"A year spent in artificial intelligence is enough to make one believe in God"
About this Quote
A year in AI, Perlis jokes, is enough to drive a rationalist to theology. The line lands because it’s aimed at a very particular kind of hubris: the engineer’s recurring belief that intelligence is just a hard but finite programming problem, and that a few more breakthroughs will make minds legible, reproducible, and obedient. Perlis flips that optimism into mock-conversion. If you spend long enough building “intelligence” out of brittle rules, leaky abstractions, and data that refuses to behave, you start to feel the gravitational pull of mystery again.
The subtext is less piety than an acid appraisal of AI’s tendency to overpromise. Perlis came up in an era when early AI often sold itself with the confidence of a new science and delivered, in practice, systems that worked impressively until they didn’t. His “God” is a stand-in for the irreducible: the combinatorial explosion, the edge cases, the shifting definitions of what counts as understanding. It’s also a wink at how quickly researchers end up speaking in metaphors - “knowledge,” “reasoning,” “learning” - because the formal language keeps slipping out of reach.
Context matters: Perlis was a computer science pioneer with a talent for aphorisms, and the joke reads like insider gallows humor from someone who has watched bright people mistake cleverness for comprehension. The line isn’t anti-AI; it’s anti-triumphalism. It’s a reminder that the closer you get to building something mind-like, the more you confront what your tools can’t capture, and how much faith - in models, in metrics, in the next paper - sneaks into a field that prides itself on proof.
The subtext is less piety than an acid appraisal of AI’s tendency to overpromise. Perlis came up in an era when early AI often sold itself with the confidence of a new science and delivered, in practice, systems that worked impressively until they didn’t. His “God” is a stand-in for the irreducible: the combinatorial explosion, the edge cases, the shifting definitions of what counts as understanding. It’s also a wink at how quickly researchers end up speaking in metaphors - “knowledge,” “reasoning,” “learning” - because the formal language keeps slipping out of reach.
Context matters: Perlis was a computer science pioneer with a talent for aphorisms, and the joke reads like insider gallows humor from someone who has watched bright people mistake cleverness for comprehension. The line isn’t anti-AI; it’s anti-triumphalism. It’s a reminder that the closer you get to building something mind-like, the more you confront what your tools can’t capture, and how much faith - in models, in metrics, in the next paper - sneaks into a field that prides itself on proof.
Quote Details
| Topic | Artificial Intelligence |
|---|
More Quotes by Alan
Add to List
