Mental Acumenh2

I’ve been thinking about what the difference is between thoughts, ideas, information, and concepts.

Then, there’s also thinking, deriving a conclusion, understanding, et cetera. A process that transforms a series of thoughts into a brilliant idea.

One obvious thing is that these are concepts in the realm of mental, mind, and cognition. Borrowing a model from computer science, these are data and computation. So let’s call them:

• mental representation: a structured internal symbol the mind uses to encode something about the world.
• cognition: a mental process that transforms representations across various fidelity levels.

The mental representation can go from low-fidelity to high-fidelity:

1. Senses
2. Perception
3. Information
4. Knowledge
5. Concept
6. Thought
7. Idea

What’s important is not how comprehensive the bullet points are. Rather, the subjective comparison between them in terms of level of fidelity.

The cognition can consist of sensing, perceiving, thinking, and understanding. The cognition process transforms mental representation from low-fidelity to high-fidelity.

There’s also memory in the cognitive faculty. The ability to store and recall mental representations is part of the cognitive process as well. I’m not going to talk about the difference between short-term and long-term memory.

Discrete Thoughtsh2

What is interesting here is that reality works in a continuous fashion. Time flows continuously. Space exists continuously. We sense inputs continuously.

Only when a recurring pattern is identified and categorized, a discrete meaning emerge. If we’re able to capture that and symbolize it with a name, we will have a stable symbol for that specific, discrete meaning. Examples are cat, house, glass, which are nouns. It can also be more abstract such as globalization, hard, biology. What matters is that each of them is different, distinct from the other, discrete.

Logic and reasoning also operate in a discrete context. Computer works in discerete, 1 and 0 (binary).

So I think, the more we can identify patterns, categorize them, name them (symbolization), and segregate the symbols, the more we can gain clarity of thinking.

Strong Engineerh2

Thinking simplistically, the technical talent of an engineer can be reduced to mental acumen and domain-relevant skills. The ability to hold various complex concepts, performs transformations on them (mental gymnastic!), store temporarily, recall at will, and manipulate it again is the important raw-ingredient. Combine it with domain-relevant skills we get a comprehensive and robust, I’d say, framework how to hone and test an engineer’s technical talent.

Read Investable Candidate for my complete take on engineering talent.

Connection to AI and vector embeddingsh2

One of my criticisms about the AI industry is how the industry heavily relies on vector embeddings, a series of floating-point numbers that we don’t fully understand. We try to derive new training/inference methods, new computing processes (CPU->GPU), and new papers/research on a foundation that’s not robust.

But, thinking about it, we don’t necessarily need to fully understand the continuous blob of floating-point numbers. What matters are the symbols, the distinct meaning of those vectors. In fact, not fully understanding them unlocks the ability to represent reality more completely as floating-point numbers operate continuously. Though limited as CPU has 32/64/… bit limit :)