(no subject)

[yvannairie]

I woke up absolutely goddamn confused over the news about how Microsoft and Amazon have employees going through written/audio data for improving recognition and this is apparently considered a breach of privacy by some.

Because like... no shit they're doing that, how do people think handwriting recognition works?? How do they think image/audio recognition is trained? Of course there's The Eventual Human who looks at that data and explains to the computer what it means.

Information doesn't come out of nowhere, and I'm frankly surprised people are surprised by this.

Comments

[yvannairie]

Like what did people think "retaining anonymous data for analysis" means?

"This service that uses your voice and face to function has its employees look at your face and listen to your voice."

[cimorene]

lol @ paying money for the privilege of turning all your money over to one of the 5 or so biggest and evilest information tech companies in the world and then later getting all upset about privacy.

[yvannairie]

Yeah, that, but more just..... who did they think does all the error-checking and user calibration for stuff like Alexa, Ciri, Cortana, Touch, etc etc?

How do people think that kind of optimisation happens?

Who do they think made these services?

[cimorene]

Well, it's funny once you have the slightest insight into how neural networks (or any algorithms... I think they aren't all neural nets?) work, but I suppose the blindingly obvious answer is that most people don't think about those things because they don't have any idea how algorithms work; artificial intelligence is essentially magic to them, much like the working of technology in space opera.

Dunning-Kreuger effect, I guess?

Although even if you were under the impression that Siri or Alexa did it all completely using a supercomputer and that they were each approaching human levels of intelligence, I don't know why they would think that made their data significantly safer from the parent companies. Since they would have programmed these hypothetical genius AIs, their data would still be on their servers... even if you don't suspect them of a desire to use it against you, there's still the possibility you won't like the next way they figure out to use it to sell you things.

[yvannairie]

And all of that gets me wondering about how much of this is that as internet was popularised, internet safety lessons... somehow stopped? Like, the Facebookification of internet somehow masked that this information being out there, freely isn't neutral?

[cimorene]

I know, this is one of my pet peeves. -_- I've read that it probably is due to Facebook and that whole paradigm of not trusting strangers was left behind for "Only people who have something to HIDE would hesitate to make numerous personal details and photos readily searchable".

But there's also probably an element of people not really knowing much about how things work too.

[yvannairie]

Yeah. Which I definitely see as having become a thing once Facebook started to become a thing. My teachers, for example, were really diligent about teaching us netiquette and how information moves in the internet (i.e. once it moves, you can assume it can be found forever because at no step is it guaranteed that it isn't copied). I've had people express shock when I explain that the only thing stopping all of their personal data and messages and calls made online from being recorded was because the service providers promised they wouldn't.

They've never made such promises about the services we've been discussing -- in fact! The opposite has happened!! They often explicitly say that they'll be using the data "anonymously" to do all the same things with it that people get worked up about when they know their names can be connected to it.