The world is buzzing with the claim that computers “can do X better than humans.” So then why did ShmooCon use animal picture identification to throw off bots during their most recent ticket sales? We’ll look at how computers are learning and more importantly how they are not learning… or alternatively how woodchucks are like moose. A simple framework will be introduced to help the you understand how a computer approaches problems. Then we’ll continue by examining potential computer security applications of machine learning to prepare you to evaluate its usefulness the next time you encounter a product that promises a “deep learning” or “AI” solution to all your problems. Throughout the discussion we’ll recognize the large datasets needed for machine learning algorithms to perform reliably and the resulting implications on privacy, whether it is your personal photos or your corporate intellectual property.
Samantha Livingston is a software developer in Boston, MA. Her interests include incorporating, exploiting, and promoting the use of exploratory data analysis tools in the design and development of robust software solutions for real-world problems.