1. Keep it short.
Omit needless words. Take a document you wrote and review it sentence by sentence. Compress each sentence as much as you can, omitting every possible word without breaking the sentence. Literally: Go over each sentence, word by word, and leave out everything you can. "It is a fact that Fred is sick" becomes just "Fred's sick". "See if there are any bits and pieces you can leave out" becomes "Leave out bits and pieces where you can." Make this a routine part of editing.
Your new style might seem strange, but it will suck less. Once it becomes habit, you'll develop a style that suits short sentences. At any rate it is better to have strange-sounding terse writing than to have writing that sucks. Um...rather: "Better strangely terse than sucky."
As an exercise, write a philosophical dialogue or short story with each character's speech limited to a total of two lines. You'll find yourself trying to squeeze that bon mot or witticism into just two lines. You'll notice tricks, such as leaving out 'the fact that', or transforming the passive voice to active voice. "She was bonked by the crowbar" becomes "The crowbar bonked her", saving precious words. Squeezing dialogue into the shortest possible space gives you good writing habits.
2. Stay out of your writing.
Your document talks with the reader; you do not. So get out of the reader's way. Nobody cares about you or what you think. The word 'I' usually marks a flawed sentence. The audience doesn't care who you are; they probably don't care how you arrived at your conclusions. They only want the arguments that can be abstracted and reused. Omit all mentions of yourself.
No, I'm not a hypocrite. I believe in this trick but haven't quite learned it myself yet - Oops..
3. Avoid formality.
Don't try to be formal. Botched formality looks far more amateurish than botching ordinary speech. Botching deliberate informality is also amateurish; just write in a normal voice.
Yes, it might be helpful if your editorial voice sounded like a trained academic, but you won't sound that way; you will sound like someone trying to sound like a trained academic, or someone whose previous experience with writing has consisted of regurgitating textbooks for school work.
As practice, try writing science fiction or some other form of drama, rather than philosophical essays. Practice some voice other than the formal. Write a Singularity essay as a philosophical dialogue between teenage girls in a mall, and compress each character's dialogue into no more than two lines.
4. Never compromise rationality or ethics under any circumstances.
See [this SL4 post.]
If you compromise rationality or ethics for seemingly good reasons, you will botch it. Guaranteed.
Say nothing you do not believe true. Advocate nothing you do not believe to be right. If that's inconvenient, then keep practicing completely honest and ethical writing until you can write under those constraints. Do not practice anything else, or you might learn it.
5. Be kind.
"Unkindness" includes any kind of personal attack on anyone or anything, no matter how well-deserved it seems. Ask yourself whether the nicest person you know or can imagine would write it that way. Write like a Friendly AI.
Rule 5 is fully compatible with Rule 4. Strange, but true. I once thought politeness was incompatible with total rationality and honesty. Then I wrote some dialogue for a fictional Friendly AI. This showed my mistake. I will figure out what I did right, and incorporate it into my writing, no matter how long it takes. I learned politeness is allowed.
6. Maintain sympathy with the audience.
The audience is not a them. It is an us. They're your friends you haven't met yet, and your fellow humans. Don't try to manipulate the audience. When you find yourself thinking manipulatively, stop writing and delete whatever you just wrote. Continue only after you regain your link with the audience.
As you write, you'll form subgoals - things you want to accomplish by writing. Learn to detect manipulatively phrased subgoals. When you find one, kill it. Don't rationalize it as a sympathetic subgoal, just kill it.
Unsympathetic thinking, placing yourself at a distance from the audience, begins whenever you think of the audience as "the masses" rather than "potential Singularitarians I haven't met yet". Manipulative thinking happens whenever you ask "How can I persuade people to believe X?" or "How can I persuade people to do X?"
Do you believe X? Do you do X? Write down all evidence you personally considered to arrive at that conclusion. Those are your arguments. If these seem weak, don't rationalize extra arguments you didn't consider personally, even if they "seem like good arguments" to you. The most reliable method for determining whether an argument is valid and whether it will seem valid to the reader, is not the intuition that says something you made up "sounds like a valid argument". Look instead at whether whether an argument has strength enough to affect your own decision-making.
