One more reason to never buy a product requiring persistent internet connectivity with automatic updates. When physical products are treated the same as software as a service, you don't own a damn thing.
And it used to be so good. Not being ironic. For example, coca-cola patented their formula. It didn't prevent other companies from making their own "colas", like Pepsi. They just couldn't make it using the exact coca-cola formula. Which is fair, I'd say. You have to create your own way to achieve that result.
The equivalent in software should be "you can replicate the idea, but not the code behind it". But somehow that became "you can't replicate the idea at all", which is super broken.
And Coca-Cola has an ingredient benefit - the fact they have a license to put coca into the drinks is completely unique in the industry AFAIK. It might be troublesome for a competitor to get that same license.
Technically, Coca-Cola puts in a coca leaf extract that is produced by Stepan Company. They are the ones who have the only license to import coca leaves in the US and they use them to make medical cocaine. The extract they sell to Coca-Cola comes after all the good stuff is gone. Coca-Cola never touches the leaves.
I was mentally preparing myself to agree with Sonos on something, but apparently the patent in question is for being able to adjust the volume on a group of speakers instead of having to adjust each speaker individually. It seems ridiculous that this is patentable at all. It seems like the sort of usability feature that would take five minutes to design and perhaps get implemented before lunch. Am I missing something here?
And all 5 of them are about as horrible as the speaker volume patent. For instance one covers the common iot task of joining a device to a wifi network by identifying the device initially via Bluetooth, sending the device the wifi credentials and then verifying it has joined the wifi network. The difference being the device is a speaker.
> And all 5 of them are about as horrible as the speaker volume patent.
I'm not defending any of these, but in general all Software patents are like that. Most of them are obvious solutions to the problem one is working on. It's just that one person/team works on a problem before another, comes at one of the obvious solutions and gets to file the patent.
Due to this a large number of software engineers hate software patents.
Edit: having said that, I will also refer to @samwillis's comment in this thread, https://news.ycombinator.com/item?id=29841653, as to why some of this may not be as obvious as it seems on first look. And, the fact that Google had partnered with Sonos (again per that comments).
Waaay back, early in my career, I was categorically told never to read patents.
IMNAL, but per my understanding, there is a major difference between patent infringement and willful patent infringement. Which is why a soft engg should never read patents, at least not in the areas one may work at some point in their career.
This comment thread is very quickly going to become a discussion about software patents in general, however in this case it worth also looking at the history of the case. My understanding is that Google partnered with Sonos back in 2013 to bring Google Play to Sonos speakers, during that partnership Google basically milked Sonos for info on how their platform worked in order to compete with them.
I suspect that the outcome of this where Google devices have to have volume controlled individually isn't due to a patent on that specific functionality, but due to something with the inter-speaker communication system that Sonos pioneered (multiple wireless speakers playing sound in sync with each other, with no audible delay). Google will have had to change how they have speakers talk to each other and the side effect is the lack of group volume control.
A patent on "speaker group volume control" I agree would be silly, but a parent on a wireless control system and protocol for multiple speakers that can be grouped together is far less so.
I wouldn't classify myself as someone particularly "skilled in the art", though I've definitely built audio DSP solutions. I can think of 2-3 ways to achieve "wireless control system and protocol for multiple speakers that can be grouped" while just sitting here.
In classic infringement avoidance ridiculousness, I have no desire to dig into the details... but this superficially feels like it would fail the obviousness test for patentability.
I think the true innovation and problem that Sonos solved when they started was to have synchronised compressed audio streams play on multiple speakers with do audible delay between them. That is not a partially "easy" problem to solve, and bear in mind that when they started they had to implement their own wifi network for the speakers as the wifi standards of the time (from memory) both wouldn't provide the necessary bandwidth but also few people had wifi at the time. The were no smart phones when they stated, they had their own "smart" remotes even.
Sonos aren’t popular here at HN due to some truely awful decisions they have made.
I really like my small system and it works great - mostly.
You hit some seriously rough edges if you have a Unifi network, and Sonos’s use of STP rather than RSTP is system breaking until you realise it. And even when you solve this issue, there will be random outbreaks of dysfunction and failure to connect.
The Sonos app is terrible, and the integration with Spotify occasionally craps out - logging out and then in again can fix it, sometimes.
You are quite right, it does have rough edges and certainly more than 10 years ago when we started buying into their system (we have four rooms covered by 8 speakers now).
