Why Those AI Doom Percentages Are Basically Made Up

A viral post skewers the trend of assigning precise-sounding odds to AI wiping out humanity, arguing the numbers reveal more about vibes than evidence.

Every few weeks, someone puts a tidy percentage on how likely artificial intelligence is to end civilization. Ten percent shows up a lot, as if risk assessment for a technology nobody fully understands can be pinned down like a weather forecast. A recent post making the rounds pushes back on that habit by simply flipping the number: if there's a 10% chance of doom, doesn't that also mean there's a 90% chance of something closer to utopia? The post doesn't argue AI is safe or dangerous so much as it mocks the confidence baked into these figures in the first place. Estimates like this get repeated by researchers, executives, and commentators alike, often without much explanation of where the math comes from. Treating them as settled science, rather than educated guesses or gut feelings dressed up in numbers, is where the joke lands. To drive the point home, the post leans on a bit of borrowed comedy: a nod to Homer Simpson's line about how people can twist statistics into proving whatever they want, since, after all, forty percent of everyone knows that. It's a silly citation, but it underscores a real frustration — that headline-grabbing percentages about AI risk often function more as rhetorical tools than as rigorous forecasts. None of this settles whether AI poses a serious existential threat or is overhyped hand-wringing. But it's a reminder, especially as AI video and generative tools accelerate into daily life, that big scary or big rosy numbers about the technology's future deserve the same skepticism as any other statistic pulled out of thin air.