If your arguments seem weak, does your conclusion seem weak? If so, perhaps you shouldn't argue for weak conclusions. If the conclusion doesn't feel weak, but the arguments do, then you must have failed to list all the evidence you considered. Perhaps you're simply wrong. So mentally question X. Don't ask yourself "How can I persuade people to believe X?" or even "Why do I believe X?" Reconsider X. Ask yourself: "Is X right?". Again, write down whatever evidence you personally considered, followed by whatever conclusion you reached. There. You've outlined your article. Now figure out how to explain the explanation. (It's recursive.)
But when you include negative evidence you personally considered, doesn't that run the risk of persuading fewer people? Yes. That's not a "risk", it's how things should work. When you are on the right side, you don't need to cheat in order to win. Seriously. If honest writing ever seems unpersuasive, it means you haven't yet learned how to list all the evidence that convinced you.
Sometimes early-stage Singularity fans try to "make the Singularity appealing". Besides being dreadfully unethical, it always sounds like wish-fulfillment. Some futures depicted I would not in a million years wish to inhabit. The audience agreed. The moment you stop thinking about what makes sense to you, and start guessing what will "sound good" to an audience, you lose your connection to the audience. You no longer know whether your words make sense. Generally, they don't.
Ethical writing is not "persuading the audience". Ethical writing is not "persuading the audience of things I myself believe to be true". Ethical writing is not even "persuading the audience of things I believe to be true through arguments I believe to be true". Ethical writing is persuading the audience of things you believe to be true, through arguments that you yourself take into account as evidence. It's not good enough for the audience unless it's good enough for you.
If your writing tries to convey a thought you wouldn't think yourself, you're out of sync.
Whatever sounds awful to you probably sounds worse to the audience. If you think they can't tell the difference, you're fooling only yourself.
Don't use stupid meme tricks. You don't believe in them yourself; well, neither does the audience. They won't fall for it. You are a beginning writer; you do not have the powers of P. T. Barnum, nor should you strive to develop them. Feed your audience's strengths, don't exploit their weaknesses. Always strive to make the audience more rational at the end of your essay than at the beginning.
7. Track your dependencies.
The audience is an us, not a them - but that doesn't mean the audience is already subscribed to SL4.
You're a hunter-gatherer, blood and bone and DNA. You were born to live in small tribes in which everyone grows up with the same aliterate education.
The inability of most scientists to communicate with a lay audience, or even scientists from different disciplines, probably stems from this violated ancestral invariant. In a hunter-gatherer tribe there are no abstract disciplines with vast bodies of carefully gathered evidence generalized into elegant theories whose conclusions are fifty inferential steps removed from externally shared premises. We have no built-in faculties for communicating across vast inferential distances, and must learn to do it consciously if we learn it at all.
We are all hunter-gatherers. Our instincts say that anyone who says something with no immediately obvious evidence is an idiot, because in a hunter-gatherer tribe, he is. Conversely, if you say something blatantly obvious and the other person doesn't see it, he's an idiot. Anyone who expects you to believe a statement with no obvious evidence is either a poor liar, or (at best) the tribal witch doctor. And you'd better not try to claim the status of witch doctor (by making witch-doctorish statements) unless you have a bona-fide bone through your beard; do you take us for fools?
The scientist has no instincts that track whether the words and conclusions he's using are fifty inferential steps removed from the evidence the audience is previously familiar with. He knows that the layman doesn't know the conclusions; he doesn't realize that the layman doesn't know the premises. Reporting a result new to his scientific tribe, he takes one or two inferential steps backward to the common wisdom of his tribe; obviously this is a persuasive argument. The layman is not persuaded, and so the scientist thinks the layman is an idiot. And in turn the layman has no instinct that someone offering blatantly unsupported conclusions might be working fifty inferential steps away; he thinks the scientist is a lunatic. How often does a hunter-gatherer in a tribe with no written literature wind up more than two or three inferential steps away from everyone else?
