Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Cached Thoughts, published by Eliezer Yudkowsky on the AI Alignment Forum. One of the single greatest puzzles about the human brain is how the damn thing works at all when most neurons fire 10–20 times per second, or 200Hz tops. In neurology, the “hundred-step rule” is that any postulated operation has to complete in at most 100 sequential steps—you can be as parallel as you like, but you can’t postulate more than 100 (preferably fewer) neural spikes one after the other. Can you imagine having to program using 100Hz CPUs, no matter how many of them you had? You’d also need a hundred billion processors just to get anything done in realtime. If you did need to write realtime programs for a hundred billion 100Hz processors, one trick you’d use as heavily as possible is caching. That’s when you store the results of previous operations and look them up next time, instead of recomputing them from scratch. And it’s a very neural idiom—recognition, association, completing the pattern. It’s a good guess that the actual majority of human cognition consists of cache lookups. This thought does tend to go through my mind at certain times. There was a wonderfully illustrative story which I thought I had bookmarked, but couldn’t re-find: it was the story of a man whose know-it-all neighbor had once claimed in passing that the best way to remove a chimney from your house was to knock out the fireplace, wait for the bricks to drop down one level, knock out those bricks, and repeat until the chimney was gone. Years later, when the man wanted to remove his own chimney, this cached thought was lurking, waiting to pounce . . . As the man noted afterward—you can guess it didn’t go well—his neighbor was not particularly knowledgeable in these matters, not a trusted source. If he’d questioned the idea, he probably would have realized it was a poor one. Some cache hits we’d be better off recomputing. But the brain completes the pattern automatically—and if you don’t consciously realize the pattern needs correction, you’ll be left with a completed pattern. I suspect that if the thought had occurred to the man himself—if he’d personally had this bright idea for how to remove a chimney—he would have examined the idea more critically. But if someone else has already thought an idea through, you can save on computing power by caching their conclusion—right? In modern civilization particularly, no one can think fast enough to think their own thoughts. If I’d been abandoned in the woods as an infant, raised by wolves or silent robots, I would scarcely be recognizable as human. No one can think fast enough to recapitulate the wisdom of a hunter-gatherer tribe in one lifetime, starting from scratch. As for the wisdom of a literate civilization, forget it. But the flip side of this is that I continually see people who aspire to critical thinking, repeating back cached thoughts which were not invented by critical thinkers. A good example is the skeptic who concedes, “Well, you can’t prove or disprove a religion by factual evidence.” As I have pointed out elsewhere,1 this is simply false as probability theory. And it is also simply false relative to the real psychology of religion—a few centuries ago, saying this would have gotten you burned at the stake. A mother whose daughter has cancer prays, “God, please heal my daughter,” not, “Dear God, I know that religions are not allowed to have any falsifiable consequences, which means that you can’t possibly heal my daughter, so . . . well, basically, I’m praying to make myself feel better, instead of doing something that could actually help my daughter.” But people read “You can’t prove or disprove a religion by factual evidence,” and then, the next time they see a piece of evidence disproving a religion, their brain completes the pattern. Even some atheist...