An application of AI that I think will be possible in the near-term future, if there are not already people working on it: lie detection. There is a school of behavioral psychology that believes that people's behavior changes in subtle ways that betrays them when they knowingly tell a lie, even if it is not obvious to most observers. It is these kinds of subtle, hard-to-identify cues that machine learning excels at discovering and making use of, even if the AI has no profound understanding of the behavior that it is identifying.
One possible application: a political action committee has an AI continuously monitor C-SPAN video feeds of people in government together with witnesses called before them to give testimony. Whenever the red light on top of the AI goes off during a segment of the video, political operatives look further into the matter in order to dig up dirt on whoever was equivocating. Eric
