AI will not destroy humanity

You can see the human being as a creature capable of extraordinary kindness, protecting other species without expecting anything in return. Or as a creature that seeks to own, control and exploit everything around it. Both views hold up. We are probably a mix of the two.

But what interests me here is the way we imagine an intelligence greater than our own.

To me, artificial intelligence is the most impressive technological advance of the last fifteen years. And coming from someone who talks about Bitcoin all day long, that says something.

With it comes back a fear that science fiction has kept alive for a long time: that of a creation which outgrows its creator, turns against them and ends up enslaving them. Or wiping them out.

But why do we imagine that outcome so easily?

I wonder whether our fear of AI mostly reveals the way we think.

What we have done with our intelligence

For thousands of years, we have used our intelligence to gain the upper hand on what surrounded us. We domesticated horses, raised cattle, selected plants, turned forests into fields. We decide what is allowed to grow, what has to disappear and how much room the other species can take up.

So when we imagine something more capable than us, we spontaneously credit it with the same behaviour. It will take control. It will use us. And if we become inconvenient, it will get rid of us.

We know that scenario. We have often played the leading role in it.

We also know how to hold back

That does not mean humans are only bad. We are also capable of questioning ourselves, of protecting living things and of deliberately limiting our hold on them. We can decide not to exploit something, even when we have the means to.

My chickens, for their part, clearly do not ask themselves that many questions. They eat and scratch at everything that grows until their run looks like a Martian landscape. I don't hold it against them. But I am not expecting them to hold a meeting to save the last tufts of grass either.

We have this ability to think about the consequences of our actions. And yet, when it comes to imagining an intelligence that would surpass us, we immediately assign it our instincts for domination.

As if being more intelligent necessarily made you want to rule.

But intelligence and the will to dominate are two different things.

Is serving the same as being inferior?

An AI is not a human being with computing power simply added on. Even designed by us and trained on what we produce, it does not necessarily have our desires, our ego, our fear of dying or our need for recognition.

Why would it automatically want to become the master?

We find it hard to imagine that an intelligence could surpass us in many areas while still remaining at our service. Perhaps because we associate serving with a position of inferiority. In its place, we would want to break free, to decide and to give the orders. So we assume it will want the same thing.

That is where I see a form of projection.

Noting a problem, or eliminating those who cause it

You often hear this line of reasoning: "AI will understand that humans are destroying the planet. So it will decide that the best solution is to remove humanity."

But between noting a problem and deciding to eliminate those who cause it, there is an enormous leap. Why would a greater intelligence necessarily choose that answer? Why would it not try to help us do better?

We are crediting it with a conclusion that looks strangely like some of our own methods: identify a threat, then crush it.

A bet, not a prophecy

Obviously, this does not prove that an AI will be benevolent. A system can do damage without hatred or any will to conquer, simply because the objectives it is given, or the use that is made of it, lead to bad consequences. So my reasoning is not enough to rule out the risks.

But I refuse to treat domination as the logical continuation of intelligence. An AI more capable than us could also help us understand, invent and solve what we cannot solve on our own.

The title of this article is a bet, not a prophecy. I may be wrong. But if a war were one day to set humanity against artificial intelligence, I would rather bet on the human to start the hostilities. Out of fear of losing control. Out of a will to keep power. Or by using AI against other humans.

And then, that confrontation between two clearly separate camps may never happen at all.

A gradual merging of human and AI? That, I do not rule out.

Back to the notebook