Before the discussion gets too heated: the study authors are at pains to point out that this is simply a PoC. It’s possible that further studies will show overfitting and/or training bias (though clear to me what bias might exist in the turker approach they took).
Very I nteresting work. Presumably the point of the announcement is to encourage agencies to fund more grant requests.
Reminds me of the thesis in James Scott's Seeing Like A State. This is where the effort to make things legible (easier to count, to measure) lead to standardizations which on turn shaped policies. For example, old growth forests were strategic assets (masts for wind-powered warships). To make then easier to count, trees were replanted in neat rows (which had the unintended sideeffects of losing eco diversity for a healthy forest).
Big Data and AIs are an extension of that idea. And while the machine learning doesn't require as rigid of a classification scheme, resulta will still be understood through the lens of what can be made legibile. For example, classifying cars to estimate socioeconomic status assumes that most people within the population owns and drives cars.
Somewhat relevant, "you're not allowed to smile in your passport photo" is so that the face recognition machine has it easier to identify you...
Maybe people who don't like the surveillance state should run around in Joker make-up, and at least there's an answer to "You want to know why I got these scars?"
The title reads a little disingenuous. Aren’t they using deep learning and Google Street View to predict the make, model, and year of a vehicle? Then, separately, they’re using predicted vehicle information to guess the demographic traits of the person who lives there.
Still, applying image recognition on Google Street View is a pretty cool idea. The location information gives you a whole other set of data to join with. For example, you could recognize flags hanging on people’s houses and join that with immigration data (there are a lot of Italian flags where I live), or recognize species of trees and join that with climate data.
The factoid about San Francisco in the article confirmed my perception that American cars are just not very popular is actually a distortion of the Bay Area bubble I live inside of.
Edit: from the narrow view of my kitchen window in West Oakland I can see 2 Toyotas, a Subaru, a Ford Focus and the VW e-Golf in my driveway: 20% American. There'd usually be another suby in my driveway, but my son is home from college and borrowed it.
Those toyotas were likely assembled in the usa, and from probably as many american parts as the ford. 'Domestic' and 'import' dont mean anything these days. They are all made by large trans-national corporations for whome citizenship is fluid.
That's quite interesting, but I wonder if cars in garages could skew the data. Garages and garage size affect what cars can be kept in a garage (e.g., extended cab pickups might not fit), and also might be correlated to geography and affluence.
Probably, but not much. In my experience, the vast majority of garages in the US are used as storage for "stuff", not cars. This probably changes as people get wealthier and want to store their more expensive cars indoors (and can afford to put their miscellaneous stuff in paid storage if they'd like to).
Might also be different in urban cities where street parking is scarce and leaves your car more vulnerable to breakin.
If you care about your paint the garage is way better than leaving your car out in the sun in the south west. Not as annoying as clearing snow day to day but still a concern.
Hey, yep :-) I leave my vehicles out in the summer when all sorts of yard and other stuff is spread around the garage. But come mid-fall that all gets cleaned up so the cars can go in and let the plow do its thing. There's much less manual cleanup needed than if there's a car in the driveway.
You’ll certainly find people using garages for cars in places where you get snow and ice, and the temperature at night is below freezing.
Living in northern Wisconsin, for example, if you don’t park in a heated garage, snow and ice might not melt off your car for the entire winter. Parking in a garage means you won't have to scrape snow and ice off for 10 minutes standing in subzero weather. Parking outside, one also might need to plug the vehicle into an engine block heater on particularly chilly nights or risk the vehicle not starting in the morning.
In wealthy regions, the cars might frequently be missing from StreetView.
For example, in Woodside & Portola Valley (CA) most homes have garage or other off-street parking. A single home might house a Prius, a Hummer, a Volt, an extended cab F150 and a McLaren F1, all invisible from the street.
What matters is how many cars can be seen by google. In both very rich and very poor areas cars are kept away from the road. This may work in city suburbs but not in downtown cores or rural areas. Drive around beverly hills and you will see more pickup trucks belonging to contractors than bmws belonging to residents. They park in garages. Workmen park on the street.
I find writing to be much easier than drawing or painting, because I have an easier time with interpreting discrete characters and then interpreting them through a dictionary in my brain accrued over my lifetime. Why are you asking differently of AI?
It could be illegal for them to use that. Perhaps the party does the redistricting which an assembly member introduces into the legislature. To use the database, it would either need to be publicly available or else the executive would have to do the gerrymander. (Which might be how it works. But I live in a place where good government is valued.)
Why on Earth does the legality matter? It's not like they'd ever admit they are gerrymandering. What data they use to optimize their gerrymander is irrelevant.
i wonder how visual methods like this could monitor for sign of voter fraud. you could probably film polling places and use peoples faces as ids, and clothes and gender and age to get some reasonable estimates
Because You can use this to discriminate charge even $100 for this and it locks out poor and homeless people or you close places where you can register in the areas of where more of your opposition live.
i feel like i accidentally stumbled in to a r talking point. but ill take the bait: the id doesnt do anything to show you irregularities between the demographics and the results
Citizens are not required to have government issued photo ID.
If you make government issued photo ID a requirement to vote, you have violated the constitution and prevented legal citizens from voting.
Voter fraud is not a real problem. Election fraud is, and election fraud is often carried out by removing legal citizens' right to vote, just as you described.
When you see a state making laws about what ID you need to vote, you are watching election fraud in action.
> Section 1. The right of citizens of the United States to vote shall not be denied or abridged by the United States or by any State on account of race, color, or previous condition of servitude.
