Wednesday


What is the least thing one might expect from an AI program? As far as creativity goes, drawing a connection between politics and pickup cars is nothing short of far-fetchedness, or so it seems. But, the truth is, that is exactly what a recent study published in PNAS touched on. No matter how weird you think this conclusion is, as you read this article, you will soon find out that the principle is actually quite simple, yet powerful enough to predict voting results!

As Artificial Intelligence technologies continue to infiltrate every aspect of our lives, it was only a matter of time until we see an AI program that predicts people's voting decisions. At first glance, the relationship between politics and AI seems tangential at best. However, upon a deep analysis, this association makes perfect sense, since both government policies and Artificial Intelligence applications heavily make use of data. For instance, the US government conducts a yearly nation-wide survey (i.e., American Community Survey) to collect socioeconomic data such as jobs, races, education, gender, and so on. The goal of the ACS is to derive an efficient way to distribute federal and state funds on a yearly basis. However, one problem with such approaches is that they need a lot of resources, and the data they gather don't generally reflect reality in real-time, as this information usually has a one year lag behind the "actual" data.

This is precisely what researchers from Stanford University, the University of Michigan, both Houston's Baylor College of Medicine and Rice University, tried to solve using deep-learning algorithms. The team used publicly available data from Google Street View cars to accurately determine income, race, education, and voting tendencies among others. Through the analysis of 50 million Google Street's images taken from 200 cities, the scientists were able to establish a surprising connection between cars and voting choices. For instance, analyzing Google Street's images extracted from cities where the number of sedans is higher than the number of pickup trucks, showed that the underlying city is likely to vote for a Democrat candidate in the next presidential election, with a 88% chance. Conversely, cities with a higher proportion of pickup trucks will likely choose a Republican in the next election, with a 82% chance.

Studies like these might have huge implications on our lives in the futur, as governments will be able to gather different forms of data in a quasi real-time fashion, which is both fascinating and scary at the same time, since these data could have an impact on our privacy. The bottom line is that calling Trump mister pickup truck man might be the perfect comeback Kim Jong-un could use in response to the former calling him "rocket man"!
Back To Top Icon