Honestly, that's fair. In my experience Google does a lot toward being trustworthy since it's in their long-term best interest. But at the end of the day I'm just some random guy on HN comments, so you're right to be skeptical.
You shouldn't trust anyone. There's nothing stopping them from doing it again, or continuing to do it and publishing that nice writeup. Of course when your data is on a 3rd party server, that 3rd party has access to it. Why do we keep pretending like anything else is the case?
Of course when your data is on a 3rd party server, that 3rd party has access to it.
You realize that oblivious third-party storage can exist, right? For example, Tarsnap. It's the the durable-storage equivalent of end-to-end encrypted messaging. The client does all the encrypting; and the client is open-source; so you don't have to trust the third-party at all.
This is a bit more challenging when you want clients to be able to operate over an index of the data, rather than just the raw data (i.e. you want them to be able to search through their stored email)—but it's fully possible to architect a service such that all indexing happens on the client (perhaps distributed between many connected clients in the case of a shared data-source like a Slack workspace), where the backend is just obliviously storing and returning E2E-encrypted copies of generated index chunks, in the same way it obliviously stores and returns E2E-encrypted copies of the data itself.
Going by the uncanny-valley accuracy of their suggested auto-replies, they're training their natural language neural nets with every message sent over their wires.
I mean, maybe high-level aggregate data like "X% of users got emails about shopping" might still be in some spreadsheet somewhere, or in people's brains. But no, I can't think of a way any personalized data would still be around.
Comments
Doesn't gmail scan your mail for data to target ads?
In the past. We (I worked on gmail ads during this time) stopped doing it about a year ago.
https://blog.google/products/gmail/g-suite-gains-traction-in...
Thats good, but something tells me I still shouldn't fully trust Google/Gmail with things like this.
Honestly, that's fair. In my experience Google does a lot toward being trustworthy since it's in their long-term best interest. But at the end of the day I'm just some random guy on HN comments, so you're right to be skeptical.
You shouldn't trust anyone. There's nothing stopping them from doing it again, or continuing to do it and publishing that nice writeup. Of course when your data is on a 3rd party server, that 3rd party has access to it. Why do we keep pretending like anything else is the case?
You realize that oblivious third-party storage can exist, right? For example, Tarsnap. It's the the durable-storage equivalent of end-to-end encrypted messaging. The client does all the encrypting; and the client is open-source; so you don't have to trust the third-party at all.
This is a bit more challenging when you want clients to be able to operate over an index of the data, rather than just the raw data (i.e. you want them to be able to search through their stored email)—but it's fully possible to architect a service such that all indexing happens on the client (perhaps distributed between many connected clients in the case of a shared data-source like a Slack workspace), where the backend is just obliviously storing and returning E2E-encrypted copies of generated index chunks, in the same way it obliviously stores and returns E2E-encrypted copies of the data itself.
I mean instead they're just training their 'auto-reply' AI systems which they probably expect to be more valuable than advertising anyway.
Every interaction you have through google gives their neural nets a little more data to predict everyone's next reply.
Now I get ads in gmail relevant to what I search on google.com
What is the monetisation scheme of Gmail now? Just a free enticement to buy the G-suite?
Going by the uncanny-valley accuracy of their suggested auto-replies, they're training their natural language neural nets with every message sent over their wires.
They still have ads for the free version, just not based on your emails.
Is any of the data used from that collection/analysis still around?
No.
I mean, maybe high-level aggregate data like "X% of users got emails about shopping" might still be in some spreadsheet somewhere, or in people's brains. But no, I can't think of a way any personalized data would still be around.
Yeah, we believe you. wink wink nudge nudge.
A couple of years ago a friend sent me a barbershop quartet video and I spent a few days getting hairdresser ads ...
Solution: use an adblocker.