Hacker News Reader

Comment by imgabe | Hacker News Reader

Isn't that exactly how Google's voice recognition works? And Siri's too I assume? Your phone doesn't have the processing power to do speech-to-text on the fly, so it sends it to a server to do it. I don't see a vast government conspiracy behind this. If they can make the TV a bit cheaper by putting in a less powerful processor and taking advantage of the always present high speed Internet connection, of course they would do that.

View on HN

> Your phone doesn't have the processing power to do speech-to-text on the fly, so it sends it to a server to do it.

This is what I don't get. I used to play with Microsoft's SpeechAPI back in 2007 and it was pretty decent at real-time speech recognition - almost perfect, if you limited your program to a pre-defined command grammar. It was all done completely off-line, real-time on a machine that had much less processing power than your average smartphone of today. I don't see anything inherent in speech recognition that would require so much processing power that the phone can't handle it real-time. The only reason I see is that continuous moving of everything to cloud because it's easier to capture revenue stream that way.

While I agree that a component is revenue, I also think the "big data" factor of it is important. All of those usage stats (Where/when/what was said/) would definitely help to shape future products and integrations. (50% of people want to order pizza, maybe we should integrate some pizza thing vs. the 5% that want to do advanced math, etc)

EDIT: I'm also one of those guys who's like "no I'm not signing up to X service because they don't need more data about me" but I get the double-edged sword and how more of the right data makes certain decisions easier.

I couldn't agree more with your comment. Maybe there is a way to collect data about user only after they have been anonymized, so the data would contain only what user_xxx does, with no way to map user_xxx to his real identity.

I don't think it's trivial, but I hope to see some regulations coming in the way the data are collected and stored.

The point is that it is a choice, a deliberate choice. Without sending the date, i.e., your output, to corporate servers, they can't profit from your person. Hate to say this and many may believe that it's over the top, but humans are ever increasingly becoming nothing more than a commodity. Kind of like a cow on a ranch. You monitor it and groom it and prepare its feed in accordance with stages of development and tweak your inputs all in order to squeeze the optimal amount of profit out of the cow....I mean human.

Unfortunately, even YCombinator, just like a vast majority of tech companies, are guilty of supporting this notion. There is no respect or regard for privacy and ownership of your person. if it escapes your flesh and bones, it is deemed fair game to harvest by anyone. You and your actions and your speech and your image and your location are a commodity to be optimized for profit. Say moo, mothertrucker.

Uh, this is just a total misunderstanding of how and why these systems work the way they do.

Of course command grammars are recognized on the device, otherwise you wouldn't be able to turn your TV on using a command or use basic functions in a speedy manner.

And, of course they send audio data to a server -- because your TV does not have a complete language model for up to date web search or even tv guide installed a priori. i.e. good reasons that have nothing to do with "capturing revenue stream".

Sigh...

I wonder when we'll get commercials that run voice commands on those TV's themselves...

'TV Install Samsung Super Social Awesome App, Yes, Yes Yes'

If it is anything like the last time this happened (google 1-900 santa) the commercial will be pulled in under 10 minutes.

Wow, this happened already? I only remember an ad recently that turned on people's Xbox Ones. http://www.bbc.com/news/technology-27827545

Yes, phones do this too. But the massive difference is that you have to consciously bring up Siri and you know when it's listening and when it's not. The TV appears to be constantly listening for audio commands and is sending whatever it hears upstream for processing. Someone in these comments said that you can turn the TV on through a voice command, which means that it is listening even when it is off and merely plugged in.

It's not sending the audio continuously, in the same way that Android phones with always-listening features ("Ok Google") aren't. There is dedicated keyword-detecting hardware.

Isn't Siri listening by default now in iOS 8 whenever the phone is plugged in?

I don't think it's on by default. I believe I had to turn it on in the settings after upgrading. In any case, it's not streaming everything it hears. The "Hey Siri" detection is done locally, even though the subsequent voice-to-text is not.

Even assuming that the lack of processing power on the TV is a correct theory, your comment misses the point here.

The issue is not about the intent of the manufacturer. Noble intent or not, the story is about what is enabled by this device. Any time you rely on noble intent of a 3rd party you are putting a lot of trust in them. Especially when those third parties can monetize your data without you ever knowing about it.

Technology provides the tools for mass surveilance and globally undermining trust, but it also provides tools for removing the need for trust, and apparently this smart TV adopts a "just trust us" approach.

From my understanding, these systems send voice to servers to recognize queries, but they detect the prompting signal ("OK Google" or "Hey, Siri") entirely on the device, and both systems make a "Listening now" sound when they switch from one mode to the other. So, presuming that software is working as described, when you speak near an Apple/Android phone, your speech isn't being sent to the cloud in most cases.

Among other things, running the radios all the time to send audio to the cloud would be a battery-killer.

https://support.google.com/websearch/answer/2839743?hl=en

http://stackoverflow.com/questions/17616994/offline-speech-r...

Sadly, it appears that Google voice recognition follows yuor network connection, so you can't keep it in offline-recognition mode unless you turn off networking when talking.

Except for Siri (and presumably Google), you have to actively press a button and trigger it to listen. It's not constantly listening and uploading.

Almost. The phone can do the TTS on the fly, but it can't calibrate itself without server-aggregated data. So it needs to send off SOME voice captures to central in order to work at peak performance.

Great, so have the user explicitly put it in training mode.

Genuine question, that makes all the difference in the world for me: Does the TV require me to push a button to activate the voice commands?

There is a button on the remote to activate voice.

"There is a button on the remote to activate voice."

More likely: There is a button on the remote to make users believe they have the ability to disable voice data relay to someone's remote servers.

What are the chances such TVs will have network controlled power-up capability? You only think you turned it off.

It's funny that people seem to be accepting always-on GSM network connectivity of the electronics in control of their cars, but oh my god, not the TV!

The TV can't drive you at 100MPH into a palm tree on remote command.

Just to be clear, I think 'smart TVs with network connected microphones' is a completely insane idea. Another on a long list of hanging offenses.

> oh my god, not the TV!

Dude. I stare at my TV. If it stares back, that's scary! And if it's listening to me!?!? All the time?

The messed up part is that ALL of our devices will soon be watching and listening to us. Our thermostats, our smoke detectors, our TVs, our phones ... I just installed a fancy internet connected lock on my door. In fact, I installed a fancy electronic toilet seat too... oh well.

DEA: this household hasn't had a decent bowel-movement in a week and they have no active prescription for narcotics, deploy SWAT.

> It's funny that people seem to be accepting always-on GSM network connectivity of the electronics in control of their cars

Shit, I'm not. OnStar freaks me the fuck out; one of many reasons I'll never buy a GM car.

Reply on news.ycombinator.com