Nothing in here about "shall not be denied or abridged because of lack of photo ID".
As long as acquiring an ID is equally painless for all demographics.
Introduce a fee, limited licensing locations/appointment times, or complex applications, and the indirect result is often selective discrimination by race. That's still unconstitutional.
Do you believe that the list quoted is intended to be restrictive or illustrative? Do you believe that citizens right to vote should be denied or abridged in general, or just in the case of photo ID (which I doubt was a thing at the time the constitution was written)?
You do not need to prove you are a citizen to vote. Voter registration requirements vary state to state. Typically states will request some form of identification, like a social security number, or a birth certificate, or something like that, if you do not have a photo ID, so they can verify who you are and that you did not vote twice. A government issued photo ID, or other proof of citizenship, is not required. That would be unconstitutional.
Edit: California will even allow you to use a utility bill with your name and address as valid ID to vote, for example. No photo ID or citizenship requirement, and no voter fraud happening in California either.
If requiring a government-issued photo ID is, as you claim, unconstitutional, then why is requesting identification of any sort (SSN, birth certificate, utility bill) not also unconstitutional?
>no voter fraud happening in California either
If the current method for identifying people doesn't require a photo ID, by what means are we able to claim that there is no voter fraud in California? How would we know that is the case if legally there is not a mechanism to catch or even detect hypothetical cases of it happening?
> If requiring a government-issued photo ID is, as you claim, unconstitutional, then why is requesting identification of any sort (SSN, birth certificate, utility bill) not also unconstitutional?
because goverment-issued photo ID can be denied or made prohibitively expensive to large numbers of poor people who's interests might not align with the people in power. Other forms of identification are simpler to obtain by the poor.
Most functioning democracies have the means to assess the level of voter fraud without requiring photo IDs. There are a bunch of civil servants who have that exact job; I think the onus should be on people who believe there is systemic voter fraud occurring to provide some evidence before additional restrictions should be put into place.
Also, democracies were also functioning fairly well long before photo IDs were invented.
many places use simple ink-on-finger from fingerprinting the voter. I believe they're more common in less-developed places. Other places simply accept non-photo id. In Canada most people just present their voter card, which is mailed to them - it doesn't have a photo on it. However even without the voter card, other non-photo id options are acceptable.
For those who insist on it, why is photo-id neccessary? Why is the photo-id level of assurance the bar that needs to be crossed? Any system can be cheated, but some are simply good-enough, and they balance the drawbacks of higher-security systems.
I don't have the same kind of lock on my front door as the bank does to their vault. The lock on my front door isn't perfect, but it's good enough for the threats it's likely to face, and avoids the drawbacks that having a bank-vault lock on my front door would.
I'm also genuinely curious why the people who insist on photo-id for voting believe the threat of voter fraud is high or likely enough to warrant that level of assurance compared to a slightly simpler system without the photo. Do you really live in a place where you'd expect enough voter fraud for it to have a material impact on how your democracy functions?
I'm not in America but I've had to vote in many elections. I don't think I ever presented any ID ever, they just took my word that I'm the one who I said I was. (I think there was once a state election when they posted me a token that I was supposed to return to vote.) Yet the elections here are widely regarded as high quality.
But if your goal is to prevent double voting, you would just use the inks they use in some newer democracies. Then you know John MacIlroy only voted once because you can see it on his fingers. This is probably even more reliable because you won't cut off your finger to hide the ink, but anyone can go get a false id. And no-one is disenfranchised because someone stole their wallet last week and they haven't received their replacements yet.
Such a system would need to be massively perfect to catch the three people actually doing real voter fraud. That aside, filming people as they vote to see if they are voting the correct way, as people of thier backgrounds and neighbourhoods are expected to vote? Better get some great lawyers, if any would even attempt to defend such a thing.
Waiting in line is part of the voting process. Any sort of measurement of people waiting in lines will be a constitutional issue. What will the outcome be? A measurement saying too many poor people voted in an area?
There is also the MSM's tendency not to investigate, due to political correctness, the government's tendency to stone-wall to avoid looking incompetent, and the politicians' tendency to not pass laws that would hinder their chances of winning the next election.
One thing about voter fraud: a lot of people who complain about how much it's happening love to cite the existence of inaccurate voter rolls. Those exist in spades, and no one doubts that. But every election, we compile a list of everyone who voted, so it should be a very easy matter to translate inaccuracies in registration to actual instances of voter fraud. Note that instead of citing actual counts, there's almost invariably a citation of potential fraud--and that's because the actual counts do not bear out the claims.
Of course, the biggest reason that inaccurate voter registrations exist is because people don't contact the registries to tell them they've moved, and the government hasn't put much effort into trying to track down people when they move and actually update all the various registries. Cynically, one can suggest that the reason for the latter is that politicians would rather continue using the inaccurate voter registrations as an excuse to conduct voter intimidation than actually fix the problem.
It's definitely the case that, in Australia, when automatic registration was introduced it was introduced and supported by the left-of-centre parties who pundits suppose would benefit from having extra voters on the roll, and weakly opposed by the right-of-centre parties who pundits suppose would be hurt from having the voters on the roll.
But honestly, the idea that the government should track people down to put them in a registry strikes me as really unamerican, especially when you put it that way. I can imagine Democrats who would stand to benefit from it being like "That sounds yuck".
Very I nteresting work. Presumably the point of the announcement is to encourage agencies to fund more grant requests.