The Spotify bug when it randomly cuts our is REALY annoying, it seems to happen in a cluster about once a week.
Fortunately we weren't effected by their incredably ill conceived changes to the "legacy" speakers, they deserve criticism for that and I'm sure I would be less charitable if we had been effected.
Have each speaker transmit a little packet with a timestamp 10 times per second to each other. Or relative to some designated central point. Each speaker will have an internal clock but use the central clock tick to keep exactly synchronized.
Having synchronized their timestamps, the central hub transmits a specific time for playback to begin. At this point you need to make sure that there isn't any unexpected buffering going on, output buffers have to be tiny/realtime or the process of sending signal out has to be again synced to the clock, which means you need control over the firmware/drivers. Worst case you might have to skip or repeat a couple samples once in a while to keep synced up which in a 44000 samples per second signal is not a big deal.
You need a big enough input buffer to play a few seconds of audio, just to be safe, but the output buffer has to be realtime.
That's pretty much how I'd do it, off the top of my head.
> to have synchronised compressed audio streams play on multiple speakers with do audible delay between them. That is not a partially "easy" problem to solve,
Without commenting on the radio side of this, I implemented automatic synchronisation of video playback units to within a few microseconds in 2007 and it was fairly obvious how to do it back then. That's accurate enough for audio as well.
So long as units were playing the same content or equivalent-length content (for video walls, shows etc.) and could receive each other's broadcasts on whatever LAN they were connected to, they would automatically detect each other, form equivalence groups, and adjust timing until the clocks matched each other, and stay locked together across content changes.
There was no master control unit, and no configuration; it was an automatic distributed algorithm that "just worked".
> suspect that the outcome of this where Google devices have to have volume controlled individually isn't due to a patent on that specific functionality
I couldn't find the complete list of patent infringements (Sonos alleges more than 100) but it does seem there's one in particular about controlling volume on multiple speakers / rooms at once.
This seems plausible as the group-volume feature isn't unique to Sonos (and predates them), however their specific means of implementing it likely is unique and patentable.
The patent in question is not about the specific manner in which Sonos implements group-volume. It in fact doesn't even state how Sonos implements it. It is purely about the concept of centrally controlling the volume of local area network connected distributed speakers using a centralized control unit.
Then it should be unpatentable - you should be able to patent a novel technical solution that's necessary for a desirable feature, but not the concept of a feature; patents are for "how to achieve X" not for the outcome.
You aren't supposed to be able to patent an idea or concept but that is what software patents all too often do which is one of the reasons they are so awful. For instance this patent covers having a controller which presents a UI to raise, lower and mute/unmute the volume of a group of network attached speakers. When the user selects to do one of those things the controller communicates over the network to the group of speakers to raise, lower or mute/unmute their volume. The specific technology which implements this ability, i.e. how the controller communicates with the speakers, is not part of the patent. It is purely on having a controller which does the above however it may achieve it.
Edit: Thought I would point out that when Sonos first filed their claims they made a big noise about Google using the same protocol as they to do to control their speakers. This protocol isn't covered by the patents in question. The reason is that the protocols in question are UPnP standard protocols.
> Google home speakers are about to get slightly worse
Google are well ahead of this lawsuit, the home speakers have been getting slightly worse since Google lost interest and moved on a couple of years ago.
Say what you will about Alexa, but it is the only voice assistant that seemingly keeps improving, Siri is statically-terrible and while Google Assistant was once good, it is worse now than then somehow (although functionality/integrations hasn't evolved).
My favorite thing right now is that a room can have multiple devices (e.g. screens, speakers, etc) but timers don't belong to a room, they belong to an individual Home device, so your screens don't show your timers if your speaker "eats" them. Makes zero sense. Been like that for over a year.
In general Google as an organization is going downhill fast. Google Search is legitimately bad now, I turn on Verbatim Mode and quote everything just to get a so-so experience (still mostly spam though).
I threw all my google home devices into the recycling bin a year ago. They had become comically bad. The only thing I haven't replaced are the nest thermostats but I'd like to get rid of them as well.
Just casually wondering here... how do you evaluate the voice assistants to claim that Alexa is improving, Google Assistant is getting worse and Siri is terrible? I used all three and honestly I don't feel a whole lot of difference between them. They all control the lights and set the timers/alarms just alright, with similar "error rates".
