Should You March Against AI? One Analysis Says Maybe Not
A 3-hour protest might do less for AI safety than you think.
On November 21, PauseAI UK will hold a March Against AI Extinction in London. The goal is to raise awareness about the risks of artificial intelligence. But if you're thinking of going, you might wonder: does my presence actually make a difference? One writer did a back-of-the-envelope calculation to find out.
The analysis starts with the idea that protests are useful because they generate media coverage. But how much does one person add? With an expected crowd of 670, each marcher increases the crowd by only 0.15%. And since media coverage grows slowly with crowd size, that translates to a tiny bump in coverage—just 0.073% more for that march. Then, considering that this march is only a small part of all AI-risk coverage (about 0.9%), the total impact of attending is a minuscule 0.00066% increase in overall coverage.
So how does that compare to other ways to help? The writer estimates that this tiny coverage boost is worth only a few hours of direct AI safety research. In other words, if you have three hours to spare, you might contribute more by doing research than by marching. The analysis uses a model where impact equals 1.5 times the log of research hours plus the log of press coverage, suggesting that research has a more predictable payoff.
However, the writer acknowledges several caveats. The sample is small (only seven protests), and the estimates are rough. Media coverage is also unpredictable—sometimes a small rally gets a big outlet, sometimes a large one gets ignored. The analysis also doesn't account for the intangible benefits of protesting, like community building and personal motivation. So while the math suggests marching may not be the most efficient use of time, it's not a definitive answer. The decision ultimately depends on what you value and how you think you can best contribute.
- One marcher adds only 0.15% to the crowd, which translates to a tiny increase in media coverage.
- The estimated impact of attending is equivalent to just a few hours of AI safety research.
- The analysis has limitations: small sample, rough estimates, and unpredictable media.
Why It Matters
If you care about AI risks, this analysis suggests your time might be better spent on direct research than on protesting.