I'm not sure about computer image generation but you can (relatively) easily fingerprint images generated by digital cameras due to sensor defects. I'll bet there is a similar problem with PC image generation where even without the EXIF data there is probably still too much side channel data leakage.
An extent presentation (IMHO) by Sam Vaknin on how the very way modern social networks operate even down to their form are deeply destructive. The "Internet Hate Machine" is real and it was created by anti social deeply narssistic nerds.
The groupthink is real, and it's coming for a skull near you.
But in all seriousness I think it's a mix of bots on the dead internet leading the monkey see monkey do paradigm. If you see 80 out of 100 people doing a thing then you get swept up in the flood. Even if 50 of those 80 are bots.
That's because AIs can't survive by eating their own output. The only solution they know to ward off model collapse is more human input. They needyou to use AI to feed the beast. And if it's built into your office apps, they get that data for free.
That's part of why every service and system are getting integrations, It's not for us it's for data harvesting.
In the end that's what "Windows Recall" will be used for. Access to every moment of every user for every app... Can you imagine the training data that would provide? An AI that could run any program ever created.
I work in MSFT although not in office org. Based on my experience, the reason is far more trivial. Someone has a half year goal (KR) that says I/my team will increase engagement by N% from X to Y. Some people, whom I don't respect, when presented with a goal like that immediately start doing this (tfa) kind of stuff. Many people, when towards the end of the period some of their genuine (i.e. delivering good stuff) bets don't pan out and the numbers don't number, start doing things like this or generally throwing stuff at the wall to see what sticks.
I bet there was a meeting where someone axed the off button because numbers.
I did a brief stint in office and back this up. There’s a no malicious grand scheme, just the the loudest mouth in the room this quarter calling the shots. It’ll be someone else in 6 months demanding a different color of shit thrown at the wall.
> Based on my experience, the reason is far more trivial. Someone has a half year goal (KR) that says I/my team will increase engagement by N% from X to Y.
How is does this contradict the comment you are replying to?
It implies there's no nefarious intent to collect some training data. In my area at least the only user data I'm aware of is used for measurement of engagement in anonymized aggregated form. Engagement metrics still exist, because supposedly on yet higher level they translate to revenue, not because of training (unless you count thus feature works do more of it as training). I assume the office org is not different.
Because it increases the prestige of your department when you can say 'we developed features which are now used by X% of users'. If you've ever wondered why every new feature in a Microsoft product seems to need to be used, this is why. It's so the team that implemented it can justify themselves.
> Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization's communication structure.
