Also Dan Luu's essay 95%-ile isn't that good, where he claims that even 95th-percentile Overwatch players routinely make silly mistakes, suggesting that you can get to that level by not making mistakes.

Stop saying wrong things

by lc 5 min read1st May 202034 comments

71


Lots has been said within these servers on finding ways to say new exciting correct things. Or correcting things many think are right, but didn't know were actually wrong. This is a relentlessly optimistic perspective. I have gotten grand mileage by merely trying to hold back from saying or acting on information or ideas I know are wrong, as opposed to worrying about thinking better, or not making mistakes, or being more rational when I am concerned with the truth. This other part is the meaty part, the intense part - the struggle to stop mouthing the words, to stop myself from willingly burning more money or time.

"Why on earth would a self-identified rationalist, with full knowledge and consciousness, submit to incorrect information?", you might be asking. Boy, wouldn't I like to figure that one out. You see what I just did there? I refrained from saying something wrong. I could've made some plausible sounding thing up. I could've even made up something that was less plausible, but which my listeners might be too polite or confused to dispute me on. Normally, in polite conversation, I'm inclined to. But not this time. Not here, at least, of all sacred places - perhaps as a consolation prize to my ego.

Most people just don't even try to believe correct things, and make major life decisions based on transparently bad logic. This straightforward fact is revealed just by listening to people speak to one another, just by hearing them describe themselves and their motivations, and thinking internally about your own motivations. No one needs to be a rationalist to understand these pathologies, and no one needs to be a rationalist to recognize them. Think of all the people who start and don't finish learning a new language, yet with clear intention at the beginning to make full use of time well spent. Do most of these people actually believe they will finish learning Chinese? Do they just refuse to look up online statistics on the hours it takes for full time diplomats to learn Chinese, and blindly trudge 25 minutes a day on Anki, because they can't imagine that their new three year plan to master Mandarin might be something that deserves thirty minute validation? Or perhaps they fully understand how they might make better decisions, and just... don't. Because... I don't know. Insert fleshed out paragraph here. Insert Robin Hanson linkpost here.

Take a look at this presentation by one of the early employees at Etsy, a massively successful startup. Etsy made 600+ million in revenue last year, meaning the company has apparently outperformed at least 99.99% of all startups ever. Lot of child prodigies in that 99.99%. Lot of Eliezer Yudkowskies in that 99.99%. And for half a decade, they also apparently refrained outright from using math or data to make decisions about how to grow an empire. These very intelligent people - who are at least smart enough to build a multibillion dollar company- they took five whole years to come up with the idea to A/B test their 6 month plus projects, and even longer to build prototypes. The author's described method of A/B testing, which you can glean from his other posts, is toddler fodder - not that it doesn't work, but lots of people here could probably improve upon it immediately if they were trying. The pervasive problem he's describing, however, is that they decided not to try - they didn't need a better statistical methodology, the company wasn't using any methodology many years after it was already getting press in the New York Times. Etsy as an organization, did not care or want to be correct at predicting what its users would want, beyond going through the motions of writing plans on whiteboards in front of their friends.

Most people, like Etsy, are beyond not using math. Almost everyone is beyond not using math the vast majority of the time, borrowing the mathematician's notion of "almost everyone". The thing Etsies need to do is try at all, not be pointed to the sequences. Like most people, I know I should work out. For a long time, I didn't, at least not seriously. I would have this running gag, though I considered it very Productive^TM loop and not at all a gag at the time, that since working out was so very tough, every time I worked out I had to make sure conditions were perfect. You know, for microhedonics, and spoon conservation. I needed to "trick" my brain into being comfortable working out, and along with that came a host of "precommitments" and chanting internally "you MUST work out two or three times a week you must you CAN'T not do it". Then one day, when I was about to decide not to go because it's cloudy or it rained last night or some such nonsense, I thought to myself:

You know, there really is nothing in my brain that changes the dynamics of the choice I'm about to make. No matter what I say to myself about the impeding circumstances, if I don't somehow exercise today, I'm going to be very slightly physically weaker by tomorrow and have a very slightly shorter life expectancy.

