I desire many things. I invent ideas to satisfy my desires. Whenever I learn something new, I compare the new fact against my database of desires. If my new fact could be used to satisfy a desire then I get an idea.

Some desires are universal among humans.

  • I want to be rich without putting in much effort.
  • I want lots of people to think I'm attractive.
  • I want a space station.
  • I want an electric car that costs $100.
  • I want a time machine, an invisibility cloak and a magic wand.
  • I want to live for a long time in good health.
  • I want to make my friends happy.
  • I want an ethical source of tasty protein.
  • I want to save and/or destroy the world.

Common desires are hard to satisfy creatively because too many people are competing to satisfy them. The possibility space of solution tends to have been picked clean. Either common desires are already satisfied or common desires are irreducibly hard to satisfy.

Cultivating Weird Desires

The trick to coming up ideas is to cultivate weird desires. Weird desires often arise from caring too much about things that don't matter (yet).

  • I want my keyboard inputs to be as efficient as humanly possible. (Solutions: Vim, Spacemacs.)
  • I want to learn foreign languages as efficiently as humanly possible. Anki isn't good enough. (Solution: write my own ML-powered flashcard software.)
  • I run LineageOS to evade adware, bloat and root protections on my Android phone.

Another way to cultivate weird desires is to be stingy with your money.

  • I write meta-ML software to automate architecture search because I need machine learning software but I do not want to hire lots of data scientists.
  • I lived in a hallway for two years to save money in college. I lived in a kitchen cupboard to save money in China. I like capsule hotels.
  • I use use AWS Lambda to avoid paying for more expensive EC2 instances. I use S3 to avoid the minimum price of a traditional RDS.

Weird desires motivate weird behaviors in a self-sustaining chain reaction.

  • I wrote a bash script that uses xrandr to adjust my screen resolution to the resolution of my drawing tablet because my drawing tablet's drivers for Linux did not map correctly to i3.
  • I do data science in Hy because my meta-ML software requires custom macros for its UI.
  • I am reading Tae Kim's Guide to Learning Japanese because I want to better understand the poetry of Zen Master Ryokan.


Justin Ith and I will discuss "How do you come up with ideas?" via Clubhouse @lsusr at 7pm Pacific Time on March 15th, 2021. (Note: Clubhouse runs only on iPhone.)

New to LessWrong?

New Comment
6 comments, sorted by Click to highlight new comments since: Today at 9:34 AM

Do you have a good system for saving and prioritizing the ideas you have?

Building a habit of noticing a new idea and writing it down (even if it's silly) has increased my overall output dramatically.

System? Yes. Good system? Eh.

My phone and laptop each contain three text files of ideas: one for blog posts, one for fiction and one for software/business ideas. New ideas go on the top. I don't organize them.

Another angle to view this is "coming up with ideas is compulsory if you want to optimize literally everything". Bonus: when you practice holding off proposing solutions, ideas are usually better. 

Some desires are universal among humans.

  • I want to be rich without putting in much effort.

This is particularly meaningful message to me. It is about a year ago when I found a youtube channel talking about "small work big money." It isn't just a discussion channel, but they try to carry their desire to action, by drawing characters, making hoodies, and communicating with users and running and expanding their business; it was just impressive that they stretch their ability in the theme of "small work big money." 

Since then, I believe human desires have the potential to influence people, even if it is just a dream.  


In case it interest you, I wrote an explanation about how to put an arbitrary scheduler in Anki http://www.milchior.fr/blog_en/index.php/post/2020/02/22/How-to-plug-a-scheduling-algorithm-into-Anki to work also with mobile software Please let me know if it present any interest to you. I'm extremely interested in alternative scheduler

Disclaimer: I was paid to develop the add-on behind this post, am a mainteneur of ankidroid and am quite related to anki developement in general

I have some ideas on how to make this argument better. First of all, I don't think some of the desires you listed are universal among humans. If something is universal that means every human shares that desire. I personally wouldn't want to be rich without much effort or own a space station and i'm sure there exist people in the world who don't have the space for a car or might have religious guidlines that they follow that discourage the use of electricity. A safe way to ensure that a trait or catagory is universal is if the trait / catagory you listed encompasses the sphere of all possible outcomes. For example let's say all humans are either (1) happy, (2) sad, or (3) neither happy or sad. This is a universal trait since no matter what you choose, there are humans that fall under one of these catagories. If you rewrote your list in this form you could fix these problems. For your weird behavior argument I think weird behavior could be defined in the following 3 catagories: (1) Weird behavior that is productive (2) weird behavior that is unproductive, (3) weird behavior that is neither productive or unproductive.