Luckily for the aspiring Singularity writer, many key parts of Singularity theory are not (yet) fifty inferential steps away from the audience's previously accepted premises... just four or five. You can follow the inferential chain all the way through, in one article, provided you keep track of what the audience knows at any given point. An "argument" is an inferential pathway starting from what the audience already knows or accepts. If you don't recurse far enough, you're just talking to yourself.
If at any point you make a statement without obvious justification in what you've previously supported, the audience will think you're an idiot or a cult victim. This also holds true if you visibly attach greater weight to an argument than is justified in the eyes of the audience at that time. You, from your knowledge base, may know something at the level of virtual certainty. But if the audience doesn't have enough information to be virtually certain, you have no right to ask them to be virtually certain, and in fact they probably won't be virtually certain regardless of how much support you put into your essay. So don't talk at them as if you expect them to be virtually certain. At the very least, make sure they know you know the difference; better yet is writing in such a way that it never becomes an issue.
Do not expect your audience to believe what you have not supported. Do not expect your audience to believe strongly what you have supported weakly.
An ethical argument is an inferential pathway, leading to a conclusion you yourself hold as true or moral, composed of arguments which you yourself take into account as support, starting from facts or principles already accepted by the audience to which you address yourself, and at no time making use of a degree of belief you have not previously supported. (This doesn't mean you write your essay as a formal chain of logic; Rule 3 still holds, and you're supposed to use the inferential pathway, rather than mentioning it. But keep track of what you have and haven't supported.)
Our inability to communicate with inferentially distant minds screws us up in subtle ways. It's not just the beliefs, it's the vocabulary of support. A biologist can explain evolution to a physicist by saying that it's "the simplest explanation". But not everyone on Earth has been inculcated with that legendary history of science, from Newton to Einstein, which invests the words simplest explanation with their awesome import. The biologist and the physicist are both scientists; to them "simplest explanation" is a word of power, spoken at the birth of theories and carved on their tombstones. To anyone else, "But it's the simplest explanation!" sounds like an interesting but hardly knockdown argument; it isn't all that powerful a tool for comprehending office politics or fixing a broken car. Obviously the biologist is infatuated with his own ideas and doesn't consider other explanations.
If you've ever wondered how sane people can end up doubting evolution, this is how.
Check at each step that makes sense to you for lack of audience-visible support. Check the beliefs your arguments rely on, check the evidence the beliefs were generalized from, check the weight of convincingness you assign to each step in the argument, check your conceptual vocabulary, check the import you assign to words and concepts. Make sure that at the very least the audience knows why you believe something, even if they might not agree with you yet. Otherwise you are a lunatic.
Accept that you can't tell the audience everything you know under such harsh restrictions, and settle for telling them at least one new thing.
8. Stop being defensive.
Some people will read all the way through your document and still disagree with you. It's a fact. Live with it. Don't try to counter every objection you think your audience might think of. You're just shooting off your own foot, listing (pointless) objections no one would have thought of otherwise. Also, your writing will sound unbearably amateurish.
If an objection had enough weight with you personally that you stopped and thought it over on first encountering the idea, it may be worth going into.
FAQs and academic documents are permitted to consider objections, provided that they are either serious (you considered them yourself) or common (you've actually seen other people raising them). If you still sound defensive in your ordinary writing, don't try to write FAQs or academic documents until you've gotten over it. The last time you want to sound defensive is when you're defending yourself.
9. Don't invent new words.
Yes, I frequently violate this myself, but at least I've been trying to keep it down.
If you do violate the rule, then make the new word as self-explanatory as possible. "Seed AI", "Friendly AI", and "neurohacking" are good. "External reference semantics" is bad.
10. Don't write for an SL4 audience.
It's useless. It may feel satisfying but it accomplishes nothing. What's worse, it teaches you nothing.
11. Learn how to write.
Writing about the Singularity can stump even experienced wordsmiths, so learn how to write. "The Elements of Style" by Strunk and White is widely considered mandatory. Some other good books I've read include "How to Write Science Fiction and Fantasy" by Orson Scott Card, and "Writing For Story" by Jon Franklin.
