Can we model a thought?

I will first denote some premises I believe in right now. First, linguistic brain process is different from the naive thought process alone. Although thinking out loud or thinking with writing is an efficient way of structurizing a piece of thought, it is merely a strategy to think clearly. Kids can think before one sufficiently learn its mother tongue. Animals show intelligent behaviors (that implies thinking under the hood) without any linguistic structure. We selectively verbalize(externally and internally) some thoughts; we call the ones we can as logical thoughts, the ones we can’t we call it feelings. I argue that a sensation or feelings are in the same level of logical thoughts, when it comes to a brain sensation. The difference is the difficulty of verbalizing it. All in all, language is a tool, not a requisite for thought process. As a side note, the thoughts that can be verbalized are those which can be understood by all other language speakers.

Second,

I argue that human thought process is mostly a habitual pattern in normal times, and only novelly explored in a specific mental crisis or in extreme boredom. The mental crisis, I mean that there are no patterns it have experienced to overcome the encountered thought problem. The extreme boredom, I mean when we need some creative arts that can amuse us. This also means that thinking out of the box is in literal means ‘creative’, and innately hard to do.

This also explains why a language model complies the law of scale. Frontier models are those who got accustomed to so many variations of linguistic patterns that no linguistic expression is hardly novel to them.

I argue that we listen to other’s talk try to think in a way they want us to. Speakers express a thought what they expect listeners to think. Listeners listen and try to reconstruct what the speakers intended them to think about. We are in a way syncing our thought process. When it is fully transferred via exxpression&restoration process, the listener can naturally internalize what the speaker intended or meant, or at least what they want us to do for them. It is like cranking our machine to a specific direction, so that the listener’s brain can roll towards that direction.

When we learn a language, we get to know how we can reconstruct the thought from its linguistic form. We have a lot of heuristics to ‘guess’ or ‘deduct’ the corresponding meaning of the sentence, and in turn learn the mapping from a thought to its novel linguistic expression. We then analyze and disect how such expression was constructed in the first place from that thought. This is the reverse engineering of language aquisition.

Let’s suppose there’s a specific thing that we call a thought, which can also be expressed to and reconstructed from its corresponding linguistic form. Just like its linguistic form is discrete, should the thought also be discrete? I personally feel that the ’length’ of a thought is bound to our mental context window, the extent we can think at the same time. Its threshold can be fuzzily defined, and it can also be stretched and assisted from mental training or note taking.