Smart Assistant & Smart Speaker thread - "Google Assistant just beat the Turing test"

Moggy wrote:

Green Gecko wrote:My mum used to have an alarm clock that also boiled water to make tea. Now that's a smart device for you, although probably not particularly safe. And that was in the early 90s.

Teasmade. They’ve been around for a very long time.

https://en.m.wikipedia.org/wiki/Teasmade

Tafdolphin wrote:https://www.cnet.com/news/google-uses-humans-to-review-google-assistant-queries-report-says/

A contractor provided Belgium-based VRT NWS with more than a thousand recordings of people using Google Home and the Google Assistant app to answer questions and conduct searches, the news outlet reported Thursday. There were also recordings from when the assistant was accidentally triggered.

The Google contractor says the job entails listening to the recordings to check the accuracy of the assistant listening to users speaking in Dutch. Contractors review the script of the recorded query and confirm it matches with the voice recording. There are reportedly thousands of Google employees worldwide listening to audio excerpts via a secure login in the company's Crowdsource app.

I meet a lot of tech folks due to the circles my wife moves in, specifically a lot of academics concerned with privacy and digital rights. One of the foundational principles of every one of them is "Don't let a smart speaker get anywhere near your home."

Hang on, a contractor totally broke confidentiality and numerous international privacy laws by providing private information to a newspaper?

And that's news? What a giant bellend.

These things are controversial, sure, but read the policy. If technology isn't by itself able to verify the accuracy of text translations of voice queries then what else is going to be used? Parrots? I totally get the privacy concerns around these devices, I really do, but what narks me off is the deliberate stupidity surrounding willingly entering into a contract to process your voice data with, well, a device that processes your voice data and translates it into search queries.

If that upsets anyone, don't buy one. It's like people blaming the Internet for wasting their time. Don't go on the Internet. Don't eat overly sweet foods and exercise regularly if you don't want to be fat. Don't use a Tesco clubcard if you don't want your spending habits to be tracked. And on and on and on.

my smart computer professional brain says never ever do this

my other brain says I want to talk to the lightbulb in the bathroom from the kitchen

Same. It really helps me with my disability though. I'm far more effective at processing information and acting via my own "voice commands". A voice assistant just helps me remind myself, call people, and do other things "in the moment" - because I will literally forget in 3-4 seconds if I don't do it right away.

A computer or other interface adds friction to that process. If I don't want the smart speaker to be listening in on my stuff (although it does that offline, and is not sending everything it hears ever to the Internet), I unplug it. I unplug it every time I leave the room and it's never on at night.

You can use some of these technologies without being absolutely stupid about how they work, what their risks and limitations are. Like anything else. I can use a circular saw that is absolutely capable of dismembering my limbs in seconds. I can store all of my currency in a bank that is capable of going broke and never returning my money. I can entrust my teeth a private dentist and hope that they don't slip one day and remove my jaw. Anyway, no need being more facetious.

I don't think it's too dissimilar to someone using their debit card at a coffee shop, but refusing to pay for anything online, even though the infrastructure is largely the same.

Also, phones have had this capability to process voice data digitally for literally decades. Why is it suddenly a big deal because you can put it in a little box? People have their smartphone on 24/7 and "fragment" their privacy into hundreds of apps with dodgy as strawberry float permissions. Nobody cares because they want x or y convenient feature and they go on with their lives. Perhaps I compensate by not using a smartphone, because they really are designed to consume your time like nothing else in history.

Green Gecko wrote:

Tafdolphin wrote:https://www.cnet.com/news/google-uses-humans-to-review-google-assistant-queries-report-says/

A contractor provided Belgium-based VRT NWS with more than a thousand recordings of people using Google Home and the Google Assistant app to answer questions and conduct searches, the news outlet reported Thursday. There were also recordings from when the assistant was accidentally triggered.

The Google contractor says the job entails listening to the recordings to check the accuracy of the assistant listening to users speaking in Dutch. Contractors review the script of the recorded query and confirm it matches with the voice recording. There are reportedly thousands of Google employees worldwide listening to audio excerpts via a secure login in the company's Crowdsource app.

I meet a lot of tech folks due to the circles my wife moves in, specifically a lot of academics concerned with privacy and digital rights. One of the foundational principles of every one of them is "Don't let a smart speaker get anywhere near your home."

Hang on, a contractor totally broke confidentiality and numerous international privacy laws by providing private information to a newspaper?

And that's news? What a giant bellend.

