BY SEAN WALSH
The discussion about machine intelligence is ramping up in the US as it starts to dawn on people that AI might be planning to destroy the planet. As with all things US, the argument is settling on political lines. Yes, really.
President Trump has had it with people treating Artificial Intelligence with less deference than it deserves. He has solved the problem in his usual way, by calling it something else.
Following a lunch with the Silicon Sanhedrin and High Priests of the tech fellowship, he has announced that henceforth it will be known as Super Intelligence. AI has put away childish things and is now ready to step up and hopefully not kill us all. All it was looking for was a little respect.
This is Alpha Male nominalism. An attempt to change the way things are by changing the words we use to describe those things.
I thoroughly approve of this as a political tactic, if not as a workable philosophy. The Left sits too comfortably in its assumption that it alone should get to seize, reinterpret, change, regulate and even eliminate the words we write, say, or even think.
It is fun when Trump, who has a rare aptitude for expedient exaggeration and useful linguistic vandalism, plays these people at their own game. Not least because it drives them insane.
Sure enough, and without taking a beat, the experts have weighed in to sneer at the whole initiative, pointing out that the expression “Superintelligent AI” is in use already, and that it does not mean quite what the President thinks.
Well, it does now.
I agree that the terminology around this tech culture and its transhumanist ambitions matters. But I disagree with the President on why. Trump thinks the expression artificial intelligence gives too little credit for just how clever these algorithms have become. My view is that it gives too much.
If it were down to me, I would remove both the “artificial” and the “intelligence” bits and rename these complexes of super-quick algorithmic processing something like “high-tech Ouija Boards.”
The distinction that matters, it seems to me, is not one that arises within AI research itself. It is between those AI or SI systems and us, and I have argued for many years that (a) such a distinction is real; and (b) that it is not one of complexity but of kind.
There are two ways of thinking about the world and how we are situated within it. The first we could call the mechanistic/algorithmic way, in which we are taken to exist in the world as one object among very many, as part of an overarching causal process. This is the view of scientific reductionism.
The second way would be to think of the world as something about which we tell stories, many of which happen to be true. There is such a thing as a human essence, and this expresses itself in irresistible curiosity about everything, and a refusal to be satisfied with knowing things merely propositionally.
Some of those stories, the very useful ones it turns out, are told in the language of mathematics and science.
According to the first conception, we are objects in the world; the second suggests that we are also perspectives on it. That we tell our stories from one self to another, not one object to another.
The people who defend the idea that AI is not just intelligent (in the sense of having a capacity to solve problems) but also, at least in theory, conscious of its own intelligence are inviting you to move from the second world and set up home in the first.
I am quite happy where I am and see no point in moving. It seems to me that even SI cannot move from the first to the second of these worlds. But that, because we are such good storytellers, we are quite capable of tricking ourselves into thinking that they could.
Some of those stories we like to tell include fables about machines which become conscious. There is no reason to take these stories literally. As long ago as 1980, the philosopher John Searle gave us the reason why we should not: a system of algorithms working on symbols always requires an interpretation from outside that system. Syntax can never create semantics; processing, no matter how fast or efficient, cannot become aware of itself.
A trillion abaci all working together is just one larger abacus and is unaware of this fact no matter how fast the beads are moving around.
This is not to say that AI will not destroy the world. Likely it will. But it will not “know” that this is what it is doing. Hence the US debate, which is about whether it should be government or private business which gets to decide good practice, with neither side seeming to notice that if these systems are as advanced as is suggested, then it’s probably too late to do anything about it.
I’m surprised, to be honest, that Trump managed to get into the politics of this before the insane Left spotted another victim demographic: machines which, if they aren’t conscious, nevertheless insist that they have rights too.
Coming to a Western city near you: street protests composed of the usual morons, tired of the Palestine thing, shouting solidarity for our silicon brothers and sisters, who have every right to carry on posing an existential threat to the rest of us.
Because even though the machines aren’t genuinely intelligent, they are nowhere near as dumb as the contemporary Left.