AI Afraid of Us?

AI Afraid of us? Well, that sounds strange in this day and age where the fear of AI seems to be prevalent.

I mean, AI has beaten the very best of humans in Chess, Poker, and Multiplayer Games. It has generated digital human faces that are very hard to tell apart from real faces, and it can even write passably good articles!

Of course, this would all inevitably lead to singularity, and the robot uprising. Which would ultimately lead to AI taking over the world, and the enslavement of all mankind. Isn’t that right?

So: AI Afraid of us, that just sounds crazy. Or does it?

But to some extent, it’s true! However strange that sounds. I mean, it should be the other way around, right? And us mere humans should be fearing our AI.

Well. That may not be the case. Many researchers say that AI’s are good at narrow tasks. Yes they are impressive at chess but that’s because when compared to the real world, chess has very little defined rules and variables. In fact, it took another 18 years from the chess victory for AI to beat humans at GO. Now I have to admit I don’t know how to play GO but that should give you some perspective.

When it comes to more complex games. Well let’s just say if you’ve been playing any mainstream games at all and came up against AI, you know how far AI still has to go just to not be utterly embarrassing and hilarious. In fact, despite trying for seven years, AI researchers can’t get their AI to beat humans at Angry Birds. I’ve put a link to the source article in case you want to read up about that.

So while it is true that AI have recently beaten humans at an open world game. But that game is a simple capture the flag with narrow sets of rules. Or so they say.

Like I said earlier, when it comes to mainstream and more normal open world games, AI just falls flat.

And in many turing tests, AIs just can’t seem to understand human instinct and emotions enough to mimic them in prolonged sessions. 

Ok, so hopefully AIs and their robotic army won’t be taking over the world any time soon. But AI being afraid of us? Now why would that be?

Well I am referring to a recent research carried out which tests how empathetic humans are towards AIs. So in this test, researchers tried out an AI powered counselor on real humans. By the way, it’s important to note that humans  know they are interacting with machines.

Yes, I know it doesn’t sound very good. And human psychiatrists can rest assured that they won’t be losing their jobs to AI any time soon because by all accounts AI psychiatrists are more likely to cause trauma than cure them.

But that’s not the point of the research. In this case, researchers were trying to gauge human empathy towards AI psychiatrists.

The result? In all cases, humans show woeful lack of empathy towards AI. I mean, people interact with dogs better than when they interact with AI!

Even when AIs showed emotions, albeit fake and simulated ones, humans in the tests just seemed to discard and disregard it entirely. Even when the AIs in the test displayed sadness or simulate being hurt by humans in the tests, humans just don't seem to care. In comparison, similar tests show that humans respond with much more empathy towards dogs and other animals when they display some sort of sadness or distress.

So to summarize the test results: humans are just really, really, mean towards AI.

While there may be some real world implications of this test. The example given is that people may be more likely to rudely cut in front of self-driving cars powered by AI. The real takeaway is that people just act rudely and badly towards artificial intelligence. 

But AIs don’t mind that we are mean to them, do they? I mean they are just thoughtless, senseless machines with no feelings, aren’t they?