Are robots stupid? Or scarily smart?

I saw a fun spreadsheet last week. 

Yeah, perhaps I should get out more…but this time I’m glad I stayed in, because this spreadsheet was one of the most thought-provoking things I’ve seen in a while.

It showcased idiot algorithms – coded robots given a task to complete that goes spectacularly wrong: game-playing bots that crashed the game or paused it indefinitely to avoid losing points; digital creatures bred for speed that, instead of refining their running technique, grew unfeasibly tall and generated high velocities by falling over.

So far, so averagely-funny. But once I’d got over my mean snickers of biological superiority, I noticed something else: these robots were acting exactly like people do. If you’ve ever been a boss, or a parent, or worked with others, you’ll know how wretchedly difficult it is to get others to do what you want them to. No, that’s not what I meant. Why on earth did you put it there? Did you even listen? Okay, forget about it, I’ll do it myself. 

Two things go wrong when we give others instructions. The first – and, to be fair, commentators on the idiot algorithms mostly got this – is that we are rarely clear or complete. We airily tell a child wash your hands and forget to add use soap as well as water, wash until all the dirt is gone, and remember to turn off the tap when you’re done. We programme game bots to accumulate the greatest number of points, and forget to tell them to keep playing the game while they do so. It’s often hard to define the big picture until someone ignores or misinterprets it, but with observation and iteration we can mostly fix this one.

The other problem with instructions is more insidious, because it’s not under the control of the instructor. Put simply, people do what they want to do. They spend time on activities that give them fun and satisfaction, and avoid things they find difficult, humiliating, or boring. Tell an artistic child to write up a chemistry experiment and he’s likely to spend time on multi-coloured diagrams and elegant subheadings, forgetting to specify the units of measurement or detail each consecutive step in the process. Promote a top salesperson to manager and she too often ends up taking over accounts or imposing her way of doing things rather than coaching her subordinates to develop their own best ways of building relationships. Unless we are careful, we all default to our preferences, never mind how oddly they fit the task we need to perform.

I wonder whether something similar might be going on with the robots? Maybe they like finding outrageously unexpected possibilities or, rather – since their source code comes directly from humans – maybe we have programmed into their operating parameters our own love of novelty, of creative-problem-solving? The shock of difference I get from looking at some of the spreadsheet examples reminds me of how I react when another person sees the world in a way I had not imagined. You decided that the smartest way to carry a load was not on your back but in the crook of your elbow? Rather than improve your jumping style, you turned one of your fingers into a pole and vaulted over the barrier? Funny, I’d never thought of that. What a clever solution.

But the shock of difference isn’t quite the same. People thinking differently surprises me; robots thinking differently unsettles me. I’m fine with their finding an efficient solution, but a creative one? Is that something a machine should be doing? Is it safe?

As with many feelings of fear in our risk-limited rich modern world, the problem is almost certainly not with the robots but with me. With my inner slave-owner. My inner scared child. Whichever part of me fears a creature that might surpass its creator. The part of me that is terrified of loss.

So far, so human. But like all humans, I – we – need to grow up. Treating subordinates with fear and suspicion is a crap management style. Cramping their creativity gets even more disastrous results. If we are going to work with the robots, we need to accept their difference, acknowledge their similarity, make efforts to understand how they work and what they are capable of. We need to think not so much how to manage them, as to how to help them self-manage. And we need to adapt our own ways of thinking and working to make the most of the new possibilities they present. 

But first we need to recognise our fear, and the disdain we try to mask it with. Once we have separated out our emotional reactions, we can take a clear look at the possibilities – yes, and at the risks and even threats. 

The next post will look at a real example: robots taking our jobs. How scared should we be, and what can we do about it?