In the last few weeks there were two pieces on how we as a mankind should be weary to give too much control to artificial intelligence (AI). One by the The Guardian and one by the director of the Rathenau Institute. And since it is very healthy for our society to have a discussion on this subject, and because I think both pieces are missing the point, I will give a short commentary.
Plane crash caused by automation?
Both authors write about Air France flight 447 that crashed in 2009. On that flight the automation that assisted the pilots malfunctioned for a short while, forcing the pilots to take over the controls. Because the pilots were not used to flying manually anymore, they could not identify what was wrong and tragically crashed the plane. This case, the authors claim, proves that too much dependency on AI is dangerous. Because when we are dependent on technology too much, we cannot take control anymore when we are in an emergency situation. Tho decrease this dependency they propose we should find a hybrid form of human/AI control, which sounds to me like the worst of two worlds, in stead of the best.
I would say, the argument should be made completely the other way around. The automation on the plane only had a temporary malfunction, it worked again after a short while. It was the human factor that eventually caused the plane to crash. If anything, this is an argument for more automation and less dependency on humans.
Just like a lot of people I have a washing machine at home. Every few days I fill it with clothes and detergent, push some buttons, and after 2 or 3 hours I collect clean clothes. It will probably break someday, or the onboard computer will do something I did not intend it to do. Leaving me ultimately without clean clothes. This would be unfortunate, and could be seen as me being too dependent on technology. But I would never want to go to a situation where I would have wash my clothes by hand, because that is so much more boring, time consuming and dangerous. The same can be said about the airplane crash. Although it is exceptionally tragic, as a society we would not want a situation where all planes are flown more or completely manually.
So are The Guardian and the director of the Rathenau institute completely wrong? Should we automate everything? Not completely. We should definitely automate everything that technology can do better, and that is boring, dangerous or time consuming. There is nothing wrong with that, and we have been doing it for millennia, using technology to make our lives easier. But we should not automate activities that are fun, or that define us as human. For instance, automating human communication with AI is in the long term far more damaging to us than automating cumbersome chores. So when discussing the dangers of automation, do not fall into the trap of resisting technology, because it automates tasks, but be skeptical when it automates tasks that make us who we are.