Yes, devices are (mostly) not listening to us now, for two reasons: one of the reasons being, as I said in my OP example of the moms and the strollers from the Netflix documentary 'The Great Hack" you mentioned, they don’t need to listen. Through geolocation, together with the info they’ve already gathered, they already know plenty about us.
The other reason is because it’s simply been logistically impossible or at least impractical, given the technology up until now, to listen to everything we say and parse it meaningfully into useable info. But with the advent of AI I imagine it will, or already has, become trivially simple to do. At first it may only happen with obvious opt-in situations: if you say “hey Alexa, add ‘X’ to my shopping list” you can’t be too surprised if you start getting served ads for ‘X’. But how long before your phone does always listen? It geotracks us now, and I don’t remember opting in to that (it was probably buried in some 10,000 word user agreement I checked but didn’t read).
How soon before you’re lying in bed with your significant other, phone on the nightstand, talking about taking a romantic trip to, say, Aruba, and the next day you start getting ads for vacation packages to Aruba? Someone, go ahead, tell me I’m being paranoid.
last time I checked, those AI-frontends that you can install on you cel. ask you to give it access to you microphone … and in android you can opt for: yes/no/just when app is active/only this time … so that should take care of that worry.
also the whole geolocating … how does that work in an appartment complex with 400 different 30-40m2 sized appartments spread out over 27 floors. How will “the internet” know if you are on the 5th/15th/25th floor … when you all have the same “gps-address”, as you are living atop each others?
anybody with more in depth gps knowledge caring to weigh in?
The geolocation on typical phones is probably good to within 20 meters / 60 feet, although it can be much more accurate than that. Vertical location is much less accurate, and it’s also poorer inside.
Figuring out how to tell when two people are interacting versus living in the same building, passing by, or just near each other, is the “secret sauce” that companies are trying to perfect. Some simple examples of the rules they might build in are:
if the location is an apartment building, proximity doesn’t suggest interaction
if two phones are close and no others are nearby, it does suggest interaction
two phones moving together for an extended period suggests interaction
two phones in proximity mutiple times on different days suggests interaction
As @TroutMan says, geolocation is more used for detecting certain ways two devices are in close proximity to one another, to determine whether their owners are likely interacting. Or to track your movements, as in my example upthread where my phone apparently tracked me to my lunch at Qdoba and served me ads for Cholula hot sauce as soon as I got back to work.
They don’t need to geotrack you at home; they have plenty of ways of knowing that you are reading or posting from a certain device- the sites you visit and especially log into, as well as a number of data points they can grab. Even if you are on a VPN network to try to hide your online identity, I believe trackers are reading your MAC address (that’s Media Access Control, not Mac as in the computer company), which is as uniquely identifiable to your device as your home address is to your house or apt.
There are ways tracking companies have of identifying us that we probably can’t even imagine. For example, did you ever wonder about those simple captchas that just have you check that you’re not a robot? How is that even detecting whether you’re human or not- can’t a robot learn to check a box? The answer is:
Does the cursor move in a non-straight line, like a hand moving a mouse? Human. Or does it move in straight lines across, up and down, like an etch-a-sketch? Robot. Similar for those “click all squares with a traffic light” style captchas. It’s not so much seeing whether you click every box with a single pixel of traffic light in the frame, it’s seeing how you do the clicking.
Extrapolating from that, people might even have a ‘signature’ to the way they move their cursors onscreen. allowing another way for them to be individually identified. Pure speculation, but I wouldn’t be surprised.
I don’t understand why this is so difficult.
Why can’t a computer learn to move the cursor to imitate a human hand?
It should be easy to write a program that tracks a few thousand samples of cursors being moved by hand. That data is just the averages of a bunch of X,Y coordinates which create a jagged line , you could use the data anywhere, to create a human-like line which ends on the pixel labelled "check the box that says I am not a robot. "
It should be simple arithmetic. What am I not understanding?
I’m sure it can be, and I’m sure it probably will be or has already been programmed to do so. It’s an arms race I imagine, with algorithms being written to fool captchas, and captcha developers having to create different methods to detect the robots.
My point was more that there are many more methods to detect not only people, but individual people, than we might imagine.
