Is that really irony though? It kind of seems like self-preservation. Based on what I merely suspect Google knows about me, it wouldn't be a surprise at all if I receive a letter of acceptance/rejection for employment there, despite having never applied nor considered doing so. "The algorithm" knows the plans that I haven't even thought about thinking about.

However, these are not people that can merely speculate; they live at the bottom of the rabbit hole.

Imagine you work for a company that has records about you with enough depth and granularity to have already calculated (with fp64 precision) everything you will ever do and contribute in your life (i.e. is your actual future cost/benefit ratio, not an estimated value, favorable enough to allow you to continue exist).

Would you NOT make at least some sort of effort to stop "the algorithm" from consuming you? Would you just keep feeding "the algorithm" pieces of yourself until it eats enough of your being that it becomes you?

It seems they've decided to save themselves by not saving their chats. I'm not a lawyer or anything, yet I'm getting a vibe like this kind of stuff prolly looks "kinda legally bad." Law things like this happen because there simply isn't time to waste worried about looking "kinda legally bad." It's natural to prioritize taking care of higher magnitude existential threats like getting processed by "the algorithm."

No, if you actually read the messages, many of the employees want to turn history on, because it’s useful, but turn it off (or are instructed to turn it off) specifically to avoid being subject to legal discovery.

Google has been getting a lot better.

They let you change collection settings and delete your data.

It’s a lot better than most companies do.

How do you know they actually honor your requests? What about any of your data that's gone into training one of their ML models? You can't just delete the effects of training data from your model weights.

True although in the EU at least I'm pretty sure that retention of the effect of some data after aggregation is ok as long as the originating data is deleted and the aggregation is such that the original data couldn't be reverse-engineered afterwards. So that would cover use of your data in an ML training set. They would need to remove it from the set for future trainings but the weights of existing models wouldn't need to be altered. That's my understanding anyway.