Responsible Intelligent Systems


Remotely human-assisted artificial intelligence for self-driving cars?

Nissan’s Maarten Sierhuis promotes the idea that remotely operating human intelligence can fill in the gaps left open by artificial intelligence falling short in dealing with ‘corner events’ for self driving cars. The idea is that as soon as a self driving vehicle meets a situation that is too difficult to handle for its AI-mechanisms, a human takes over by remote control. I think this is not a viable idea, for several reasons. First of all, it complicates liability and responsibility questions. Second, I think it will not work. Who is going to decide on when too switch from AI control to human control? If it is the remote human, he will have to monitor the car all of the time, which means that hiring a chauffeur is probably a simpler idea. If it is the AI, the problem is only partially solved, if at all. And then there is the additional problem of the remote human intelligence having to grasp a situation instantaneously without having experienced or wittnessed the events that led up to it. Third, I think it is maybe not really needed. Indeed, there will always be situations where a human driver performs better than an AI driver. But these will be fewer and fewer and will eventually be vastly outnumbered by the situations where it is the other way around.

Nissan’s Path to Self-Driving Cars? Humans in Call Centers