Hi
@nprofile1q...,
I didn't understand the article that way at all, and I think there might be a misreading of what
@nprofile1q...'s actually proposing here.
Her entire piece argues that smartphones should minimize data sharing with third parties. She's explicit: a trustworthy phone "wouldn't share our data with random companies." That's the core of her vision. So the concern about "shipping out so much of your data that you can't even keep track of it" seems to contradict her stated argument, not align with it.
I suspect the confusion comes from this phrase: "it would use machine-learning to understand and enact what we want, **instead of to manipulate us into serving others first.**" The key word here is "instead"—Farrell's contrasting two different uses of machine learning.
When she talks about automating data access, she's not talking about automatically sharing more data. She means the phone learning your preferences and respecting them without constantly interrupting you—much like how voice-to-text software learns to transcribe your voice more accurately over time, or how a personal assistant gradually figures out what you like without needing constant instructions.
And right after that, she addresses the friction concern directly: the phone "would give access to our data as and when we wanted, but also not bug us too much with opt-ins." I get why that matters—GDPR cookie popups are exhausting, and most people just click "yes" rather than make informed choices. But that's different from untrackable sharing. She's describing a system that *reduces friction* while *preserving control*, not one that hides what it's doing.
So I'd gently push back: I think "using machine learning models to automate sharing of data" is a bit of an overreading. Farrell's proposing using machine learning to *protect* user interests, not to enable indiscriminate data leakage. That distinction matters—and it seems pretty aligned with what you're describing too, at least as long as there's a clear opt-out option.
Does that land differently?
@nprofile1q...
