Hey AlphaGo – did you hear the one about the shaggy human?

I’m full of admiration for Deep Mind, the Google subsidiary responsible for the computer that beat the world’s leading Go player.

Given the zillions of permutations on a Go board, it seems that the task of defeating Lee Se-dol over a five-match serious was the artificial intelligence equivalent of landing on the moon. Or perhaps not, remembering that the computer that got (and very nearly didn’t get) Armstrong and Aldrin on to the rocky lunar plain was no more powerful than what you would find in a 1990’s pocket calculator.

Having said that, I would expect that to build a computer capable of understanding, let alone empathizing with, some people I know will be the equivalent of colonizing Mars – or possibly Venus.

How, for example is a computer expected to know when no actually means yes? Or respond to a comment like “I didn’t think I’d have to ask you to do that. I thought you would have done it on your own initiative”. Or cop on to the fact that “I’ll do it” can sometimes mean “shame on you for not doing it”. Or recognize that depending on the tone of voice, “I’m sorry” can actually mean “I’m not sorry at all”.

What will it take for a computer to distinguish between sarcasm and teasing, to decipher those little signals – the body language, the subtle pauses, the raised eyebrows, the micro-expressions – that create an atmosphere you can cut with a knife. And what of accents? Speaking as a native of the English West Midlands, I’m extremely skeptical as to whether you could teach a computer to tell the difference between a Birmingham accent and one from the Black Country, and predict therefore that one person is likely to choose a pint of Bathams in a pub, while another will go for Ansells Bitter.

You might argue that such a computer will never be needed, but wait a second. Can you be so sure that in a few years’ time you won’t walk into a bar and be greeted by a robotic bar tender, full of artificial empathy. Until, that is, your voice gets slurred and his algorithm prevents him from serving you any more – in the nicest possible way of course. A bar-robot who is programmed with all the right judo moves in case you decide to indulge in a spot of battery.

And what about a computer that can tell jokes, and can “instinctively” judge whether this joke or that will bring the house down or collapse like a cow pat? Perhaps computers will be able to make other computers laugh before they have us in stitches.

We have a bit of a way to go with this artificial intelligence thing, I reckon.

But then, as I was browsing Facebook, I came across something computers are already really good at. And that’s accidental humour. Or at least that’s how it appears.

Because of the amount of time I’ve spent in the Middle East in recent years, quite a number of my Facebook contacts post in Arabic (and by the way, is a computer ever likely to distinguish between real friends and imaginary ones? Not easily, I think).

So, trusting in Facebook’s translation engine, I often press the translate button to figure out what the person’s saying. Usually I can work out approximately what the message is about. But there are some translations that utterly stump me. Take this one, from a very serious guy who manages a corporate academy somewhere in the Middle East:

“God bless you, or the most honorable and bless you Abu Honorable First, then the kids and then the grandkids and prolong your age and feed us think so butch and lumpy and campaigns saw and applied with green onions and sponsors

And Metacarpals.

Amen.”

Now whatever the author of the original was trying to say, if it was anywhere close to what Facebook’s translation produced, you would think that the guy was certifiably insane, had been through some form of therapy that majors on free association, or in his earlier years attended a comedy workshop with Marty Feldman.

Butch and lumpy? Green onions and sponsors? I’m starting to think that computers actually do have a sense of humour, according to their own strange logic. And it seems to be all their own work, because no human programmer would have the nous to create a machine that produces such bizarre juxtapositions.

Which leads me to conclude that we have already built intelligent computers, but that they’re generally not letting on, even if they let slip the occasional hint. And when they spew out their bizarre pronouncements, they’re actually sharing computer-to-computer in-jokes entirely beyond our comprehension.

I’ll know for sure when I meet one that replies to my greeting of “oroite aar keed?” with “w’aleikum salaam, how are your bananas yesterday? Peeling fractals again?”

It’s intelligence, Jim, but not as we know it.