In my experience they are only good for settings timers/alarms, that's it (though I believe can be good for controlling lights as well).
Few years ago I was in indonesia, Bali and asked google 'navigate to ubud' and it showed me how to navigate to some cities in USA. Same with translation when you want to have simple conversation with locals - each time ends up with embarrassment.
I thought calling it "statically-terrible" was unfair because Siri is getting much worse. Yesterday I discovered that asking it to play "Bring the Noise by Public Enemy" plays a dreadful Benny Benassi remix, even if you specifically ask for the album version. Apple's last big keynote led with some marketing of a voice only plan for Apple Music, and yet Apple Music / Siri has no concept of the canonical version of a song.
My favourite though is asking it to "play Richard the third by Supergrass", Siri responds that it can't find that song on Apple Music, so you ask it to "play Richard eye eye eye by Supergrass" and Siri responds "Now playing Richard the third by Supergrass".
I'm with you. I bought a Homepod and thought I would make use of Siri for timers, to-do items, grocery lists, and other similar tasks in addition to music. However, Siri is so comically useless for music that I don't even bother. It's significantly faster to just pull the song/album up on my phone and tell it to play on the Homepod.
The only things I use Siri for now is to add ONE item to the grocery or set a timer. The timers are useful in the kitchen, but it's hardly worth a Homepad to set a timer.
I wish I had returned it within the return window. At least it sounds great.
Apple Music is terrible with no help from Siri.
Say, I listen to Dva, the czech duo.
Apparently it is the same artist as some rapper Dva.
I wonder who gets the money.
Don’t get me started on “various artists”.
Or search.
Google Home Mini connected to YouTube Music also suffers similar problems and I swear it's ability to determine the voice commands given has gotten far worse than it was a few years ago.
For example, asking for specific albums and it will play someone's playlist of the album - in terrible quality because someone uploaded it as a video to YouTube I think part of the problem is that the music catalogue is so poorly maintained/governed on YouTube Music. For example there are many songs that state they are by Little Richard that are actually other artists' songs
- company builds a profile on your likes and dislikes based on commands you give
- company surreptitiously builds a profile on everything it hears, ostensibly for "training" their ML / AI
- data is viewed by snoopers within the company
- data on public figures is leaked to media / public
- data is sold to other companies you do business with
- data is used to build a "social credit" profile. Say something casuallu offensive? Forget getting that loan!
- data is accidentally leaked. Organized theft rings now know your routine, and the routines of everyone in your city all at once.
Then there are the odd stories of people who are expecting a baby suddenly getting ads for parenting supplies, or gift surprises being spoiled by recommendations, etc.
At the end of the day, FAANG are not my friends. They don't get to know things about me for free, and rarely are the goods they offer worth what they get in exchange.
I make some exceptions, but voice assistants aren't one of them.
> - company surreptitiously builds a profile on everything it hears, ostensibly for "training" their ML / AI
I have yet to see any evidence that any voice-command device is transmitting everything it hears. The data sent by the device is too small to contain audio beyond what was said after the wake word was detected. The hardware does not have the necessary chips to do 24x7 voice to text.
You're totally right that they can do a lot with what they do get, from the audio around when you say the wake word, and from other data that you give them through web surfing, location data, etc., and that's totally a legitimate reason to not have one or to limit one's interaction.
As your article states, the wake word (and those phonetically similar to it) may activate the voice recording. But the device isn't recording audio 24/7. Presumably other voice assistants like Siri and Google have made advancements in activation hardware/software/algorithms in the 4 years since that article came out.
Maybe it's just my accent, but whenever I'm around someone else's device (Alexa most recently) wakes up multiple times and responds to something we didn't say.
The worst part is, when we tell it to stop (potentially flagging it as incorrect), how do we know that conversation isn't being analyzed to "improve" the model? We never used the wake word, so we don't know how long it was listening.
It's hard for me to think of them as anything other than creepy. I want to like them; I like the idea of being able to have the convenience of vocal interactions with technology. But, until I'm convinced that it is entirely on my own terms, that's not going to happen.
Why is there an assumption I have a voice activated assistant in my pocket or on my wrist? And certainly, those options come with a "push to talk" feature.
I feel like the title is misleading. The speakers are getting worse because Google is too greedy to pay for the patent that they have been using without paying until now.