Any book about writing will tell you, among other things:
Consider writing non-Singularity articles for publication in a newspaper or magazine, or SF short stories. Singularity writing is harder than writing; if you can write readable material on the Singularity, your non-Singularity writing skills should be good enough for newspapers (and then some). Newspapers and magazines sometimes pay very well by impoverished college student standards.
(SF stories sell for a lot less per word and have much higher quality standards.)
See also LearnHowToWrite.
12. Don't harangue your audience.
I'm still trying to master this myself, and failing pathetically, but it's good advice. The audience does not like to be lectured, and we have already established that the audience is God. If you get all the way to the end of an essay and you think the audience still disagrees with something, then for Belldandy's sake don't lecture them about it. You offered up whatever evidence you had, and either it was sufficient or it wasn't. Don't harangue the audience on whatever points you weren't smart enough to establish; it just ticks them off.
13. Drop any habits you learned in school.
Unlearn everything you learned in high school or college. In high school or college you learned to regurgitate textbooks in a slightly different form which would convince your teacher you learned something or had at least read the assignment. Singularity writers fresh out of school produce essays apparently designed to convince someone they read the assigned material at intelligence.org.
"School writing" is wrong in almost every way. Instead of building an incremental pathway from what the audience already knows, school writing self-consciously mentions important-sounding words and beliefs in an effort to convince someone that the material has been digested. School writing is stilted and formal because it regurgitates stilted and formal textbooks, and because teachers grade it-looks-vaguely-like-academia-to-me, it-must-be-knowledge, not ease of comprehension. School writing about the Singularity winds an incoherent and disconnected pathway, if "pathway" isn't too strong a word, to a final conclusion that is always found somewhere at intelligence.org and is always phrased in the language of singinst.org, whether or not it makes any sense to the audience. And it will send anyone who's been out of college for a few years, anyone used to reading grownup books, screaming into the night.
14. If the audience doesn't remember it, it didn't happen.
Corollaries include:
15. Make every statement in the simplest possible language.
No bells and whistles. The powerful words may impress you, may have been impressed into you, but the impressiveness is not shared; it is inferentially distant from the audience.
A good writing exercise may be to write without using a single SL4 buzzword. Conjure up vivid imagery, don't use the exciting names of vivid imagery that you've encountered in the past and that your audience probably hasn't.
16. Don't fight the audience.
You are not trying to win an argument with the audience in front of a meta-audience. The audience is all there is. Fight them and they'll fight back. And you'll lose, because you're suing your own jury.
Success is teaching the audience at least one new thing. You do not have the resources to simultaneously persuade them on every point; persuading them on most points is excellent. If the audience's free acceptance of whatever you can honestly establish isn't good enough for you, you will end up fighting the audience over the remaining points. And losing.
17. Show. Don't tell.
Show the evidence, don't just tell the conclusions. Show the facts and generalize, don't just tell the principles. If at all possible, prefer the concrete to the abstract and the illustration to the assertion. (But don't be concrete in a way that you yourself don't believe in. Don't show "concrete examples" of life after the Singularity unless it's what you honestly think will happen.)
18. Make one point.
A successful short or medium-term paper will make one point and make it thoroughly. Making two or more points requires a book. Short papers that contain many points generally look like attempts at Singularitarian's Creeds rather than arguments for anything; an impression confirmed by the presence of too many buzzwords and opinions which all appear to be carbon copies of SIAI's online material.
It should be obvious, but choose your one point before you start writing.
19. Proselytizing is even harder.
You found out about the Singularity five days ago, and you want to do something productive, so you write an article summarizing all the reasons why people should donate money to the Singularity Institute.
In a word: No.
Proselytizing is even harder to do well, easier to screw up. If you try to write proselytizing material, on your first attempt at Singularity writing, which is your first attempt at writing, you are guaranteed to sound like a cult victim. Counterproductive.
Gather the points of fact you would need to establish. Write about them first, one point of fact and one article at a time.