I pretty much work out as often and as intensely as I figure is aligned with my goals now. I just do it, because I know working out is a good idea and I know it's usually still a good idea regardless of the weather that day. It's amazing the quality of life increases I have made by merely not trying my damnedest to believe incorrect things. The domain name of this website would lead me to believe most of you are attempting the same, but I don't see a lot of the discussion about this that I would expect. I don't see people discussing the same problems that I fight. The books and skills I pour hours and hours into, then forget; the university I study at because a friend I no longer hang out with decided to attend; the job I take without applying anywhere else, researching salary or internship expectations, or learning any negotation or marketing strategies. How distant and conceptual the prediction market is in the face of my absurd self-sabotage. How foreign timeless decision theory. Am I so unique? I refuse to believe it.

Another example. I have some very specific life goals. Specific enough to the point where they're specifically motivating when I think about them. Specific enough that I've written them down and drawn up detailed explanations for myself of what they might look like. For as long as I can remember, up until the last year or so, I haven't actually done anything to achieve them. I want to achieve them very badly, and it hurts to think how far away I am from them. They're not impossible goals, or goals that only 0.1% of the population can usually achieve, or goals that require Herculean self-sacrifice on my part. But I refused to actually engage in the loop of trying seriously to come up with effective ways of completing my objectives. Instead, I would do something that I pattern matched with being productive, and then retroactively explain to myself why those actions were pushing me forward. Reading Real Analysis textbooks and doing every proof, despite not liking real analysis that much for its own sake. Rearchitecting my commercial app for the third time, despite hating the tedium of grand refactorings and cursing myself every step of the way. These decisions weren't intelligent, to me or anyone I explained them to. I just didn't care. Whatever internal psychology of my brain led me to learn Real Analysis and then forget about it within six months, it wasn't impeded by any sort of underlying motivation for truth or self-interest. For most of my life, and even now much of the time, I simply did and do not care about being less wrong, much less more right.

Clearly, I have recently started to gather a few methods for maintaining against this kind of open self-sabotage. Obviously, they're insufficient, or else I'd be busy conquering the world by now, what with my ability to make effective decisions and believe things that are true. My main strategy is to, in the theme of this post, stop saying wrong things. Or at least try to stop. You can never fully suppress the urge to go headfirst into a conversation, present ill-formed and poorly validated ideas, and bask in your own intelligence. Trying, however, does more than you might think. First, you just end up being an asshole pretending to be honest, because that's what people end up doing when they try to be honest for the first time. But then you just end up building a self-perceived-identity around honesty and unmitigated goal seeking. You actually start to think about the things you say and do. After all, your friends (they don't care in reality but this is your internal monologue speaking) might judge your efforts negatively if what you're doing doesn't actually accomplish anything. That's motivating. But it's not good enough - you're friends are insane too. Then you take the meme too far. You start engaging in reasonable dialogue with your own internal thoughts, fully independent of others. Why am I buying the large case of Oreos at the grocer if I don't plan on eating them all before I come back next? Why don't I just launch or attach to tmux automatically in my bashrc instead of having fifty different unlabeled Urxvt windows open? Why have I been trained to wipe my ass of poop with a series of small, dry sheets of paper towels? These are some dangerous ideas that are popping into your head all of the sudden.

That's when you start actually saying correct things. Your newfound intolerance for ideas that don't make sense produces more than enough insight for the few conversations you have regularly. Suddenly, you realize, you don't have to say incorrect things to keep conversations going. You can just talk about more boring stuff, or better yet, become your social group's local pragmatic pedant. And maybe, just maybe, you start appropriately acknowledging reality. That last part hasn't happened yet. But I can feel reality creeping up on me. It's always been there - I'm just starting to work with it instead of around it.

71