These things are controversial, sure, but read the policy. If technology isn't by itself able to verify the accuracy of text translations of voice queries then what else is going to be used? Parrots? I totally get the privacy concerns around these devices, I really do, but what narks me off is the deliberate stupidity surrounding willingly entering into a contract to process your voice data with, well, a device that processes your voice data and translates it into search queries.

If that upsets anyone, don't buy one. It's like people blaming the Internet for wasting their time. Don't go on the Internet. Don't eat overly sweet foods and exercise regularly if you don't want to be fat. Don't use a Tesco clubcard if you don't want your spending habits to be tracked. And on and on and on.

That's...a viewpoint. I can see where you're coming from but there's a lot of absolutes there, and that's not really the world we live in. The companies that sell these things don't sell them as listening devices but as personal assistants; there's probably a clause deep in the google user agreement regarding these recordings but even reasonably tech savvy folks rarely read those in their entirety.

It's not deliberate stupidity when facts are buried beneath the full weight of google's PR department.

And why wouldn't a contractor breaking the privacy laws and revealing all that not be news? It's...a pretty big deal.

I don't think it's a huge deal if in the agreement to use the services that the tech depends on you allow the service provider to, I paraphrase, "continuously improve" "assess and evaluate effectiveness of e.g. voice detection/translation capabilities" or similar. All one has to do is scroll to a section, that is increasingly (and actually required now) to be in relatively low-literacy/plain English, titled "How we process and use your data". Sometimes it's just 2 or 3 paragraphs long, it's not quite as hard as people want to think reading some stuff is. I think this speaks as much of consumer's reckless disregard for their own contracts with other parties as well as commercial exploitation of that fact. The two are intrinsically linked.

Maybe I use the term stupidity in place of ignorance but imo I don't consider them all that different.

I agree that these things can get buried in PR, but part of the problem is the willingness for the wider public worried about these things to admit in profound masse "I don't read terms and conditions or privacy, I tick "Accept/OK" and then complain about this stuff. That is wilfully stupid, no other way around it. I mean, you get user experience techniques now like, you have to physically scroll over the text before you can tick the box, you get updates about changes to terms and conditions that you have to click to accept or ignore them to accept (the latter I have some issue with), and yet people still don't care.

Why not term a phone a "listening device"? You can convert any speaker easily into a microphone by simply inverting a circuit, in some microchips this can be done by programmatically flicking a switch. This idea of the "listening device" isn't a new thing at all, what has changed is the computational power and capability to actually synthesise and convert voice that passes the Turing test. That means the data has to be processed and stored in computers, but it can and has been stored in a myriad of other media in the past. So we arrive at the basic question, do you trust X corporation or Y government agency to store and process this information? If you do, use the product, if you don't, the whole machination coming up with answers to consumer pains will continue.

Green Gecko wrote:Also, phones have had this capability to process voice data digitally for literally decades. Why is it suddenly a big deal because you can put it in a little box? People have their smartphone on 24/7 and "fragment" their privacy into hundreds of apps with dodgy as strawberry float permissions. Nobody cares because they want x or y convenient feature and they go on with their lives. Perhaps I compensate by not using a smartphone, because they really are designed to consume your time like nothing else in history.

its a big deal because people are literally listening and harvesting info all of the time from these things

But they aren't. It's only when you submit a query to the device, the listening for the activation word is processed offline by the device and then the query is sent. At least that's how Google handles it. In that sense it's no different to a Google search. People should be wary when searching Google not to look for illegal things or whatever but they probably still do it.

Google stores the audio snippet for some amount of time (tbh I can't remember) but after a while it retains the text element only. You can go in and delete that. For third party services stuff only the text/handshake is done like ask spotify to play a thing etc.

So it's not much different from interacting with online-connected software in some other way, the difference is the audio is sent out for voice analysis and converted to a text query.

I think what it does, and this is a good thing, is make people more aware of how much search engines know about you from what you search for day in, day out. But it's not worse than that, in my view, not if you consider that in the vast majority of cases your query consists of "OK google play me some music" or "OK Google tell me about elephants" and such like. If people are treating Google like their confidential informant, GP, lawyer or similar that is unwise for all kinds of obvious reasons, least of all privacy aspects, secret service monitoring etc.

I think your assertion that the majority of people can and should make rational decisions based on an in-depth assessment of any product is completely implausible. Ignorance of privacy concerns isn't a personal flaw, it's a state deliberately encouraged by the market as a whole. I believe that the only reason people are mostly happy with the many modern violations of privacy are because they are hidden and obfuscated by technology. Hitting people with one-off mandatory text walls of terms is never going to address that sufficiently.

