Thoughts: Knowledge and Intents

A.I. that Thinks Creatively, Part 5

From Narrow To General AI
9 min readApr 27, 2020

[This article is a work in progress]

Part 1, Part 2, Part 3, Part 4

In the previous section you saw how you develop skills. Skills help you face and address problems instead of avoiding them.

In this final section, you’ll see how the ideas in all the previous sections come together to enable you to build useful thoughts. I’ll start by showing how you make two general categories of thoughts: knowledge and intents.


By “knowledge”, I mean a thought that reflects how things are in the world. What the capital of Thailand is is a type of knowledge, as is what cereal you ate yesterday morning. The second example is specific to a particular time and place, but they are still both thoughts that reflect how the world is or was.

Some knowledge is so general that it lets you predict the future. It can be as mundane as what happens when you turn up the volume of a speaker. It may also be knowledge from the sciences, such as what happens if you move a magnet next to an iron bar, or mathematical knowledge like what the sum of two numbers is. Unlike these factual examples, memories are will likely never repeat, such as who you met on the street yesterday and what he said.

These two are not significantly different. The main difference between them is their consistency. A memory that is consistent enough becomes knowledge. For instance, if you see your brother eat scrambled eggs this morning, that is a memory. It has no real predictive use. If you see him eat the same tomorrow and the next day, then it becomes knowledge, namely “my brother eats scrambled eggs for breakfast”. You could use that to guess what he will eat the day after as well.

With both memories and knowledge, one experience (a sight, sound, etc) causes you to produce a related, useful thought. Here are some examples:

Thailand + capital ⇨ Bangkok

Speaker volume + turn up ⇨ loud noise

Me + breakfast + yesterday ⇨ Bran flakes

To be clear, these are all concrete memories, they have not yet been turned into abstract concepts. These connections all reflect something the world showed you. You could call them all “truths” in the loosest sense of the word. So how do you gain them?

One prerequisite of knowing the truth is that you have to be interested in it. Truth only seems to appear when you ask a kind of question; and the answer matches the question you asked.

Look around you. What do you see? Your answer, i.e. the truth, may include things like “pen”, “desk”, or “person”. But it is unlikely to include “politics”, “1 + 1 = 2”, or “a method for scrambling eggs”. Why? Because the latter three don’t answer the question asked. So how do you decide what qualifies as a valid answer to the question?

To start, you could never identify an object you weren’t familiar with. Before you can identify something you have to have a template for it.

Take a look at the image below:

What is it an image of? You likely don’t know (I didn’t either). So when you are answering the question, you try to liken it to something you already know. You might guess that it’s a machine for spooling rope, or perhaps a wind-up toy. Both of those are types of things you know how to identify from previous experiences.

When I asked you what it was an image of, your motivation at the time was to identify an object that you don’t immediately understand. As long as you haven’t identified it, you have a “problem”. Questions without answers are problems — you try to neutralize them with an answer. Identifying it, by which I mean the act of looking at it and giving it a name or purpose, would be the solution to this problem.

I’ll explain now that the image above is of a machine that slowly and automatically turns food over a fire to roast it. A heavy object pulls the rope down and the gears turn at an even speed.

You now have an answer. Even if you don’t exactly see how that could work, there is enough information in what I’ve said for you to accept it. You’ve gained a new piece of knowledge.

Assume I’m being truthful. What did I say that solved the problem and made you store my answer as a new thought? The phrase “a machine” played some part. If instead I had said that it was a picture of a man who wrote romance novels, you would likely not have accepted or even remembered that answer, except perhaps as a joke.

You saw in the previous article that every thought is created by a ‘skill’. The skill you are using here is your ability to identify something as a machine. Seeing metal gears, and reading that something is “a machine”, together, make for an acceptable solution to the original problem, that of identifying an unknown object. In fact, you probably already guessed that it was a machine before I told you.

After that, every additional detail I gave about the machine you either remembered or weren’t able to retain. Which one happened, depends on if you had the skill in place for it. Those of you who are already familiar with how weights and gears can be used to create steady rotation in clocks probably remembered that this was an example of that, when you saw the rope and gears, and heard my explanation. The rest of you may have glossed over that detail. Most of you probably remember that it is used to roast meat over a fire because this is a purpose you are already familiar with.

What you’re starting to see is that you can only remember a fact if you have an existing template for it. If my explanation instead had used arcane and convoluted technical jargon, like “retro-encabulator”, then only those of you familiar with this specific jargon would be able to remember it. You can only learn what you already have the skill to learn.

[Work in progress: how exactly, and when, thought connections are made]

You’ve acquired new knowledge. You’ve connected the sight of the machine above to other pieces of information I gave you. Most importantly, you connected it based on a template of problem-solving that removed the ‘unknown’ and replaced it with one of your prebuilt categories.