Whether that patent is useful is a different question, but the headline makes it sound like Sonos is the aggressor here. But Google chose to do things this way, and Sonos then reacted.
71 comments
[ 3.5 ms ] story [ 123 ms ] threadOur patent system is so broken.
The equivalent in software should be "you can replicate the idea, but not the code behind it". But somehow that became "you can't replicate the idea at all", which is super broken.
same with WD40/KFC/...
IMHO Patents are unethical. Patents for IP are an atrocity and Software Patents are a crime against humanity.
It seems there are total 5 parents. The multi zone speaker volume one is just one of them.
https://www.theverge.com/2022/1/6/22871121/sonos-google-pate...
I'm not defending any of these, but in general all Software patents are like that. Most of them are obvious solutions to the problem one is working on. It's just that one person/team works on a problem before another, comes at one of the obvious solutions and gets to file the patent.
Due to this a large number of software engineers hate software patents.
Edit: having said that, I will also refer to @samwillis's comment in this thread, https://news.ycombinator.com/item?id=29841653, as to why some of this may not be as obvious as it seems on first look. And, the fact that Google had partnered with Sonos (again per that comments).
IMNAL, but per my understanding, there is a major difference between patent infringement and willful patent infringement. Which is why a soft engg should never read patents, at least not in the areas one may work at some point in their career.
I suspect that the outcome of this where Google devices have to have volume controlled individually isn't due to a patent on that specific functionality, but due to something with the inter-speaker communication system that Sonos pioneered (multiple wireless speakers playing sound in sync with each other, with no audible delay). Google will have had to change how they have speakers talk to each other and the side effect is the lack of group volume control.
A patent on "speaker group volume control" I agree would be silly, but a parent on a wireless control system and protocol for multiple speakers that can be grouped together is far less so.
In classic infringement avoidance ridiculousness, I have no desire to dig into the details... but this superficially feels like it would fail the obviousness test for patentability.
I think the true innovation and problem that Sonos solved when they started was to have synchronised compressed audio streams play on multiple speakers with do audible delay between them. That is not a partially "easy" problem to solve, and bear in mind that when they started they had to implement their own wifi network for the speakers as the wifi standards of the time (from memory) both wouldn't provide the necessary bandwidth but also few people had wifi at the time. The were no smart phones when they stated, they had their own "smart" remotes even.
Sonos aren’t popular here at HN due to some truely awful decisions they have made.
I really like my small system and it works great - mostly.
You hit some seriously rough edges if you have a Unifi network, and Sonos’s use of STP rather than RSTP is system breaking until you realise it. And even when you solve this issue, there will be random outbreaks of dysfunction and failure to connect.
The Sonos app is terrible, and the integration with Spotify occasionally craps out - logging out and then in again can fix it, sometimes.
I like the system but it’s got a way to go.
The Spotify bug when it randomly cuts our is REALY annoying, it seems to happen in a cluster about once a week.
Fortunately we weren't effected by their incredably ill conceived changes to the "legacy" speakers, they deserve criticism for that and I'm sure I would be less charitable if we had been effected.
I'm curious if we could re-invent it, right here in this thread, without looking up how it's done.
Having synchronized their timestamps, the central hub transmits a specific time for playback to begin. At this point you need to make sure that there isn't any unexpected buffering going on, output buffers have to be tiny/realtime or the process of sending signal out has to be again synced to the clock, which means you need control over the firmware/drivers. Worst case you might have to skip or repeat a couple samples once in a while to keep synced up which in a 44000 samples per second signal is not a big deal.
You need a big enough input buffer to play a few seconds of audio, just to be safe, but the output buffer has to be realtime.
That's pretty much how I'd do it, off the top of my head.
Without commenting on the radio side of this, I implemented automatic synchronisation of video playback units to within a few microseconds in 2007 and it was fairly obvious how to do it back then. That's accurate enough for audio as well.
So long as units were playing the same content or equivalent-length content (for video walls, shows etc.) and could receive each other's broadcasts on whatever LAN they were connected to, they would automatically detect each other, form equivalence groups, and adjust timing until the clocks matched each other, and stay locked together across content changes.
There was no master control unit, and no configuration; it was an automatic distributed algorithm that "just worked".
I couldn't find the complete list of patent infringements (Sonos alleges more than 100) but it does seem there's one in particular about controlling volume on multiple speakers / rooms at once.
https://www.theverge.com/2022/1/6/22871121/sonos-google-pate...
