Alcohol-impaired driving fatalities

It is a curious feature of the universe that as researchers have grown increasingly enthusiastic about training machines to think, humans have grown correspondingly worse at the relatively simple task of operating vehicles while sober. Between 2010 and 2022, these two phenomena climbed in near-perfect synchrony, as though every neural network paper published to arXiv somehow licensed someone, somewhere, to drive home at midnight with impaired judgment. One wonders if the correlation works in reverse, or if it simply reflects humanity's profound commitment to being wrong about two completely different things at the same time.

The explanation, warmly, is almost certainly that both trends follow the same underlying current: economic expansion. As the economy grew during this period, more people drove more miles (creating more opportunities for fatal impaired driving), and simultaneously, more funding flowed into AI research, which meant more researchers publishing more papers to arXiv. It's rather like noticing that pizza consumption rises alongside drowning deaths in summer—both spike because more people are outside, doing things, including some things they shouldn't. A specific example: between 2010 and 2019, US miles traveled grew roughly 25 percent, while arXiv submissions in AI categories grew from around 3,000 papers annually to nearly 15,000, both riding the same wave of economic confidence and technological investment.

What we are looking at, then, is not a universe where artificial intelligence causes drunk driving, but rather a universe where two entirely separate human endeavors happen to rise and fall together, perfectly correlated, perfectly meaningless. This should trouble us slightly, not because it suggests a hidden mechanism, but because it reveals how readily our pattern-seeking brains will bind two strangers together with statistical ribbon and call it truth. The real correlation is probably just growth itself.