When AI mistakes rocks for people

A recent report suggested that visitor numbers at the Giant’s Causeway had increased significantly. The reason wasn’t a sudden surge in tourism, but something more technical. An AI system used to count visitors had been mistaking parts of the landscape, including rocks, for people.

It’s the kind of story that’s easy to laugh at. The mistake feels obvious once you see it. But it points to something more subtle about how these systems are actually used.

Tools like this are often brought in to track footfall, monitor usage, and give a clearer picture of what’s happening in real time. Once they’re in place, the numbers tend to be taken at face value. In this case, the output looked reasonable, even though the interpretation behind it was wrong. Nothing obviously broke, so nothing was questioned.

That’s where it starts to matter. These numbers don’t just sit in reports, they shape decisions. Funding, staffing, planning, all built on data that looks solid but might not be. A small error at the point of collection can quietly ripple into much bigger outcomes.

It also highlights something people tend to forget. AI doesn’t understand context, it works from patterns. A human would instantly spot the difference between a person and a rock in that setting. The system doesn’t “know” that difference, it just matches what it sees to what it expects. When something falls slightly outside that expectation, the mistake isn’t always obvious.

The issue isn’t that AI gets things wrong. That’s expected. It’s that the mistakes don’t always look like mistakes. They look like normal, usable data.

Which leaves a slightly uncomfortable question. If we’re relying on systems like this to tell us what’s happening, how often do we stop and check what they’re actually seeing?