This raises a question: where did your identification skills come from? Why is it you can identify things such as machines, or food, or other more specific categories? To understand that, you have to look at the motivation behind your desire to identify objects. I mean the “parent” motivation that created both your desire to identify the unknown, and the types of things you would accept as answers.

The parent motivation in this case is your aversion to harm. Not pain itself, mind you, but your aversion to things that you predict could cause you pain and discomfort. In this sense it is quite a low-level aversion and arose early in your life. As a child, you grew averse to the unknown because it could be harmful. You can feel this basic aversion creep in when someone asks you “hey, what is that?”

Generally, though not always, questions that start with the word “what…” trigger this identification stress. Given the fluid nature of languages, you can’t connect stresses and skills to individual words, parts of speech, or phrases, per se. You learn through experience to connect various combinations of words, sights, sounds, etc, to aversions.

Identifying the unknown is only one of the many motivations that you use to gain knowledge. There are other motivations, such as the desire to prove yourself to others, or the desire to help and teach others.


Given everything above, intents are comparatively easy to understand. By ‘intent’ I mean a thought that visualizes a preferred outcome. Your career aspirations are intents. Your mental shopping list is an intent. How you will answer your friend’s question when you meet her again is an intent. In each case, as above, experiencing a situation causes you to imagine some other thought, which is your goal.

A clear example of an intent is the one you have for a drawing before you start it. You see a blank page, and in your mind you see the outline of the intended image, what to draw, and where: a bird above a woman in a field. As you draw the field, you then imagine more details, such as hills, bushes, etc, and where they should go. Each of these steps is a minor intention of its own. In each case, you first see something in your mind, then you try to realize in the world.

Now let’s look at details of generating intentions, as seen when you play a game of chess.

[WIP: chess example]

Intents and Knowledge are the Same

In the previous two sections you saw two different ways of using thoughts: forming knowledge and having intents.

  1. Knowledge, broadly speaking, is about what the world tells you is true. Memories, predictions, facts all reflect how the world was at some time, and may be in the future.
  2. Intents are what you’d like to see happen. Plans, strategies, designs, wishes, even fantasies are types of intents.

But how does your mind separate these two types? From what you saw above, it seems both are created in the same way, by using a skill to solve a problem in thought.

They are, in fact, not separate. Although intents and knowledge seem incompatible they are, to your mind, the same thing. Let’s look at an example to see what I mean.

When you see the following image

you might think of the word “pen”. This seems as much a statement of truth as any. Surely, the English word for an object falls into the category of knowledge.

But English is an arbitrary language to choose. You could also think of the French or Russian word for this object. Nothing in the pen itself inherently implies the word “pen”. The word you choose for this object is based on what your society agreed to call it. If tomorrow your society all decided to call it “blargle”, that would become the new truth.

So why do you think of the word “pen” when you see one? Well, if you want to tell an English speaker what you saw above you would use that word. When you’re having a conversation about this image, it is useful to have the word pop into your head immediately on seeing it. That way you can say out loud what is in your head, instead of searching around for the right sounds. The word is your intention as to what you will say.

Let’s look at another example. Imagine a friend of yours asked you to multiply 121 x 131 in your head. He then leaves to give you time to work it out. You ultimately come up with an answer: 15,851. You memorize the answer to give him when you next meet. You have an intent for what you will tell him. But it is also a mathematical truth.

As you saw in the previous article, every thought solves a problem. You only remember a fact if it was useful for you to remember it, if remembering it had a purpose.

This is true even of your personal memories. Imagine you went to a store to buy carrots for a friend. Glancing at the shelf where they should be, you find it’s empty. You remember the empty shelf and go back to your friend. In this way, the empty shelf is ‘truth’. But if you had gone to the store to buy him pizza instead, you might not have that memory. You only created it because you have to explain to your friend why you came back empty handed. It’s what you intend to tell him.

You often don’t realize why you create memories and what use they serve. Though it may seem to you that you remember arbitrary facts, there is too much in the world to remember for you to retain it all. You might remember a funny incident because it would be a good story to tell others. You may remember spiritual truths and aphorisms, because thinking about them later will bring you happiness or success: they are how you intend to make yourself happy.

Every piece of knowledge is also an intent. Even a scientific truth such as the model of the atom, or the number of planets, is an intent for what you will tell others about it, which is why you think of these facts in your local language.

You often hear it said that “people believe what they want to believe”. A more charitable way of expressing this sentiment is that people believe what it is useful for them to believe; or rather, people know what it is useful for them to know. This way of saying it makes it seem more reasonable. Why would you do anything else?

[Work in progress: Intents do not mean you have to act to make them happen, they can be desired outcomes]

[Work in progress: A.I.]

Part 1: Imagination

Part 2: Solving Useful Problems

Part 3: Motivation

Part 4: Defining Skills

Part 5: Thoughts: Knowledge and Intents



From Narrow To General AI

The road from Narrow AI to AGI presents both technical and philosophical challenges. This blog explores novel approaches and addresses longstanding questions.