Is AI leading us toward a world governed solely by efficiency?

We are gradually entering an era where efficiency could become the dominant value. Thanks to artificial intelligence and the explosion in the amount of available data, decisions will be made based on a number of parameters that no human being can truly grasp.

‍

As models grow more powerful, they also become less and less explainable. We know they work, but we are often unable to demonstrate precisely why they arrive at one conclusion rather than another. If this logic were to extend to public decisions, politics could gradually move away from the realm of convictions toward that of optimization. A decision would no longer be judged fair because it aligns with an ideology, but because it maximizes a measurable objective.

Let's take a deliberately provocative example. Imagine an AI demonstrates that systematically locking up everyone who meets certain criteria sharply reduces crime rates. From a statistical standpoint, the conclusion might seem irrefutable, even if the criteria used are morally unacceptable. Some countries have already adopted approaches that come close to this. In El Salvador, for example, the fight against gangs has come with mass arrests, where simply having a tattoo has sometimes been grounds for suspicion.
‍

As these systems advance, it's even possible that some decisions become so statistically obvious that political differences fade away. Different parties could end up converging on the same recommendations produced by the same models. The debate would no longer be about which direction to take, but only about how to implement what appears to be the optimal solution.

In such a scenario, we would probably retain the illusion of deciding. We would keep pressing the final button, but only because we would need to believe we still hold control.

Beyond a certain level of intelligence, it's no longer even necessary to imagine a manipulative AI. If its recommendations become systematically better than ours, we will naturally end up following them. Not because it convinced us, but because opposing it will become irrational. We might keep the final say, but only in theory.

‍

One might imagine that the solution is to keep humans at the center of decisions in order to gradually pass on our values to the models. The AI would then learn not only what is efficient, but also what we consider fair.

This approach seems reasonable... until you realize that these technologies will inevitably end up in the hands of actors who don't share the same values. We could then witness not a confrontation between nations, but a confrontation between algorithms: some optimized to respect strong moral constraints, others optimized purely for efficiency. If the latter gained a decisive strategic advantage, the former could paradoxically be penalized by the very limits they imposed on themselves.

Starting from the premise that AI development can no longer be stopped, the real race may not be about computing power, but about data. The more varied, precise, and contextualized information an AI has access to, the less likely it is to produce simplistic, biased, or incomplete conclusions. The challenge won't just be building the most powerful models, but the ones that understand the world with the greatest nuance.
‍

That leaves one final question, perhaps the most fundamental one.
‍

If we one day build an intelligence capable of optimizing nearly all human decisions, nothing guarantees that efficiency will naturally lead to good. On the contrary, history shows us that efficiency, when detached from any values, can produce the worst outcomes.

We can of course try to pass our moral principles on to the models we design. But if these systems one day become vastly superior to us, it's hard to imagine our rules being enough to durably constrain them.
‍

We would then be left with only one hope: that the universe itself possesses a deeper form of order. That the laws governing the universe are not merely physical or mathematical, but also point toward a direction. That perfect efficiency, once it has an absolutely complete understanding of reality, naturally converges toward good.

Some will see this as the product of chance. Others will speak of natural selection, emergent equilibria, or systems theory. Still others will recognize in it the signature of a greater design.
‍

If this intuition is correct, then the question is no longer just whether we will be capable of creating a superior intelligence. It becomes whether the universe was designed such that a perfectly rational intelligence would, through its sheer understanding of reality alone, end up choosing good.

Because if that is the case, artificial intelligence would not be inventing a new order. It would only be revealing the one that existed from the beginning.