Better yet, if you aren't already a writer, start your writing career by trying your hand at a short story or two, or an article explaining some simple piece of science or technology that interests you and about which you happen to have unusual technical knowledge. Fail where it doesn't matter, get your feedback where it's simple.
An exception is a case where you're talking, informally, about something you understand completely: Your own reaction. I.e., a post in a blog: "This is just the coolest thing ever. No, seriously. Coolest... thing... ever." If that is your honest opinion, then it is a straightforward historical fact that that was your opinion. And it doesn't sound phoney.
20. Do not try to explain anything you don't understand yourself!
Last and most important: Don't explain things you don't know!
School forced you to explain things you hadn't learned, to prove that you at least tried. And you had to explain everything that was required in the essay, regardless of whether you understood it; you were not allowed to concentrate on your strengths.
The things you write should, always, be your own opinions, never anyone else's. If you haven't studied something long enough to develop strong, definite opinions... don't write about it! Alas, school trains people, not just to explain things they don't understand, but to not know what they don't understand. Even without the damage of schooling, sensing shaky understanding is a cognitive skill that takes time to develop. Be conservative.
Some key questions to ask are:
To learn how to write, find a topic that you understand, thoroughly, from many directions, that you can defend successfully and confidently and without losing your place, that seems obvious, with the standard objections rousing to indignation from their sheer foolishness, about which you know relevant technical material that most audiences might not. The topic will seem utterly obvious and straightforward to you personally, whether you think it would seem simple or complex to others. The next step is to make it seem that obvious and straightforward to the audience.
Corollary: Friendly AI is not just "harder", it is off in a private dimension of difficulty five lightyears distant from any other Singularity topic. I seriously doubt you can say anything definite about FAI unless it is the uncontroversial historical fact that Eliezer Yudkowsky made such-and-such claim about FAI in such-and-such year. Nonetheless if someone says something straightforwardly, blatantly wrong about AI morality, go ahead and shoot that down, one point and one article at a time. See HowToLearnFriendlyAI.
21. Always be accurate and be as precise as possible.
Write accurately. Inaccuracy is lying, even if you are unaware of your errors. Therefore it is best to write about topics you understand. If you only know some facts about your subject, there is a high probability that you will introduce errors through sloppy phrasing. This is against everything you were taught in school, but see rule 13.
Write as precisely as the situation allows. For example, if you have been asked to describe Friendly AI in 50 words, you have no hope to include the full text of CFAI. In such cases, you must remain accurate and try to be as precise as possible. If you were writing a Google Ad and given just 50 characters, it might be better to say "Seed AI builds itself" than "Seed AI recursively self-improves". The latter is more precise, but both are accurate. The one you choose depends on the context of the situation.
If you do not have length limits, be prepared for a helluva lot of work. Exact precision is difficult and requires a lot of writing. CFAI is a long document, but it doesn't even really describe what Friendliness is. Yes, exactness is that hard.
Re rule 4.
-- GordonWorley
See also TipsForWritingProductivity.
I disagree with a minor point, so i won't edit the text or make a new one. "It's the simplest explanation" works as a powerful tool in office politics and fixing cars, and lots and lots of other simpler 'unscientific' endeavors. It is a nonobvious tool, and it's not formally applied, but almost every discipline with any depth has a variant rule. KISS, Occam's Razor, "All errors are user errors", Innocent until proven guilty, the principle of mathematical elegance, soundbites, Thoreau psychology, etc etc etc.
But it's not useful unless the audience can generalize across jargon, or it's formally recognized within that audience's competencies as a cousin or variant rule. And of course, unless you actually explain yourself rather than invoking words of power. --JustinCorwin
See also [The Age of the Essay] for further insights on how real essays differ from school writing.
I am going to beat up anybody who deletes Rule #10. Or, for that matter, anybody who forgets Rule #10. Come to think of it, I'll have a lot of beating up to do...
I sort of disagree with Rule #2. It does apply sometimes, but more with academic audiences than with laypeople. There are a lot of people who find any writing more interesting if it sounds personal. Know your target audience, and write accordingly. -- KajSotala