Edit: Thought I would point out that when Sonos first filed their claims they made a big noise about Google using the same protocol as they to do to control their speakers. This protocol isn't covered by the patents in question. The reason is that the protocols in question are UPnP standard protocols.
I use Home Assistant a lot and I basically had to learn yet another shitty programming language implemented in yaml.
You can use node-red for automation. Frontend stuff is still yaml but thats not rocket science.
The great advantage of HA is that it can be made to work. This sounds sarcastic but is a huge advantage compared to other systems, apps and devices
Google are well ahead of this lawsuit, the home speakers have been getting slightly worse since Google lost interest and moved on a couple of years ago.
Say what you will about Alexa, but it is the only voice assistant that seemingly keeps improving, Siri is statically-terrible and while Google Assistant was once good, it is worse now than then somehow (although functionality/integrations hasn't evolved).
My favorite thing right now is that a room can have multiple devices (e.g. screens, speakers, etc) but timers don't belong to a room, they belong to an individual Home device, so your screens don't show your timers if your speaker "eats" them. Makes zero sense. Been like that for over a year.
In general Google as an organization is going downhill fast. Google Search is legitimately bad now, I turn on Verbatim Mode and quote everything just to get a so-so experience (still mostly spam though).
Few years ago I was in indonesia, Bali and asked google 'navigate to ubud' and it showed me how to navigate to some cities in USA. Same with translation when you want to have simple conversation with locals - each time ends up with embarrassment.
I don’t think it’s a fair assessment. Sure it’s the worst out of the 3 but it’s improving, albeit at the speed of 0.3 microgoogles.
My favourite though is asking it to "play Richard the third by Supergrass", Siri responds that it can't find that song on Apple Music, so you ask it to "play Richard eye eye eye by Supergrass" and Siri responds "Now playing Richard the third by Supergrass".
The only things I use Siri for now is to add ONE item to the grocery or set a timer. The timers are useful in the kitchen, but it's hardly worth a Homepad to set a timer.
I wish I had returned it within the return window. At least it sounds great.
For example, asking for specific albums and it will play someone's playlist of the album - in terrible quality because someone uploaded it as a video to YouTube I think part of the problem is that the music catalogue is so poorly maintained/governed on YouTube Music. For example there are many songs that state they are by Little Richard that are actually other artists' songs
"Alexa, play soothing lullaby music"
"Sure, and don't forget that next week is prime week. Get great deals on all your favorite items....."
- company surreptitiously builds a profile on everything it hears, ostensibly for "training" their ML / AI
- data is viewed by snoopers within the company
- data on public figures is leaked to media / public
- data is sold to other companies you do business with
- data is used to build a "social credit" profile. Say something casuallu offensive? Forget getting that loan!
- data is accidentally leaked. Organized theft rings now know your routine, and the routines of everyone in your city all at once.
Then there are the odd stories of people who are expecting a baby suddenly getting ads for parenting supplies, or gift surprises being spoiled by recommendations, etc.
At the end of the day, FAANG are not my friends. They don't get to know things about me for free, and rarely are the goods they offer worth what they get in exchange.
I make some exceptions, but voice assistants aren't one of them.
I have yet to see any evidence that any voice-command device is transmitting everything it hears. The data sent by the device is too small to contain audio beyond what was said after the wake word was detected. The hardware does not have the necessary chips to do 24x7 voice to text.
You're totally right that they can do a lot with what they do get, from the audio around when you say the wake word, and from other data that you give them through web surfing, location data, etc., and that's totally a legitimate reason to not have one or to limit one's interaction.
The worst part is, when we tell it to stop (potentially flagging it as incorrect), how do we know that conversation isn't being analyzed to "improve" the model? We never used the wake word, so we don't know how long it was listening.
It's hard for me to think of them as anything other than creepy. I want to like them; I like the idea of being able to have the convenience of vocal interactions with technology. But, until I'm convinced that it is entirely on my own terms, that's not going to happen.
That being said, I do unplug my Google Home Minis when they're not in use.
"Google Home speakers are about to get slightly worse because Google infringed on existing patents and was caught."
https://github.com/badaix/snapcast
Whether that patent is useful is a different question, but the headline makes it sound like Sonos is the aggressor here. But Google chose to do things this way, and Sonos then reacted.