Will self driving cars be able to psychologically surpass humans?

By Joey Merkin

Can we teach a computer to read minds? Perceptive Autonoma thinks so. Well, maybe not actually mind reading, but their concept still involves predicting a person’s next action.

As any driver knows, it is often hard to determine what someone else on the road is going to do. To do so, Perceptive Autonoma utilized the skills of their machine learning experts and began studying the behavior of pedestrians, cyclists, and other drivers involved in traffic situations. They they conducted studies where they showed respondents video clips and frame grabs of different pedestrians going about their lives on the streets and asked the respondents to utilize their human qualities to predict the behavior of these pedestrians. Perceptive Autonoma then used these responses to train their deep learning algorithm to think as they predict a human would.

As their video demonstrates, Perceptive Autonoma was able to successfully teach their systems how to read and predict the actions of unpredictable humans and can now avoid many unnecessary stops and accidents in self driving cars. The final product? A visual system with little thought bubbles above everyone in the frame displaying two values: a measure of whether or not the pedestrian or cyclist is aware of the autonomous vehicle and another scale which predicts their intentions of crossing in front of the vehicle in question. Working in tandem with the lidar and other warning systems in the car, this data can then be interpreted by the car and allow it to think like and communicate with humans much more effectively.

Perceptive Autonoma received a hefty investment from First Round Capital and with their help were able to implement their successful software in autonomous vehicles around the world and as self driving cars progress, hopefully their technology will become increasingly more prevalent in our vehicles.