AI for Good at datascience.com
I went to a PyData meetup tonight, “AI for Good,” at datascience.com’s headquarters. After an initial explanation, we classified homes and related structures – manually – , as part of the Humanitarian Open Street map project to assist Hurricane Maria victims in Puerto Rico. (Visit the task manager here: http://tasks.hotosm.org) Having a room full of data scientists, developers, and other interested people clicking, dragging and tagging buildings was mundane, BUT how can we improve the system if we do not understand it?
At the end of the meetup, each group talked about suggested improvements. In our group, one gal mentioned training a data set to recognize polygons, since that is most shape of what we labeled. Another person mentioned that the training set for most of our ideas was the entire city of LA on Google maps…. Hopefully, there will be progress going forward. While I do not have any aspiration to become a data scientist, I do think it is important to understand the process, and of course, how to clean and manipulate data, as that is an important skill regardless of the setting.