Ayanna's perfect robot choices

We have to choose. Choosing tires us. It exhausts us. It makes us waste a lot of energy. We hate to decide. Because we know we'll never have enough data to do it. We need more data to make perfect choices. So we're desperately out to create perfect machines, endowed with perfect intelligence to help us make even more perfect decisions. And this would be perfect thinking, if it wasn't completely wrong.

Do we value perfect decisions that much?

Ayanna Howard, a robotics engineer at the Georgia Institute of Technology has concluded that humans do not trust robots that are too perfect. The first prototype of a robot designed to emergency evacuate people from a burning hospital proved to be a failure simply because it was too perfect. The robot, which used real-time data from the fire alarm system, never made a mistake in the corridors it chose. The most surprising thing was to realise that this perfect robot did not deserve the trust of the humans who were following it. Ayanna's team then tried to codify some of the mistakes that the robot would make. When it realised it had made a mistake, it would apologise and move on to the correct path. Humans trusted this second version much more. Perhaps because they recognised things that we humans are masters at doing? Like making mistakes? And apologising when we make mistakes?

Our disquiet about perfect choices and the need to create another intelligence that would help us decide, began the moment science revealed to us how we decide after all. We realised that, contrary to what we imagined, we cannot make any decision without involving our emotions. And the problem is that human emotions are anything but perfect. We also realise that others impact almost all of our choices, because we are social beings and we will always want to be part of the group. And the problem is that others, as we well know, are anything but perfect. We thus realised that our choices almost never result from a rational and conscious process. And that scared us. And it still does. It was more or less like concluding that the unusual decision-making method of the ancient Persians, so well described by Herodotus, was far more sophisticated than it appeared to be. They made the big decisions by discussing them twice. The first time, drunk. The second time, sober. If, when sober, the decision was the same as the one they had taken when drunk, it was maintained. If it was the opposite, it was rejected. Does this original method of making choices help explain the hegemony of the Persian civilisation for so many centuries?

But would Ayanna's first prototype be a truly perfect robot?

In our human eyes, yes. We truly believe in the perfection of algorithms. But we forget one important detail. That the data that feeds the deep learning of artificial intelligence is our data. That data that we have produced for decades and that holds all our unconscious human prejudices and contains all our stereotypes of gender, origin, race and social status, which today are at risk of being amplified by machines. When faced with the blurred image of a person in the kitchen, the algorithm identifies a woman, and behind the wheel of a car, a man, our hopes of having another intelligence at our side to help us make perfect decisions begin to fade.

In the midst of so much uncertainty and faced with an infinite number of options, of one thing we are sure. We will make mistakes. We will make wrong choices over the coming period. Many. Lots of them. Us and the algorithms. Yes, they will continue to help us make mistakes too. But maybe that's not bad news after all. If we can accept that our choices are a healthy mix between the ecstasy of seeing a colourful wheel of fortune spinning in front of us and the serene melancholy that invades us when the wheel stops and reveals the result of our choice. A kind of fusion between Damien Hirst'scircle spin paintings and Edward Hopper's thoughtful characters.

And then, perhaps, we will finally realise that the greatest illusion of all is the illusion that our life would be perfect if our choices were perfect.

We hate to decide. We hate to choose. Because we hate to lose. Every time we choose, we lose something. We lose everything we don't choose. And we don't want to lose anything, because we always want everything. If we learned to lose, maybe we wouldn't be so restless when it comes to choosing. Maybe we wouldn't be so afraid of deciding. Because in case we made a mistake and walked into a burning corridor, all we would have to do is make like Ayanna's robot. Apologize. And keep walking.