Now this knocked me off my seat. Ever notice the automobiles people drive in your town or neighborhood? Using the same concepts that enable Siri, Alexa and Google Assistant to understand what they hear you say, scientists are designing computer systems that can understand what they see. It’s called “Deep Learning,” and it’s a form of machine learning, which falls under the broader umbrella of Artificial Intelligence.
Anyway, scientists from Stanford, Baylor and Rice Universities and the University of Michigan used images from Google Street View, 50 million of them, to infer the answers to questions about their communities and neighborhoods such as income, race, education and voting patterns. Specifically, they identified cars parked on the streets photographed by Google Street View cars and matched that with existing census and other survey data. The hard part, that required “Deep Learning,” was getting the technology to identify the make, model and year of all motor vehicles encountered.
One thing that they learned is that a neighborhood where pickup trucks outnumber sedans is 82% more likely to vote for a Republican in the next presidential election. Where sedans outnumber pickup trucks, 88% more likely to vote Democrat. So what do SUVs mean? ..and what about people with garages? ..and what television networks run the most pickup truck commercials?
I’ll be really interested when their computers can identify bicycles. 😉
Source: Proceedings from the National Academy of Sciences of the USA (https://goo.gl/is2nSX)