quite simple since the device is always listening locally, and just storing a second of audio in a buffer.

It's not "collected" since it only has 1 second of audio until it wakes up.

Because every time Amazon echo comes up, people always rush it "don't worry guys, it's not actually recording anything until it has to send to the cloud" and by Amazon's own admission, that's patently false.

Um, no, that's patently true.

Unless you are talking about the half second before the trigger word? But even then - it only sends when triggered.

I am indeed talking about the 1/2 second before. Even if it's 0.00001 seconds before the trigger word (I feel dirty saying that phrase), it's the technicality that I'm discussing. Once we allow those technicalities to slide, the truth is easier to bend and/or ignore. Reinforcing that PR/marketing is not an acceptable excuse to lie (even if it's a minor technical lie) is one of the few ways to prevent this sort of subtle-switcheroo.

> I am indeed talking about the 1/2 second before.

Then you are alone in this. Even amongst technological people, never mind the general public. This amount of time is perfectly reasonable.

> Once we allow those technicalities to slide, the truth is easier to bend and/or ignore.

No it doesn't. Slippery slope might be a valid argument in some places, but not here. You are overreacting.

> to prevent this sort of subtle-switcheroo.

There is no subtle switcheroo.

It's very very simple: It only sends audio when commanded to. That is the essence of the difference, and it's all that matters.

And don't tell me "next they'll send 1 hour of audio when triggered" because they won't. You are being ridiculous if you think that's where it's headed.

Taking outrageous positions like this just makes people ignore you.

> the trigger word (I feel dirty saying that phrase)

? Why? Does it have some special meaning I am not aware of?

True, but that still does not prevent the possibility of it capturing private conversations ongoing in the room.

> The voice recognition is done on remote servers

While possible, this is not theoretically necessary, and I find this highly doubtful in practice.

What about the delay on each voice command as it pings the server? If it was generally the case that voice commands were sent remotely, I suggest usability would be 0.

Siri, and Cortana (iOS and WP8 voice recognition engines) both send the clip to a remote server to perform recognition.

There are many services that provide real time processing of large amounts of data on server cluster/farms, and more are coming. It is very much possible to perform large computations in almost real time in these situations, and you'd be surprised at how little latency people notice, particularly when there's no indication of what the actual number is.

I wasn't aware that people generally found Siri usable.

It is genuinely comic when I see what kind of questions my non-technical friends ask Siri. I can't imagine the thought process of how Siri could put together the information from the ramblings of a person much less answer the query...

That doesn't add anything to this discussion. Siri is more than usable for a lot of things - I can set alarms, timers, and make calls without reaching for my phone. I'm not saying that it's revolutionary, but it definitely works.

From my point of view it did add significantly to the discussion.

I'm surprised to learn that people do use Siri in genuine situations, so thanks for your reply in that regard; although I still feel gratified that another commenter has the same experience that I do.

It is not theoretically necessary if the local unit has sufficient capacity, which it doesn't in most current implementations. Hence why most implementations of voice recognition in consumer hardware does send the clips off to remote servers for processing.

Judge the delay for yourself.

It might be practically necessary, though, if they need to have a huge corpus of data in memory in order to perform their voice recognition.

I'd believe it, my friend had one and the delay on all commands in all forms was very 'laggy'.

The delay is approximately the same or less of what fast local evaluation would be (ping within a country should be <50ms).

If you're on wired internet, if you aren't saturating your local network, if your provider isn't having any issues, if their servers are up, if there is no DNS delay or caching issues, if...

Relying on the internet for "real time" computation is a recipe for inconsistent behavior.

The fact that people are using these features of android and apple phones shows that it's 'good enough'. I frequently use it to issue navigation commands on my phone and it fails maybe 1 in 25 times. Well worth not typing in destinations when I want to go somewhere quickly.

The alternative, for the moment, is to not have the functionality at all.

Try a Fire TV box for a while. Search on devices without a keyboard is far easier with voice recognition.

I agree the fascination may be excessive, but there are a few legitimate use cases where it's not obnoxious (anyone within hearing will likely end up listening to the TV anyway, so presumably you have their acceptance for noise in the first place) and actually useful, and searching for stuff to watch is one.

So push a button to start voice recognition? Always on is not worth the potential privacy tradeoff.

