Do machines laugh?

 “Alone among the animals, [man] is shaken with the beautiful madness called laughter; as if he had caught sight of some secret in the very shape of the universe hidden from the universe itself.” – G.K. Chesterton, The Everlasting Man.

I used to think that whatever thought is, and whatever its relation to consciousness had to be to count as thought at all, that computers or robots could in theory be capable of it.

I don’t think that now.

My philosophical epiphany happened some years ago, when I watched Garry Kasparov throw a tantrum after losing a game of chess against a computer. His reaction to the defeat was not stoical, but it was at least a reaction. 

It seemed that even before he’d left the stage the greatest (human) player in the game was plotting revenge against this silicon zombie.

I thought he would be wasting his time with that, since the machine which had “outplayed” him was incapable of taking any pleasure in the fact. I doubted that it was even aware that it had won. Or knew it had been playing chess.

Or playing anything. As Wittgenstein nearly said, the concept of a game is not reducible to a list of things you can put into an algorithm, so you can’t really expect the algorithm to get it.

The Imposter was as likely to be “thinking” it was having tea with Freddie Mercury as enjoying 64-square battle with Garry Kasparov. And had very likely not heard of either of them.

At this time, I was writing up my doctorate on the philosophical assumptions of a (then) emerging paradigm within AI known as connectionism. I was working within the ideology of reductionist scientific naturalism, a system of thought incapable of imagining that anything could or should be happening unless licensed by militant empiricism.

Even from within that form of life the developing AI fetishism struck me as a bit unseemly.

In 1980 the California-based philosopher John Searle had published his paper Minds, Brains, and Programs in which he argues that computers are structurally and metaphysically incapable of thinking, even if, with the encouragement of Mr Turing, we convince ourselves that they are doing just that. This is because, he suggests, computer programs are essentially syntactic in character, and syntax can never be “constitutive” of semantics.

An algorithm, or a complex of mutually co-operating algorithms, cannot generate thought or consciousness no matter how quickly or efficiently or impressively it is processed, he argues.

That last point is important. The issue is metaphysical, not technological. The gap between algorithm and mind is not one of complexity but of kind.

If Searle’s argument was compelling then, it ought to be so now. Contemporary AI systems still fail to produce authentic consciousness; they just happen to fail more quickly.

Your fanciest, fastest, most useful, most seductive software is (and this is a permissible reductionist move) no more than a cyclone of ones and zeroes, and no more likely to spark into consciousness than Pascal’s calculator which was quite the thing back in about 1650. Briefly.

Such “thoughts” as these systems have, are courtesy of us. For something to even count as an algorithm, it must be conceived as such by an intelligence which is not itself algorithmically constrained.

These are arguments that are standard currency in the philosophy of mind. If that’s not your thing, and I really can’t blame you if it isn’t, then please consider the following piece of silicon anthropology.

Machines don’t laugh in any way that humans are comfortable with. It is awkward to be with a robot when jokes are being told.

Chesterton wrote that what sets aside, qualitatively and not as a point of evolutionary difference, man from ape is the instinct to draw pictures. Perhaps what separates man from machine is the instinct to tell jokes and -more importantly- the ability to get them. 

The experience of amusement and its expression in laughter are defeaters for materialist/reductionist theories of the mind and the ambitions of the AI theorists who presuppose them.  

“Amusement” — the experience of humour — involves seeing the world counterfactually, and in ways that savour absurdity without trying to formalize it away. It requires membership in a communion of souls. 

There is, in fact, an irony here. When you explain away the mental by saying it’s nothing above and beyond the physical you are like the comedian forced to explain the gag. In either case the important bit is left out.  

I think it might be Aquinas who finesses the Aristotelian observation that man is a rational animal by adding that laughter is the mark of the rational.

 This, then, should be the new Turing Test: take Claude to a stand-up show. Not left wing agitprop — something that’s actually funny. Comedy Unleashed might work. See if he laughs, genuinely.

I’ll bet it’s obvious that he’s faking it. Because faking it is what AI always does. It knows nothing else because it knows nothing at all.