Power Word: Case Study

Heads up! I’m talking about the dark arts of the mind in this post. Continue at your own risk.

Power Word: Trust

A friend and I were recently discussing which magical spells we’d like to have. In an attempt to avoid the boring answer of “flight” and “invisibility”, we added a constraint: your magical spells aren’t allowed to violate any laws of physics. It’s a fun question, I’d recommend busting it out at your next festive event-party. People have some interesting answers, and my friend at the time did too, but I can’t remember what she said because I was too excited about my own idea.

Here’s what I wanted for a magic spell: Power Word: Trust. If you’re not familiar with Power Words, they’re a concept fro Dungeons and Dragons which are words so powerful that their mere utterance can change the fabric of reality. Personally, I wasn’t allowed to change the fabric of reality, but I was allowed to change people’s minds. Power Word: Trust would be a word that I could use to make someone trust my line of argument. My reasoning is often seemingly much more rigorous than my peers’, and too often I get bogged down trying to cross the inferential distance rather than the point I’d prefer to be making. And so I thought it’d be really nice to just have the ability to convince people that I know what I’m talking about. If they have their doubts still, I would provide citations to encourage others to level-up as well. And of course, I would only use my powers for good.

Of course, calling it “Power Word: Trust” is just rule of cool, but saying a single word would certainly be logistically easier than needing to draw circles in blood and chant incessantly.

Unfortunately, I couldn’t think of creating such a magic spell, so for the mean time it remains sitting in my mind, biding its time. One day.

A few days later, I read Brienne Strohl’s fantastic post about hacking social anxiety with Lob’s theorem. Go read it right now, if you haven’t. It’s absolutely fascinating and you will learn things and become a better human. If you’re lazy and aren’t going to read it despite my glowing recommendation, essentially Brienne realizes that if she believes that a long, drawn out process will fix her problems, she might as well just skip the long, drawn out process and reap the benefits without doing any of the work. And she derives it with math. Sick.

Let me stress how cool this is. It’s cooler than being cool. It’s colder than outer space during an ice age (that doesn’t make sense but you can shut your face hole, please). Brainhacking is neat and I’m especially interested in meta-level brain hacks that I can use to install other way-cool brain hacks.

That night, I woke up at 3am. This is not a usual occurrence for me, and, being in an inspired-yet-sleep-deprived mood, I decided that I’d give it a go. I was going to use Lob’s theorem to fix something in my brain.

But I didn’t really have anything on hand to fix. I’m sure there’s something there, but I haven’t identified it yet. Not being one to lose inspiration due to mere details, I decided to take it a little further. I was going to use Lob’s theorem to install a generalized version of Lob’s theorem in my mind, so that I could effortlessly install beliefs into my head when I figured out what I wanted to believe.

Power Word: Believe

Essentially what I wanted to do was install Power Word: Believe into my mind. A word that I could say to convince myself of something. There is no sense to need to go through all the work of actually believing something each and every time you want to do it, so, being a software engineer, I decided that I would wrap it in a function. Because I needed to work out the actual implementation of how this would all work, I would need to be able to write it down. And do math. I spent a few minutes searching the internet for cool arcane symbols I could use (again, rule of cool. If I’m going to be deriving math that will let me change the fabric of reality, I might as well consider myself a wizard), and came up with ℥¹, which of course is pronounced “Power Word: Believe”.

So I grabbed my notebook, and ran a bath in which I sat, meditating analytically, looking for a way of sneaking this idea into my mind. Occasionally I would write something that looked like modal logic in my notebook. I had looked up what Lob’s theorem looks like in modal logic, and I was trying to generalize it so I could install anything. This turns out to not be possible (or else you could prove everything), but I wanted it to be so. At least, in the context of my mind. I really, really wanted it to be so.

I tried relentlessly to hook this belief into my mind, but it was no use. I knew that wanting to believe is an error in thinking, and I’ve spent the last six years attempting to identify when I’m making errors in my thinking so that I will hopefully stop. And here I was, wanting to believe.

It was uncomfortable, and rightfully so. I actually gave out a cry of anguish at 4am in the morning as I made a little ground. It felt like splitting a fissure into my mind. It was the dirtiest I’ve ever felt in my life, trying to believe something that I knew wasn’t true. That’s an official No-No in the rationality handbook.

My last attempt in this session was to derive what I wanted out of Lob’s theorem directly. Recasting the necessity operator (□) as “known to be true” and the possibility operator (◇) as “believed to be true”, I derived my own version of doxastic logic, though I was not very rigorous in doing so. I’ve never been all that good at math. One of my axioms was that ◇ ◇ p = ◇ p, which is to say, “believing that you believe something” is the same as “believing something”. I believed that I had read this on LessWrong, since I remembered an article called “Belief in Belief”, and remembered that our thoughts on how we see the world are in fact our thoughts on our thoughts on how we see the world. There’s always an invisible level of indirection when talking about beliefs.

Anyway, this axiom turns out to be 100% false. Bad. Very bad. Don’t try to do mathematics with wrong axioms. Like a sports car, it might get you places, but unlike a sports car, it won’t make you happy. Finally I actually went back and read “Belief in Belief” and realized I was being stupid. I acknowledged being stumped.

I started looking for different ways of creating Power Word: Believe. As its name might suggest, this would be extremely powerful if it worked, AND there was an example of it working in the real world AND SO the only thing that was holding me back must have been my own mind. And I’ll be damned if I’m going to accept arbitrary barriers from my damn self. My first thought was psychedelic drugs. Maybe I could get into a weird state of mind where I would believe ℥ temporarily, and then use on itself to install it temporarily [ ℥_drugs(℥_permanent) → ℥_permanent ]. This is what I like to call, “thinking outside of the box”. Unfortunately(?) I did not have any psychedelic drugs on hand, and so I did not pursue this train of thought any further.

