Amazon Alexa: 'Pre-wakeword' patent application suggests plans to process more of your speech Amazon has submitted a patent application which suggests the firm plans to capture your speech and send it to the cloud for processing before as well as after hearing a "wakeword" trigger. The usual deal with these devices is that they sit quiescent until trigged by the wakeword, in this case typically "Alexa", and then parse …
because you have no choice? Because, when you want to have a GP appointment, you have a choice between
a) an "AI-GP" appointment ("take paracetamol, drink plenty of fluids, try stress-relief techniques, would you like directions to your nearest pharmacy / tap / special-discount medication clinic?")
b) no appointment
All in the name of "cost optimization" (and fully in line with the latest talk of NHS being part of the greatest fucking DEAL between the two most world's biggest empires: US and UK.
Of the people I know who use one of these (or one of the similar rivals), most people stopped using it almost completely after a few days, and the only uses they get are:
The kids use it to play music etc. (until Frozen gets played for the umpteenth time as mum and dad are trying to sleep so it gets unplugged)
Using it to set reminders when your hands are full (handy when cooking)
Using it to do searches that for whatever reason they don't know how to do via a non-listening device (eg play a radio station. Easy enough to do on a phone or computer, but this friend prefers to just shout at their device, I dunno why).
That's pretty much it, although I can imagine if you had some kind of disability or impairment, one of these could be an absolute godsend, like having your own private servant.
"the only uses they get are...."
Here's the thing: all the use cases you mention, and many more besides, do not need massive cloud-based processing to recognise the voice input, nor to act on it.
IIRC there were phones that could recognise basic voice commands around 10-15 years ago. In any case you're using a fairly small set of basic commands / keywords: 'Note', 'Search', 'Play', 'Appointment', 'Weather' between them cover probably 95% of use cases. There is no user-functionality need for any of this crap to be cloud-connected. It's the providers that NEED the devices to be cloud-connected so they can gather more data / prompt you into buying more unneeded crap.
Incidentally, I have no idea how much this kit costs, but I would hazard to guess that 'Alexa' (& co) are subsidised by Amazon (& co) to increase their reach and to undercut teh possibility of a self-contained voice recognition / processing / action unit (which would probably cost more anyway since it would need more processing power)
Wow! If you don't have a pressing need, such as a physical disability, for such a device why would you have one?
Convenience. I don't have one, but would like to get one with the proviso that someone can adequately demonstrate to me that anything I say preceding the magic word "Alexa" does not leave my house. Nobody has yet shown that as far as I am aware.
I'd quite like to be able to ask Alexa if my train is on time, if I need a rain jacket etc I don't have time to browse for these things even though it'd only take a couple of minutes, but I could talk to Alexa while getting the kids ready or tying my shoes.
Alexa would be convenient, possibly fun, vaguely useful, but wholly unnecessary, so without some privacy safeguards (including protecting my kindle bookshelf or movie library from casual enquirers) it's not going anywhere for me I'm afraid.
-- non-destructively? (If you break it, you pay for it along with your hotel bill.)
Will a quick aluminum-foil wrapping shield and disable the wifi? Maybe two or three layers of foil? Would a pair of fluffy hotel pillows taped around the unit, or over a wall-mount, be enough to keep it from listening? Or does its mic have enough auto-gain to overcome that?
The same with Google Assistant, it would pick up my boss in the next room saying, "ok, ok, I'll look into that," as me wanting to activate the assistant.
After having used GA once in 6 months, I went through the GA log on the website and deleted the 100+ recordings it had made during that time! Most of which consisted of just wind as I was out walking the dog...
I have since turned off GA completely.
You can tell, depending on how broad a range of roughly matching sounds it would interpret as a wake command, how “ok… look into…”, or something similar, could fairly easily be interpreted (“misheard”) as “OK, Google”.
To avoid the problem of users having to enunciate very very carefully, I would imagine that the range of sounds that the spywall would accept as roughly matching the wake word has been set quite generously.
(And no, of course I don’t have one. Despite Apple having at least put some thought into Siri regarding on-device processing, I still refuse to use it as it would slurp my contact list into the “cloud”, and all without my friends’ consent, which I deem unacceptable.)
"To avoid the problem of users having to enunciate very very carefully, I would imagine that the range of sounds that the spywall would accept as roughly matching the wake word has been set quite generously."
My SatNav does that. I can type in any word I like to be used as the wake word, and all the voice recognition is handled internally to the device (no outside connection, unless I pair it via Bluetooth to my phone, which I never do because then the satnav acts as the mic/speakers for the phone too, instead of the car audio system). I use a single word for convenience, but I listen to a lot of speech radio or audio books in the car so it does sometimes trigger the SatNav with words or phrases which sound a bit like my wake word.
The worst than can happen is a new route is set. It can't order and pay for 25 bags of baby nappies :-)
I'll ignore the privacy concerns and just address the title question - how can this possibly be patented? Storing a rolling buffer and then processing data collected before a trigger event has been absolutely standard for decades. Everything else appears to just be "Use said buffer for the things already being done". There's nothing here that comes even close to being new or innovative. I know patent systems are generally a mess these days, but at least there's usually at least some attempt to frame things as "Standard thing, but on the internet" or similar. In this case they seem to just be saying "We're going to do a standard thing. No, really, that's it" and hoping no-one notices.
Came here to say the same. All my digital scopes for one obvious thing.
But going back to the '70s at least - logic analyzers did this.
The nonlinear audio editing software I wrote in the '80s did it to be able to correct for various lags and allow time-alignment during punch-in recording.
