1. Audio streaming is expensive and requires lots of storage, even if ephemeral, to be done widely.
2. Natural Language Parsing and transcription is fraught with errors. Go read your visual voicemail sometime.
This, to me, is enough to close the book on this theory.
What is entirely possible is this:
Two individuals have a discussion. This discussion results is one of the individuals searching for a product in a store or online (some stores, like a famous hardware store) can detect proximity to certain items. If they can correlate interest based on a linked (or brokered) profile then they have ascertained that person A is engaged with said item. Targeted advertising is such that it is possible for you to be shown items based on your own interests, but it's also possible you're part of a larger network. Again, all a hypothesis and one hard to prove at that.
Most people are just stupid, nothing new here.
It's possible to do business without scanning your customer's faces, tracing their contacts, selling details of their transactions to aggregators, etc ... where does the line get drawn?
And that doesn't even address data that is "accidentally" collected.
This is the same phenomenon as driving with other people and suddenly asking out loud "Hey let's find such or such type of car". Whatever the make (within reason), you'll probably start finding out a bunch of them. That's because you start paying attention, of course, not because there's suddenly a surge of people driving them at the exact time you're looking for them.
If someone could come up with proof of this, they'd probably be set for life in one way or another. Perhaps they could monetize the information by collaborating with short-sellers, or simply kickstart a career using the massive media attention.
The post goes on to talk about how and why many changes and behaviours can go unnoticed, and I think the whole episode disproves the parent posters initial point.
Humans are good at pattern matching to notice trends but non-technical people always tend to blame something higher up in the stack, like the device in their hand. Or jump to something more malicious by the all-knowing powerful big tech companies.
He was making a comment about wrongly thinking causation from the situation
I think this is an obvious case of someone attributing market-rate fluctuations to the only metric they can personally see with their own eyes: Their battery level. They can’t see Uber’s network of available drivers and other factors that influence the rate, so they overweight the only data point they can see.
This is similar to the phenomenon where people have a verbal discussion about a topic and then coincidentally see an ad for that product in an app later. When we see 10s or 100s of ads per day, it’s inevitable they we’ll randomly see an ad for something we’ve been discussing. Yet some people see these coincidences and wrongly conclude that their phones are listening to their every word, using AI to parse it out (without draining their battery), and then sending those keywords to an ad server for targeting. It has been debunked many times over but it’s still a commonly held belief that all of your phone apps are listening to your conversations.
It can be difficult to convince some people that coincidences are really just coincidences. Especially in this era where Big Tech has an incredible evil reputation in the journalism and media we consume.