You Can’t Judge Memes

The ‘Existential Hope’ arm of the Foresight Institute is presently running a Meme Contest with a $10,000 prize. A panel of judges will evaluate entries based on things like their “virality potential.”

I have mixed feelings.

The memes they’re judging have already failed to go viral. If they hadn’t, you would have heard about them.

But their hearts are in the right place—they want to steer culture toward better futures. It’s just the mechanism that’s wrong.

The outcome speaks for itself: quietly deleted previously publicly shared entries.

To understand why the mechanism is wrong, we need to look at how memes behave.

Let’s do that by studying the biggest meme of the century.

The unpredictable history of the AI meme

It’s 2025. AI is unavoidable. Bear with me.

John McCarthy coined ‘Artificial Intelligence’ in 1956 for a workshop at Dartmouth—purely as memetic judo. He later admitted he chose the new label to avoid the older field of cybernetics and its dominant figure, Norbert Wiener. He said he did it so he wouldn’t have to “accept Wiener as guru or argue with him.“

Google’s Ngrams frequency for “Cybernetics” and “Artificial Intelligence”: this is what winning a meme war looks like.

Fast forward to the 2010s. Back then, the biggest problem the AI meme faced was making people not dismiss it as “The Terminator.”

Excerpt.

Eliezer Yudkowsky wrote about against this extensively: he wanted people to take AI seriously—as civilization-ending risk, not as science fiction.

The irony is that he ended up inspiring the current builders, instead of deterring them.

Sam Altman saying just that.

But here’s where it gets really good: Eliezer wrote the post screenshotted above in 2007. In it, he argued against “The Logical Fallacy of Generalization from Fictional Evidence”.

In 2023, then President Joe Biden pushed a sweeping executive order on AI safety motivated by… him having watched Mission: Impossible – Dead Reckoning Part 1, where the villain is a runaway AI called “the Entity.”

Could Eliezer have predicted this? Could the judges?

Every Single Meme is Like That

If this trajectory seems exceptional, just look at our past decade.

UK Health Secretary Matt Hancock said he pushed for early vaccine procurement as aggressively as he did because he’d watched the 2011 pandemic thriller Contagion.

Tomás Pueyo was an unknown blogger until his viral COVID essay strongly shaped public discourse and likely nudged governments toward early lockdowns.

A comedian whose campaign was literally launched through a TV show where he played the president became the actual president of Ukraine.

Bitcoin was launched in 2008 as a cypherpunk revolt. Now, the President of the United States wants to make it ‘the crypto capital of the world.’

Note the pattern: unlikely origin → memetic fitness → real consequences. And note what’s absent: predictability.

Which brings me to the meme contest.

Why Meme Contests Fail

The central mistake being made is thinking you can judge memes in isolation. But memes aren’t static artifacts—they’re organisms whose fitness depends on their environment.

I’ve recently learned Earth’s weather is unpredictable beyond a few days because weather events constantly collide with one another, creating cascading effects. Earth’s memetic landscape is like that too: memes don’t exist in isolation—they’re constantly colliding with one another, and it’s these collisions that determine their fitness. Which makes any individual meme’s future trajectory fundamentally unpredictable.

This unpredictability is fatal to the contest model. If judges can’t foresee their effects, then what exactly are they evaluating?

Memes cannot be judged outside their ecological niche, which is all other memes. Fitness is revealed only at deployment, never at design. The only judge of a meme is the environment itself.

Which is why I reposted this, despite it being bad form. Maybe deleting it was a mistake. We don’t know, and that’s the point! ‘Bad’ memes can famously take off! [1]

Conclusion

The goal of the contest was a good one. Shaping the memetic landscape toward positive futures is a good intention to hold. Look around you: everything you see began as an idea in someone’s head. Good memes matter.

But nurturing them requires a different approach than judging them in contests.

So this is my submission: a meme—in the form of an essay—about why you can’t judge memes…

…which I’m submitting for you to judge.
Good luck.

---

[1] — Just look at the most recent AI meme drama: the widely-circulated ‘AI 2027’ report claimed AGI-level systems could arrive by 2027. When the predictions didn’t pan out, critics said it should have been titled more responsibly—like the European report ‘Advanced AI: Possible Futures.’ Had you heard of that report before? Exactly. The ‘bad’ title created circulation and discourse; while the ‘good’ title created… nothing. Memetic fitness doesn’t care about your judgment.