Not thinking this is happening either. But just to speculate on a fictitious scenario..
Given a shady company, maybe even outright involved in illegal practices, where employees employ any surveillance tech they can lay their hands on. All in order to collect as much personal information as possible, from as many people as possible. As such, they won't shy away from breaking & listening in and do subsequent speech-to-text information gathering (e.g. via some Ad-obtained Windows malware, or a malicious mobile app).
They trade the collected info on the data markets, selling to anyone who'd pay. Wouldn't it be likely that this data then indirectly ends up at Google, so they can indeed target - in this case Voltaren - ads to you?
Also: If any illegally obtained information enters the data markets.. won't it be 'whitewashed' automatically as it is trading hands?
Sure, but I don't think they're listening from a technical standpoint, because the battery consumption is much less than an active network adapter would consume.
They could always either compress the audio or do a speech to text (looking for keywords rather than caring about getting the whole context correct), and batch upload the results on an interval (like when checking for email or app/system updates).
Id reckon you can boil the model down to a very small one to extract some keywords and thatll be pretty good already. Just having the ability to hear if someone is talking would take very little battery, to consequently turn on the top-10000 word model would still be barely noticeable on these powerhouses of phones we got, I think. Haven't tested it, though.
Once a minute (or five, ten, whatever), turn on the microphone for one second. If voice detected, record a bit and process. Doesn't need to be always on, and as a feature missing keywords sometimes adds to the doubt that it exists ("I said 'pfizer' and got ads, you said 'pfizer' and didn't, must be something else).
Not thinking this is happening either. But just to speculate on a fictitious scenario..
Given a shady company, maybe even outright involved in illegal practices, where employees employ any surveillance tech they can lay their hands on. All in order to collect as much personal information as possible, from as many people as possible. As such, they won't shy away from breaking & listening in and do subsequent speech-to-text information gathering (e.g. via some Ad-obtained Windows malware, or a malicious mobile app).
They trade the collected info on the data markets, selling to anyone who'd pay. Wouldn't it be likely that this data then indirectly ends up at Google, so they can indeed target - in this case Voltaren - ads to you?
Also: If any illegally obtained information enters the data markets.. won't it be 'whitewashed' automatically as it is trading hands?