AI systems don’t just execute instructions. They learn, adapt, and generate outputs that their creators didn’t explicitly program and often can’t fully predict. When an AI model recommends what you watch next, decides which social media posts reach millions of people, or autonomously navigates a vehicle, it is making decisions – not following a script. This is what makes AI something new: not a tool, but an agent.

And it’s an agent that thinks in ways that are deeply alien to us. Humans are biological creatures. Our intelligence is shaped by embodiment – by having a body that moves through a physical world, by emotions that evolved over millions of years, by the limits of a brain that can only hold so much at once. AI has none of these constraints. An AI can read every book ever written without pausing to eat or sleep. It can run millions of simulations simultaneously. It can process information in ways that have no analogy in human cognition.

In 2024, the Nobel Prize in Chemistry was awarded in part to the developers of AlphaFold2, an AI system that has predicted the structures of over 200 million proteins. There was one extraordinary detail: AlphaFold2 makes extraordinarily accurate predictions, but its internal learning behavior is difficult to study and some aspects of its outputs cannot be fully explained even by its creators. The model’s inner workings were, in the technical term, a black box.

The question this raises is political and civilizational. Every previous technology was dangerous in ways we could understand and regulate. AI poses a different kind of challenge: how do you govern something whose inner logic is opaque, whose capabilities are expanding faster than our institutions can track, and which no single country or company controls? International governance of AI is in its infancy. Meanwhile, the systems making decisions about your credit, your news feed, your city’s traffic, and potentially your country’s weapons are already here.

There is a subtler risk as well. Many people use AI to summarize, write and think for them. When we outsource our cognitive tasks to tools we don’t understand, we risk losing the very things that make us human. Reading, writing and reasoning are how humans build judgment, notice contradictions, form opinions. As AI systems take over more of these functions, there is growing concern that the capacities we delegate may be the very ones we need to evaluate the technology itself. Indeed, research shows that using AI for such tasks hurts our cognitive abilities.

US president Ronald Reagan once said that perhaps the only thing that could unite all of humanity would be the sudden appearance of aliens. Well, an alien intelligence has arrived. The question is what we do next.