Fatal dog attacks in the US

It is a curious feature of the universe that as artificial intelligence researchers began publishing their findings to arXiv with the enthusiasm of people who had just discovered fire, American dogs developed a corresponding appetite for statistical significance. Between 2010 and 2023, these two phenomena moved together so reliably that one might suspect the dogs were reading the papers, or perhaps that the papers were somehow summoning the dogs. Neither explanation is remotely true, which is precisely what makes this so troubling.

What we're probably watching is the simple physics of growth: the US population increased steadily across this period, meaning more people, more dogs, and yes, more of both tragic incidents and excited researchers. Factor in the AI boom itself—that genuine explosion in machine learning funding and academic attention that began around 2012—and you have a field expanding its output at roughly the rate that everything else was also expanding. Add to this the fact that fatal dog attacks, while rare (averaging perhaps 43 per year), fluctuate wildly based on reporting standards, breed discrimination panics, and whether a particularly notorious incident gets media oxygen, and you've got two datasets that are essentially riding the same economic and demographic swell like boats in a rising tide.

The correlation is real enough to be unsettling and meaningless enough to be instructive. We live in a world where two completely unrelated things can move in perfect synchronisation simply because they're both caught in larger currents we barely notice. Perhaps the real lesson is not that AI and dog attacks are connected, but that we are all just drifting together.