But instead of giving up, I took it around the community. I went to my first LessWrong meetup in the Bay Area, and asked the smart people there. Reactions were varied, and some people had some good ideas, but nothing worked. I ended up asking Brienne herself about it at a MIRI party she happened to attend, and while she seemed interested, there was too much festive event-partying going on to talk too much about it.

All seemed fruitless.

And then my good friend Jacob send me a stupid facebook message about how he had convinced himself that he could hear a classmate’s name in a song, and it made him laugh hysterically. Being intrigued for other reasons (and procrastination), I watched the music video and found that I too could hear the classmate’s name in lyrics I explicitly knew did not contain it. And then it struck me. I could convince myself of something I knew to be false. This is what I had been trying to do for weeks! This was it!

Weeks ago I had read the relevant LessWrong sequences explicitly around beliefs, but I had skipped one because I knew what it was talking about and it didn’t seem relevant, and boy-oh-boy was I wrong about that. The post in question was “Making Beliefs Pay Rent”, and it talks about how the only worthwhile beliefs are those which constrain your anticipations. Holding a belief should force you to expect to see the world differently than if you didn’t hold it, and if it fails to do so, it is a useless belief.

Herein was the key! I didn’t need to believe that ℥ be true, I just needed to let the belief that the (belief that ℥ be true) constrain my anticipation. Whoa. We’re getting meta here, and I haven’t personally sorted through the meta levels of what I’m saying, so there might be type-errors in there somewhere. Don’t get your shorts bunched around it, because that’s not the point. The point is this:

Installing ℥ is suddenly obvious. I just need to convince myself of the inside of the first level of brackets in ◇(∀p. ◇(℥p → p)) → ℥. If it holds for all cases, then it must be true in general! What I’m trying to express with my mathematics here, is that I needed to perform lots of thought experiments of the form “not p. invoke ℥. p.”, enough that I could convince myself that it holds for all p (via induction by example, which, as we all know, is the best kind of induction).

And so I started. I imagined myself being unable to speak to an attractive girl at a bus stop, and then ℥ escaped from my lips (as a glyph, no less!), and suddenly I could talk to this woman, and she was rapt with attention and we hit it off exceptionally well.

And then I stopped. There was a flaw in my plan. I could feel it there, but I couldn’t place where exactly. Something felt wrong. I was missing a piece of logic in my mind. And so, like I always do when I feel like I’m missing a piece of logic in my mind, I went back and read LessWrong. In particular, I reread the posts I had read recently, and then it dawned on me.

My beliefs need to pay rent. My beliefs need to regulate my anticipation. After I’ve gone through all of the work of installing ℥ into my mind, invoking ℥p isn’t going to make me anticipate p, unless p is small and trivial and self-explanatory. There is no shortcut. ℥ as we’ve explored it can’t exist, because it is a paradox: the only way to make it usable is to ensure that it won’t work.

This is to say that we can’t front-end load the work on this. We can’t construct the spell in advance. We can’t craft a general purpose spell that will let us believe useful things. We need to craft the spell each and every time. Luckily, we have a spell sketch (draw parallels to a proof sketch – we know how to construct it but we can’t generalize the procedure).

Brienne actually touches on this in her blog post when she mentions the miracle question, and while she talks a fair amount about it, mentioning that it started her along the path to her solution, I feel like she is missing the trees for the forest. The miracle question didn’t help her find the solution, the miracle question is the solution! The miracle question forces you to sit down and constrain your anticipations to your belief. It acts as the spell sketch.

The secret here is to go into your brain and say “(something is wrong. ℥_p would fix that. what would ℥_p look like? ℥_p fixes it.) → ℥_p”. We are calling a magical function into existence by observing that it would work if it did exist, and realizing that we’ve got the ability to make it exist.

Bam. Mafuckin’ wizardry, yo.

Power Word: Cause

I’ve realized what was wrong with my initial stipulation: I don’t want to believe things; I want to cause things. My (meta) beliefs about beliefs are useless, but my (object) beliefs about the world are actionable. ℥ is thus not to be pronounced “Power Word: Believe”, but instead “Power Word: Cause ”. Belief has nothing to do with it. This change in framing I think implies that I’ve successfully installed ℥. It no longer changes my mind. It now makes things happen, and that’s what really I had set out to do all along.

Brains are surprisingly finicky about framing.

Going Forwards

I haven’t yet invoked ℥, though I have absolutely no doubt it would work. Well, almost. I know that I can’t actually, directly CAUSE things. I can’t say ℥(I can breathe water) => I can breathe water and expect it to work. I need causal arrows from my belief to the world, and right now, ℥ is too weak a spell to jeopardize.

First I need to compartmentalize it. I need a context in which I can dissociate myself from it should it not work, if I accidentally overestimate my causal influence on a phenomenon. Without it, one wrong turn might cause my magic spell to self-destruct, and it seems too valuable to subject to such risks.

But what’s truly holding me back is that dark arts mind hacking is scary. It’s scary as hell. Like Moldbug says, it’s “a lot like do-it-yourself brain surgery. It requires patience, tolerance, a high pain threshold, and very steady hands”. I am not going to start casting ℥ willy-nilly, because I’m afraid of corrupting my brain. The benefits of it need to very probably outweigh the unknown risk of unforeseen interactions with other beliefs currently tenants of my mind.

I’ll find a use for this spell, but with great power comes great responsibility. But still, in the meantime, being a wizard is pretty frickin’ cool.