Hacker News new | ask | show | jobs
by NavinF 658 days ago
That model would have to be trained locally since the data never leaves the device
1 comments

So they say..
This is some serious tinfoil hattery.

The data is not small. If it was being uploaded, you'd notice.

If screenshots were OCRed and then only text was uploaded, I doubt you would notice anything.
Pssst... You can extract metadata from data and then train on those "anonymized" metadata ;)
> This is some serious tinfoil hattery.

microsoft regularly makes promises about privacy and telemetry, and regularly breaks them. They have been multiple law suites about this.

In germany we say "Wo ein Trog ist, sammeln sich die Schweine", especially in the discussion around recall.

Have they lost any of those lawsuits?
So it only counts as going back on their word if they lose a lawsuit about it?
Sure, or if there's any evidence that these aren't just BS allegations. The signal to noise ratio in this space is 0
They have it in their absolute authority to push out a patch at any time that will make the data be uploaded, even if it isn't right now.