It can measure the relative difference, so for example, it knows if you carry your device up one flight. But it is not accurate enough for absolute measurements of elevation. It cannot tell that your phone is 20 feet below a different phone.
I just did an experiment in the building I’m in, with Physics Toolkit. At rest the noise in the signal is about 5 to 10 Pa. The second floor of this building is pretty consistent 40 Pa lower pressure than the first (although these are perhaps taller-than average floors).
Given an ensemble of measurements, I suspect one could get some reasonably good estimates of who’s on what floor. Especially in a shopping-centre-type situation where the floors are quite far apart.
Just by using the same internet connection they can assume you live together. Or in my case, all 4 internet connected devices are being used by the same person.
It’s not just the line, it’s the nanosecond differences in the movement.
The best example of the internet spying on me was when I talked about having a squirrel invade my house on flakebook. For weeks, every other ad was for this one squirrel control company. That’s ads from places that weren’t (obviously) connected to Facebook
When the website diplays “check I am not a robot”, the computer begins tracking your mouse movements, speed and direction. If the data matches the correct amount of “human-ness”, all you need to do is save that data. Then a robot can move its mouse just like a human.
It’s simple math.
And it seems to me that it’s easier for a bot to do simple math than it is for the same bot to recognize the bicycles in captcha pictures.
It’s not simple math. If it were simple, those captchas wouldn’t exist.
This reCAPTCHA test takes into account the movement of the user’s cursor as it approaches the checkbox. Even the most direct motion by a human has some amount of randomness on the microscopic level: tiny unconscious movements that bots can’t easily mimic. If the cursor’s movement contains some of this unpredictability, then the test decides that the user is probably legitimate. The reCAPTCHA also may assess the cookies stored by the browser on a user device and the device’s history in order to tell if the user is likely to be a bot.
Yeah, I’m assuming it was the IP address - not likely to be anything finer-grained than that, as we don’t share any devices.
Which is interesting, because with some of the fibre-to-property internet connections being rolled out now, there is Network Address Translation happening at the cabinet - so multiple houses have the same internet-facing IP address, which in theory means some of the trackers might start recommending a thing the pervert next door is into.
If the computer can measure it, why can’t the computer mimic it?
The bot doesnt have to physically move the mouse. The bot just sends instructions as electric pulses to the screen.“Move from pixel number x to pixel y, in 0.01 nanoseconds. Now move to pixel Z in 0.015 nanoseconds. Now pause for a full nanosecond, then continue.”
I imagine human-like randomness could be modelled and realistically reproduced, but the sort of ‘random’ noise a simple computer program can generate just isn’t the same as true randomness (and the randomness in human movement is likely to be compounded by a less-than-random component that is characteristic of how people move their hands).
It could be modelled and reproduced, but it won’t be as simple as a couple of lines of code saying ‘add pseudorandom offset X at pseudorandom interval Y’
One more data point: yesterday I picked up a small purchase I’d made on FB marketplace at the doctor’s office where the seller works. Today I’m getting banner ads from the outdoor furniture retailer in the adjacent building. Might not be related, but hmmm.
If it’s a popular channel, then there are probably a lot of folks searching for that term. Though, of course, YouTube is under the Alphabet umbrella, so it could be more direct in this case.
I see the same effect when I Google crossword puzzle clues. I’m a few weeks in doing them, but then, so is everyone else who uses the Seattle Times app.
Sure, they make grandiose claims… but take a look at all of the email addresses called “trumpisapoopyhead@mydomain.com” that got a deluge of complete garbage Trump marketing.
That’s not weird at all. Ad providers have been able to tell your rough location for decades. It’s the same process behind “Hot singles in <your town> want to meet up with you”, just used legitimately.
GPS is only a fairly small part of location services, these days. As I understand it, the biggest source of information is what wifi networks are visible.
That’s part of it. But it also depends on Google (ReCaptcha is another part of the Google umbrella) recognizing that your computer has been used by a human in the past. If it’s not sure enough, it’ll serve you a “click on all traffic lights” one or something, instead.
But taking a step back, here: Why does everyone always get so upset about targeted marketing? It’s a heck of a lot less obnoxious than the untargeted marketing that we used to be bombarded with. Sometimes, it can actually be convenient, telling us about products that we didn’t know about before but which we’re genuinely interested in.
