|
|
|
|
|
|
by kinsomo
2946 days ago
|
|
|
>> Wow, they (Google) really keep everything.. They've still got my trip to Europe from 5 years ago on there, not sure how that's useful to them at this point haha. > They know you've traveled to Europe, they know all the places you went and how long you were in each one. How is that not useful? But how is deriving those facts from 5 year old location data actually useful to them? AFAIK, they don't try to drive "engagement" by inducing nostalgia with push notifications [1]. I don't really see the use of old data like that for ad-targeting beyond a coarse "travels internationally"-type categories that don't really require keeping the data around. [1] Remember when you were in Europe? Wasn't it fun? We were there too, following you. Please, please log on now to see what we tracked! |
|
|
Ad-targeting is not the only use Google has for data. Google's core view of the world is that useful signals can be extracted from data; the more data you gather, the more useful signal you can extract in the future, even if we don't currently know how to find it. In short, gather all the data, and apply machine learning to it until it produces useful (monetizable) results[1].
Maybe Google will find ways to use this data for advertising, or maybe it has an obvious use in some future project that we won't understand for decades. The point is that we don't currently know all of the uses for data, but Google is betting that those uses will eventually be discovered.
The longer the data exists and the larger the database becomes, it's value increases. Higher value means increases potential profit for a hacker. (or being sold in the future if financial stability fails)
The problem with letting Google have this data is the same bet that Google is making: we don't know all of the ways data can be used, but it's likely that many innovative uses will discovered in the future. Unfortunately, while Google is probably only considering the subset of uses that are useful for business, the full set of uses that will be discovered will inevitably contain uses that are very dangerous to some people. An obvious example is the recent attempt to determine sexual orientation from portraits[2]. Does data about where you went 5 years ago reveal something important about you when combined with the many other features Google knows about you in a big deep learning model?
[1] If the underlying assumptions are true and if it will necessarily produce useful (monetizable) results are open questions.
[2] https://osf.io/zn79k/