4 ms·

Taken to any extreme, this could easily be extended to "don't ever engage with anyone or anything for any reason whatsoever." Is there concrete evidence any of

Taken to any extreme, this could easily be extended to "don't ever engage with anyone or anything for any reason whatsoever." Is there concrete evidence any of this is happening? If not, this just reads like a recipe for paranoia. What we've seen of ML really _isn't that good-_ my phone has full access to anything I've ever typed and its predictive keyboard rarely actually predicts what I'm going to say.

mandmandam 5y ago
> "don't ever engage with anyone or anything for any reason whatsoever." On Facebook. Yeah probably a good idea and not really paranoid considering Fuckerberg's history. > What we've seen of ML really _isn't that good-_ ... Yet. And, after Brexit and Trump we've seen the difference a few percentage points can make. Add another point every couple years as the tech improves and you're mad not to be paranoid.
Pokepokalypse 5y ago
>s there concrete evidence any of this is happening? Cambridge Analytica? Doesn't even matter if there's any truth behind the science. If the personality profiling works, they can apply their MODEL from the 270k responses they DID get, against every single other user's public profile information, and get a high-likelihood result. Does that mean they can tell exactly what MY personality type is? (as much as that may or may not be a valid "thing") - - I doubt it. But can they locate a region on a map where around 180,000 of 200,000 residents fit into a particular personality type, and then target ads to that region with a degree of better accuracy than "random"? Absolutely. Index that against voting records; (which is what the "fake election audits are for" why else do you think they started with Arizona which blew in an unexpected direction for them? Also Kobach's efforts in 2017 were probably actually a data-collection operation for the GOP) - and you have a good map for how to target campaign spending in the next election. I also assume the Census data is abused in this way, (by whichever party is in-power at the time). "dont ever engage with anyone or anything for any reason whatsoever" is fine and dandy. And it's like a Zombie Apocalypse: you can protect yourself from being bitten. But you can't protect yourself from the consequences of everyone else being bitten and turning into zombies and coming after you in such great numbers, that you don't have enough ammunition stockpiled for all the headshots you'll need. (That's my analogy for "people infected with racist (and etc) disinformation are gonna vote a certain way, whether you take in that disinformation or not.")

mandmandam 5y ago

> "don't ever engage with anyone or anything for any reason whatsoever." On Facebook. Yeah probably a good idea and not really paranoid considering Fuckerberg's history. > What we've seen of ML really _isn't that good-_ ... Yet. And, after Brexit and Trump we've seen the difference a few percentage points can make. Add another point every couple years as the tech improves and you're mad not to be paranoid.

Pokepokalypse 5y ago

>s there concrete evidence any of this is happening? Cambridge Analytica? Doesn't even matter if there's any truth behind the science. If the personality profiling works, they can apply their MODEL from the 270k responses they DID get, against every single other user's public profile information, and get a high-likelihood result. Does that mean they can tell exactly what MY personality type is? (as much as that may or may not be a valid "thing") - - I doubt it. But can they locate a region on a map where around 180,000 of 200,000 residents fit into a particular personality type, and then target ads to that region with a degree of better accuracy than "random"? Absolutely. Index that against voting records; (which is what the "fake election audits are for" why else do you think they started with Arizona which blew in an unexpected direction for them? Also Kobach's efforts in 2017 were probably actually a data-collection operation for the GOP) - and you have a good map for how to target campaign spending in the next election. I also assume the Census data is abused in this way, (by whichever party is in-power at the time). "dont ever engage with anyone or anything for any reason whatsoever" is fine and dandy. And it's like a Zombie Apocalypse: you can protect yourself from being bitten. But you can't protect yourself from the consequences of everyone else being bitten and turning into zombies and coming after you in such great numbers, that you don't have enough ammunition stockpiled for all the headshots you'll need. (That's my analogy for "people infected with racist (and etc) disinformation are gonna vote a certain way, whether you take in that disinformation or not.")