Artificially Intelligent

Any mimicry distinguishable from the original is insufficiently advanced.

Miscellaneous Mediocre Models

| 1404 words

The world is very complicated. It is easier to engage with the world productively if you have simple models of how various parts of it work. These are some of the models I use. Some of them are more grounded in reality than others.

Acids/Bases

epistemic status: probably literally true

There are things that are acidic or basic. Acidic things taste sour. Basic things taste bitter. If you combine an acid, like vinegar, and a base, like baking soda, then they neutralize (and also maybe generate carbon dioxide gas). Having acid on your teeth is bad because it’ll corrode them. One of the functions of toothpaste is to be a base that neutralizes the acid on your teeth.

Water often contains calcium carbonate, which is a base. Calcium carbonate is white and chalky. If an area that is next to water is covered with white and chalky substance, e.g. your shower door, it is likely calcium carbonate. This can be cleaned with an acid, like vinegar or lemon juice.

If something that touches water is clogged, then it is likely because calcium carbonate. Soaking the relevant area in vinegar might help. I have unclogged a sink and a dishwasher by soaking the areas in copious amounts of vinegar.

Polar/Non-polar

epistemic status: probably literally true

There are things that are polar or non-polar. Polar things don’t like touching non-polar things and vis-versa. This is why oil and water do not mix. Certain substances, like soap, have both polar and non-polar bits; this allows a polar substance, like water, to wash away a non-polar substance, like grease. This suggests that if something has only be dirtied by polar things, like sugar, then using soap to wash it is not necessary.

Cleaning agents often function by being differently polar than water. As mentioned above, soap functions to bridge polar/non-polar and wash away oils. Rubbing alcohol is less polar than water, and can also help clean substances that repel water. Acetone is less polar than rubbing alcohol, and can thus clean even more things.

Working Memory

epistemic status: metaphor

Your brain has some number of memory slots that can be swapped quicker than other memory slots. You use these memory slots both for remembering things temporarily and for performing computation. If you are trying to remember many things, you can perform less computation. If you are in a distracting environment, some of these slots get occupied with ignoring distractions. If you are worried about something, some of these slots get occupied with worrying.

You also have other memory slots in other places in your body, e.g. your stomach. However, your stomach was not designed to hold memory, so storing things there manifests as pain/discomfort. This means that sometimes if you feel uncomfortable, it’s because you’re trying to remember too many things and you’ve put something important into a slot that it wasn’t supposed to go in.

For example, if you’ve forgotten to do something important, your stomach might feel vaguely uncomfortable. If you’re worried about something that hasn’t been brought to conscious attention, you may feel slightly queasy. If you’re nervous about something, you might feel jittery.

Some rationality techniques like focusing are designed to help you access the memory you’ve stored in your non-brain memory slots.

Muscle Memory

epistemic status: slightly grounded metaphor

Your muscles also have memory. The memory of your muscles is basically what you’ve used them for in the past, weighted by recency. To teach your muscles something, you just have to perform it over and over correctly. This is why it’s important to have good form when you practice sports, so your muscles learn the correct form.

Sometimes, you do something wrong so many times that your muscles learn the wrong thing. For example, sometimes people walk incorrectly, so the wrong muscles get used for walking. Physical therapy is repeating a motion over and over to reteach your muscles the correct way to move.

Graduate School

epistemic status: probably true, but ask around

Getting into graduate school can be roughly modeled as a tiered process of various requirements. The best case scenario is if there is a professor that is excited to do research with you. If this is the case then you’re basically in. The second best case is having extremely good recommendation letters. The third best case is having a good research statement/research experience. Finally, if you have none of the previous things, you can do better by having a good resume/transcript. You might note a similarity in applying for jobs. Having a connection at the company you want to work for is best, followed by a referral, followed by work experience, followed by a good resume.

One implication is that “networking” is probably more important than you think. As a side note, people are often confused by networking. A strategy that works well for me is sending someone a message saying “Hi, I think I would enjoy talking to you. If you think you would enjoy talking to me, please find a time at https://markxu.com/talk”

Friendship

epistemic status: works for me, but is probably false in general. Also note that I don’t have that many friends, so “works” might not be the right word to use (I also don’t want more friends).

The way close friendships get formed quickly is by mutual disclosure of increasingly personal information. One good way to do this is to unilaterally disclose personal information about yourself, taking advantage of reciprocity effects to mutual disclosure. This should generally be done subtly, e.g. sharing a slightly-too-personal anecdote that’s relevant to the current conversation.

Another way to make friends is through “mere exposure” type effects. If you spend a few tens of hours interacting with a person in the context of school/work, you’re in a pretty good place to be friends with them. Initiating interacting outside of the usual context you interact with them is probably the best way to proceed.

Friendships also have signalling hierarchies. At the beginning, you’re nice to each other. Eventually, you start insulting each other because “you’re stupid” no longer means “you’re stupid” but instead means “we’re close enough friends that I know you won’t take me calling you stupid to mean I think you’re actually stupid”. At some point the hierarchy flips again, and you start complimenting each other. This implies that you can escalate friendships by being slightly above the current place on the signalling hierarchy. For example, if I’m not quite sure that I’m supposed to start insulting my friend, I can go ahead and do it; most of the time, the result is that we get closer.

OODA

epistemic status: useful

The OODA loop is a model of decision making. In this model, decisions are made by first observing the environment, then orienting to possible actions, deciding which action to take, then acting out that action.

For example, I might make a move in chess by first observing the chess board, orienting to possible moves, deciding which move is best, then acting on that move.

When decisions go wrong, it’s often because of a failure at a specific step. For example, if I turn in an assignment late because I forgot it existed, that’s a failure to orient. If I took too long to do it because I didn’t use the textbook, that’s a failure to orient. If I didn’t prioritize it properly, that’s a failure of decision. If I decided to do it but never got around to it, that’s a failure of action.

One interesting thing to consider is the speed of an OODA cycle. If I can OODA in 5 seconds while you take 10 seconds, I can execute two decisions in the time it takes you to make one. In particular, I am “inside” your OODA loop, so I can outmaneuver you.

For example, if you’re a big business and I’m a small business and you have a long development cycle, if you decide to start developing a new feature, I can observe the fact that you’re developing this feature, orient towards developing it myself, decide that’s a good idea, then act and roll out the feature all before you finish developing your feature. Even though you decided to develop it first, I was able to finish it first.

One way to think of agile development methods is organizations making a deliberate effort to shorten their OODA loops.