There’s a face on Mars.

Ever since Viking took this photo fifty years ago, some people have been sure–certain–that it clearly shows a face on the planet’s surface. Of course, once we had a high resolution image from a later mission, all resemblance to a face went away.

Human beings need a story, especially when we’re trying to understand something we haven’t already classified. And so we see faces in clouds, in grilled cheese sandwiches and on other planets.

We do it with song lyrics that don’t make sense and with technology we don’t really understand as well.

Some of drivers are:

Fear of the unknown.

Novelty and the arrival of something new.

Unpredictable inputs that seem to assert some sort of intentional action and agency.

It’s no wonder, then, that LLMs and other forms of AI lead to waves of pareidolia. We ascribe a gender, a tone of voice and most of all, intent to these computer programs that are doing nothing but math. We imagine that they are lying to us, manipulating us and getting ready to take over the world.

If imagining that there’s a little person inside helps you use the tool better, that’s fine.

But made up stories that we invented to deal with our fear often make it worse. They distract us from the hard work of understanding what’s actually happening.

When the details become more clear, we’ll then have to unlearn all the personification we insisted on learning.