It’s very difficult to ignore these days, AI is slowly taking over the world. And not in the way you would see in science fiction movies, where suddenly robots are everywhere pulling the shots, telling us what do, and eliminating us if we have a mind of our own. No, it’s more in the things that we would not normally notice. Robots that are taking over the jobs of people in factories for example, or algorithms that decide which information is given or not given to us.
If you want to order or pay something on a secure site for example, you usually have to prove that you are not a robot by selecting images (so called captchas) and often this is easier said than done.
The instruction might be “select all images with crosswalks”. Then a puzzle of 9 or 12 pieces pops up with different roads with crosswalks in the fore- and background. Some puzzle tiles clearly show a crosswalk: those are easy. Others even have two crosswalks: walk in the park! But what to do with a tiny piece of crosswalk on a separate tile? Does that qualify as a crosswalk? It is indeed part of a crosswalk, but in itself I would not recognise it as one. And in the back of the picture there also seems to be something that might qualify; it could also be a bike path or a parking line, but who can tell? Doubt entered my mind, but there’s a clear time-limit to these tests. Nervously I clicked “yes”, but it was wrong. I didn’t pass my humanity-test.
On to the next puzzle: “Click on the tiles that show a cab”. Well, I would have to assume that these robots are not Dutch. If they would have been Dutch, it would be easy: all Dutch cabs have blue number plates, but these cars didn’t have blue ones. There was a yellow car in the picture and a black van. Yellow is probably a cab, but what about the black van? I often take taxi-vans and they are usually black, so I guessed “yes”, but that wasn’t correct. Luckily the next puzzle was one with stoplights and those are easy.
With these examples I would like to prove that it’s not easy to act as a real human in the eyes of a robot. People are unpredictable and take irrational and stupid decisions. Robots don’t do that. People feel pressured by time-limits. That makes them nervous and makes them take even dumber and more irrational decisions. Not robots. They don’t black-out if they see time ticking away and know they only have a few seconds left. They don’t click a button quickly, realising immediately: “shoot, the wrong one, I wanted to click on the other one”. They are not interrupted by a phone call when they are trying to remember what US number plates look like, that makes them having to take the test all over again. And finally they are very clever. They know a lot more than we do. And the people who designed them are also a lot smarter than the average person.
That’s what the robots should also consider when they take their decisions. Or does that happen already? Are they programmed to consider only pure rational decisions? Or do they take into account that average people are not as clever as the people who designed them? Do they consider nerves, black-outs, stupid decisions?
They should, but the more I think about it, the more I hope that they don’t. That would mean that they will never fully understand us, and I find that a comforting idea. As long as I fail my humanity test, I know it hasn’t happened yet!