This Simple Idea Explains Why AI Risk Snowballs So Fast

A feedback loop is when something causes more of itself. Think of a snowball rolling downhill, picking up more snow and speed as it goes. These loops drive explosive growth — doubling over and over — unless something stops them. The author argues these loops sit at the heart of our biggest risks, and that spotting them is a useful skill for anyone trying to understand how disasters unfold.

The danger isn't just how big these problems get — it's how fast. When a process accelerates, humans get very little time to adapt. The essay points to several examples: AI that can improve its own intelligence, pandemics where each person infects more than one other, climate change where melting ice makes the planet absorb more heat (which melts even more ice), and nuclear chain reactions that release energy in an instant.

Here's the useful part: if you can identify a loop early, you can often find a single point where a small fix breaks the cycle. That's far cheaper and safer than trying to stop a runaway process later. The author suggests comparing different risks by asking the same question — where's the loop, and where can we cut it?

The tricky part is that real-world loops are messy. Money and power don't always snowball in predictable ways, and human societies are complicated. But new technologies like AI, social media, and even life extension could make certain players far more powerful, faster than before. When a loop turns dangerous, the window to act closes quickly — which is why noticing the pattern early matters.