An artificial intelligence system “guesses” what the neighborhood is voting for
How many people in the neighborhood vote the same party as you? How many have a college degree? How Much Money Does Everyone Make? You would love to know, but unfortunately you can only guess. But a new artificial intelligence system for the first time is able to guess the answers to all of these questions quite accurately.
To accomplish this, it is enough to supply the relevant street photos algorithms with Google Street View and they then do their “magic” by making estimates of political preferences, educational level, income level and .a. for every neighborhood. This confirms the saying “you are what you drive”.
The researchers, led by Associate Professor of Computer Science Fey-Lee, director of the Stanford University of California’s Artificial Intelligence Laboratory, published the relevant issue in the 50th US Journal of the Journal of the American PAS Journal. , which were pulled by Google vehicles in 200 US cities.
Computer engineers used deep learning and computing algorithms to determine the brand, model and year of manufacture of the approximately 22 million cars in the photographs, which was accurately achieved for half the vehicles (52%). In the next stage, other algorithms were fed with demographic and electoral data and undertook the estimation from the standpoint of cars alone. Artificial intelligence proved unbelievably accurate in its estimates of the average household income in each neighborhood, the proportion of white, black and Asians living in each district, the level of education of residents, and how many people voted for the Democratic Party. and how much for Republican McCain in the 2008 presidential election.
In other words, it is possible to evaluate how progressive or conservative an area is, depending on the type of vehicles you see most often parked there. For example, the more sedan cars a neighborhood had, the more residents voted in favor of Obama, while the more pickup vehicles in an area, the greater the percentage favoring McCain.
If the number of sedans in a neighborhood were greater than the number of “pickups”, there was an 88% chance the neighborhood would vote Democratic. If the “pick-ups” were to exceed the salaries, then there was an 82% chance that the neighborhood would vote Republican.
Researchers said that in the future, with the increase in street cameras thanks to on-road vehicles, more and more data will be collected and an artificial intelligence system will be able to draw near-real-time conclusions.