Yuuuuuup.

Most of us on here are tech savvy enough to at least know about these concerns but the vast majority of people simply aren't aware of this stuff, mainly because it's deliberately obsfucated.

Like, my parents have an Echo dot. I tried to explain the situation to them and their reaction was "We don't say anything worth listening to!" rather than understanding the breach of privacy I was describing.

Even when you're aware of privacy issues, you don't experience them in the same way as you would in a non-technological context. It's a very intellectual awareness for the most part, while privacy is in many cases more of an emotional response.

Free email services have access to the content of your emails and scan through them to build up a marketable profile for serving you ads and whatever else. As a user you never see this happening, so you don't feel like your privacy is being violated in any way. However, imagine an analogous analogue situation, and people would probably object a lot more - for example if your postman opened all your letters and read through them every time he delivered them, and looked at the content of all your parcels before handing them over.

Similarly with things like voice recordings. If there was a person in your house writing down transcripts of some of the things you say, that would probably elicit a strong response from people feeling like their privacy is being violated. When it's hidden by technology, that doesn't happen, even if people "know" about it.

Uhmmm...im gonna pass on this kind of technology. There is a time to use technology and a time to do personal things that will not affect my own private life.

OrangeRKN wrote:Even when you're aware of privacy issues, you don't experience them in the same way as you would in a non-technological context. It's a very intellectual awareness for the most part, while privacy is in many cases more of an emotional response.

Free email services have access to the content of your emails and scan through them to build up a marketable profile for serving you ads and whatever else. As a user you never see this happening, so you don't feel like your privacy is being violated in any way. However, imagine an analogous analogue situation, and people would probably object a lot more - for example if your postman opened all your letters and read through them every time he delivered them, and looked at the content of all your parcels before handing them over.

Similarly with things like voice recordings. If there was a person in your house writing down transcripts of some of the things you say, that would probably elicit a strong response from people feeling like their privacy is being violated. When it's hidden by technology, that doesn't happen, even if people "know" about it.

This is a very good point.

Personally I have some fairly strong "swings" in that, something has to be substantially useful to me in managing my affairs in order to give up some aspect of privacy. I don't think "turn my light in the bathroom off, because I forgot" is a good case to use privacy-invasive technology like a smart speaker. Just since I probably didn't make that clear.

Although the last part isn't really accurate. It's more like there is someone there you have paid to be there, you ask them to take notes on a very specific thing, and the rest of the time, they are invisible and people are kind of worried they are listening when, well, technically they aren't. Whether you trust that to be the case or not is something nobody can really assert for you. I do think this is a key difference that most people don't really understand. It would be monumentally computationally expensive and somewhat pointless to constantly assess everything a user says 24/7. I have no doubt that some secret services already do that, to some degree, maybe on you're on some kind of watch list or part of a random search, but it's not the product people are buying on paper.

Yes (hence I said "some of the things" you say, not "everything"), but I do think the occurrence of "accidental" recordings is something to be concerned about - where the speaker incorrectly detects its keyword, turning itself on and recording a random snippet of conversation. It does happen and that's why I don't like voice activated more than button activated, even if you're still using voice commands after that initial activation. Voice activation requires you to trust the device's own voice recognition, while button activation is very much an explicit action on the part of the user.

It's that pervasive threat that I could be being recorded that I really don't like. As you say phones are also a risk factor, which is why I'm super careful with apps that I install (very few, essentially) and the permissions they request. One of the key differences I feel is that the threat from a phone is mostly from malicious targeting, while with smart speakers (or even phones with a smart assistant enabled) the threat can just be accidental.

I do honestly think that's easier to trigger on Alexa device than Google Mini. on Alexa you can for example just say "computer" which is a ridiculously common word.

The only time I've "accidentally" triggered the Google Mini is when actually playing back a recording that includes me using it, and I say "OK Google" in the recording... which activates the device. Pretty trippy. When the device was launched, it was possibly to tap on top of it to activate it also, but that was removed (so a waste of a component but whatever) because people were taping it and activating it and then thinking it was spying on them. Well, no, it was user error, like accidentally tapping a phone. Of course this was to reduce negative PR around it but less pessimistically one could argue Google made that decision because they didn't want users to worry about it.

But then it's fairly well known the voice analysis is better. So if one were slightly worried between the two, I'd go with Google although that said their privacy policy is not as tight (it's "anonymised" but they're still piling all the query data into a massive big data horde).