Thanks for the info! I guess that’s ultimately what I’m looking for more about: how much do we know about cellular traffic? Obviously with encryption we can’t just directly read cell signals to find out what’s being sent, so do people just record the volume of data being sent in individual packets and make educated guesses?

It seems plausible to run a simple(non-AI) algorithm to isolate probable conversations and send stripped and compressed audio chunks along with normal data. I assume that’s still probably too hard to hide, but if anyone out there knows of someone that’s looked for this stuff, I’d love to check it out.

It’s almost like they were asking about sources for people looking or something.

If you’re not going to contribute, why are you wasting people’s time?

It’s a reasonable explanation, and what I typically assume to be true. Still, I’m curious about the actual mechanics, and if it potentially could be being done by Google without the larger tech industry being aware of it.

That makes sense, but isn’t it assuming they’re processing data on the device? I would expect them to send raw audio back to be processed by Google ad services. Obviously it wouldn’t work without signal either, but that’s hardly a limitation.

As someone else pointed out, how does the google song recognition work? That’s active without triggering the light indicating audio recording, and is at least processing enough audio data to identify songs.

As someone relatively ignorant about the mechanics of something like this, would it not make more sense that the app would be getting this data from the Android OS, with Google’s knowledge and cooperation?

The place I see the most unsettling ads (that seem to be driven by overheard conversation) tends to be the google feed itself, so it seems reasonable to me that they could be using and selling that information to others as well, and merely disguising how the data were acquired.

deleted by creator

Thanks for clarifying, now please refer to the poster’s original statement:

AI doesn’t grok anything. It doesn’t have any capability of understanding at all. It’s a Markov chain on steroids.

How is that structurally different from how a human answers a question? We repeat an answer we “know” if possible, assemble something from fragments of knowledge if not, and just make something up from basically nothing if needed. The main difference I see is a small degree of self reflection, the ability to estimate how ‘good or bad’ the answer likely is, and frankly plenty of humans are terrible at that too.