Right now it isn't so useful, but once stuff like "put on that Wes Andersen movie where they ride on a train all the time" becomes useful, I see why voice might be preferred over a button, especially in China, where voice search already has a big share of all search.

But I agree. I'd much rather have a dumb tv and upgrade my attached boxes.

> But I agree. I'd much rather have a dumb tv and upgrade my attached boxes.

So you trust the manafacturer/software provider of your attached box more than that of the TV?

I get to choose box separate from screen. Means I can get a box with microphone if I trust the company, or one without if I dont.

What I don't like is paying for smart tv features that don't work or don't trust just to get a screen. Maybe the company that makes great screens make untrustable smart tv features, and maybe the open source smart tv of the future sits in a crappy screen.

More probably the smart tv features gets outdated before the screen.

When it is a machine that I have built myself, running an open operating system, then yes, I do give that much more trust than the TV manufacturer.

>"put on that Wes Andersen movie where they ride on a train all the time"

starts playing 'Resident Evil'

I'm thinking it's quite handy for a quadriplegics, those with learning disabilities, children, elderly, arthritis sufferers, parkinson's sufferers, thalidomide sufferers, disable in general, etc... Just because it's not useful to you doesn't mean that it's not the number one feature in someone else's life.

Sure, I'm not saying it shouldn't exist. If it's a feature you want, then you need to accept the fact that it has to send data to a remote server in order for it to do the thing you want it to do. If you have concerns about privacy, then disable it. I don't see the point of all this hand-wringing about "OMG Samsung is spying on me!". Like there's somebody listening on the other end, waiting for you to start reciting your bank account info in the living room so they can pick it up, you know, like how people sit around in front of their TV doing that, right?

Well, I frequently see the people mentioned above, especially the elderly needing help with their phones or email or online banking, and giving their passwords to a trusted family member thinking that they are the only ones listening. So yea, it happens, a LOT. And big companies like Samsung and Sony never get hacked, right? MITM never happen, that's jus fantasy, right?

If that's a concern, you shouldn't use any online service at all. There's nothing special about voice command that makes it more susceptible to being hacked than every other service that you use every single day. In fact, it's a great deal more difficult, because now the attacker has to have an equally sophisticated voice recognition system at their disposal in order to interpret the intercepted data.

Besides, when they can hack the database itself and get a list of what they know to be passwords, why wouldn't they just do that instead of hacking a bunch of voice snippets and combing through them hoping to maybe find where somebody said a password. It's stupid.

These aren't even logical arguments. I've presented use cases where voice recognition may be the only feasible way to interact. Is a voice recognition server's security as secure as my banks? Seriously, I'm done arguing with someone who fails to rebut logical arguments and just doubles down on a fairly ignorant position.

Just because you don't seem to understand the logic, doesn't mean it's not logical.

Marginalizing an entire group of people simply because it doesn't fit with your narrative isn't logical, it's ignorance.

If you're just going to make up your own things that I said, there's not really much point in me saying anything. How am I marginalizing anyone? I never said anything about voice recognition not existing or not being used. In order for it to work, it needs to transmit data to a server owned by the company providing the service. That's just a simple technological fact. If they're going to use it, they need to weigh what I think is a vanishingly small privacy risk against the benefits they get from the service. Just like you have to do for literally every online service in existence.

People can use the service and come to terms with that fact or they can not use it. If they choose not to use it, they need to live with the consequences of that decision. If it means they have no other way to access some service, then that's something they need to take into account. I can't think of any service in existence that can only be accessed by voice recognition instead of having an phone representative, physical branch, or other options. If you really, really need to access your bank account, there are plenty of options aside from speaking to your smart TV. Nobody is getting marginalized here.

>In fact, it's a great deal more difficult, because now the attacker has to have an equally sophisticated voice recognition system at their disposal in order to interpret the intercepted data.

Or you just pass the data to Siri or Cortana (or whatever microsoft is calling it). Protecting against hacking is defense in depth. If the database is well protected and monitored, attack the target that is not well protected and monitored.

I can think of at least a couple of situations where it could be useful. I've seen some people have an awful time using a remote in a darkened room. Also, free text search is useful when you are trying to search.

> So, disable it. I don't understand everybody's fascination with voice recognition. I don't find it more convenient at all.

If you have small children, it's incredibly handy. Even with Chrome, it's much easier to sneak in a quick click of a button & then say my query than to try to type it all out.

Except its often automatically enabled after software update.