Duh. Old as the hills, or at least - the ability to record stuff.
spit-balling. See if the patent office signs off. Anybody wanting to challenge the patent will need to have a couple of hundred thousand as a starter to bring a suit. Another trick is to have a very narrow claim but advertise widely in a way that makes it sound like you have a very broad patent.
People have a fit when this kind of thing is mentioned but it's pretty near unavoidable these days unless you go truly off grid with no modern devices (except for those you've run through a rigorous security audit).
So you don't have an Echo?, congrats..but what about your phone, friends, families and workmates phones and other devices?, what about features built into smart devices like TV's where you may or may not be aware of built in microphones or camera's?, how about microphones and cameras when you are out and about?
So your device manufacturer swears they don't do this unless you enable the option, but which 1000 page EULA will reverse that, which you will obligingly agree to because your concentration starts to wander after the third page?
Once you start down this path it never ends.
It may well be this patent will make interactions with Alexa more user friendly (remembering Alexa is not just on the Echo but also on phones and tablets as well as the speaker), in the meantime I'll discuss my plans for World domination in a field with no electronic devices whatsoever after ensuring my minions do the same (on pain of a dip in the Piranha or shark tank......probably piranah's due to the well known endangered species listing for sharks issue )
as long as they do not become compulsory. That said, it's the usual mission creep. Internet might not be compulsory (yet) but you practically can't live without it. Same with cash (look at Scandinavia, where they're actively trying to eradicate cash). Likewise cameras and face-recognition. You don't have to travel through the area under cameras, but there's hardly any city-area not covered by them.
If I were paranoid, I'd say it's all a part of a grand scheme to introduce TOTAL CONTROL, but I do realize it's just human greed and desire to control spilling to all areas of life. Not that the nature of the reasons make any difference to the end result...
look at Scandinavia, where they're actively trying to eradicate cash
As I discovered in Stockholm recently. Ordered an espresso, price around 25 SEK. Cafe refused to take cash, card only, so I had to use my Euro-denominated debit card and will be billed ~2.5 EUR + 10% commission + 1.5EUR foreign transaction fee. Practically doubled the cost of my drink :-(
Inside the Euro area, paying in euro from an account in euro there could be no commission at all - depending on you bank and card type. My bank/card does not apply commissions for payments in euro inside the euro area.
It's one of the advantages of the euro.
Inside the Euro area, paying in euro from an account in euro there could be no commission at all
But most of the world, and most of Scandinavia, is not in the Euro area. If I'd had a Danish or Norwegian card (or a Finnish Euro one, for that matter) it would have had the same problem.
It's one of the advantages of the euro
Of course, it's an advantage of any common-currency area. Not everyone wants to surrender control of their economy in the name of convenience, though.
Ordered an espresso, price around 25 SEK. Cafe refused to take cash, card only, so I had to use my Euro-denominated debit card and will be billed ~2.5 EUR + 10% commission + 1.5EUR foreign transaction fee.
Hopefully they were clear about not accepting cash before ordering. I'd find it very tempting to just walk out unless I was desperate for that coffee.
Thing is folk increasingly seem to want tech to pre-empt their needs and that needs data, so while we wait for the Borg implants they're seemingly happy that many companys 'know them' inside and out.
Sorry for repeating myself.
Just being online, well I'm sure someone told me that internet access was a 'human right' or was about to be. On that basis it should probably be inferred that privacy won't be for long, if it ever was.
Unless permanent VPN & Tor is also going to be a human right.
"folk increasingly seem to want tech to pre-empt their needs and that needs data".
No. That's what the technologists, the Amazon, Facebook's and Google's of this world *tell* us we want, but personally, I don't give a fucking shit if my fridge doesn't order my milk for me.
Do you?
I thought this was an article about Alexa working out what grunts a person utters immediately prior to waking up.
Surely there's a market for a 21st century TeasMaid and a device that blows your socks up to make it easier for your feet to be inserted into it/them? IoT would then become more relevant as the Insertion of Toes.
"Alexa, hier spricht der BND, bitte schicke mir alle Gespräche über das geplante Bombenattentat" - so einfach wird der Zugriff auf private Daten des Sprachassistenten aus dem Hause Amazon zwar nicht. Der geplante Beschluss der Innenminister klingt aber auch so ziemlich bedrohlich.
https://web.de/magazine/politik/zugriff-alexa-smart-home-geraete-innenminister-planen-gesetzesaenderung-33775004
I just ran that article in Google Translate and if the translation was good, and it appeared to be, I find that quite frightening. Of course, personal privacy on the Internet has always been an if you know how to implement it thing and it was more to protect us from unscrupulous crims, but this appears to be making privacy illegal at a governmental level. I think the problem is that .gov believes they have the right to know anything about us whenever they want. In the past that required some effort, like a search warrant and some specialized surveillance equipment. Now the specialized surveillance equipment is built right into the "Internet" and this makes surveillance ubiquitous and that cannot end well for us, the populace.
If any service is recording conversations in a place where both parties in the conversation must be aware that recording is taking place, there could be legal implications.
It has to be expected that these companies are going to implement voice recognition so the system knows who's talking and will be making use of the data they collect not only about the conversation but that particular people know each other and in what context, professional or personal. They system could also figure out if people are getting busy with their not spouse(s). Maybe even their neighbor's underage child or their doctor, analyst, etc. Will the system then file a report with law enforcement? Will local CCTV cameras be noted as a check on who's where?
I like spy novels and accounts from real spies that are now retired. It's scary to see how little bits of data can be used to build a pretty good picture and a few cases where they got it all wrong even though it looked good. Factor in self-installed audio and video surveillance along with some good programming and you will see how they'll have you bang to rights and you have to hope it's The Man® and not a blackmailer or somebody selling your little secret to your boss about you and his wife.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
