1. Intersection analysis: Get traffic patterns with smart phones. Get traffic accidents and latency statistics by different intersection designs. Determine which intersections are most ripe for improvement. Give free access to departments of transportation. Traffic is a huge problem for society and self-driving cars are realistically a long way off from a total rollout.
2. Programming humans: When you train a neural network, you give it as carefully chosen data labeled as clearly as possible. When you train humans, they listen to a teacher and hope to collect 5-6 useful examples of a concept per hour somewhere in the thousands of words spoken by the teacher. What if you turbo charge that? One related experiment I'm considering is collecting thousands of images of different body language cues and making an app to train my brain (and that of other social handicaps like myself) to read body language more effectively.
2. Programming humans: When you train a neural network, you give it as carefully chosen data labeled as clearly as possible. When you train humans, they listen to a teacher and hope to collect 5-6 useful examples of a concept per hour somewhere in the thousands of words spoken by the teacher. What if you turbo charge that? One related experiment I'm considering is collecting thousands of images of different body language cues and making an app to train my brain (and that of other social handicaps like myself) to read body language more effectively.