What my stroke taught me (2017)

This is quite a swampy bit of territory you're joining me in here so please excuse the mess.

I'm going to attempt a good faith interpretation here so we're clear. I want to understand your position, because I think it's, if nothing else, very interesting to contemplate these matters.

First of all, I have to clear the obvious debris out of the way.

I can read and I can cogitate lingual constructs. That includes creating monologues. Sometimes I "role-play" some professor or something and talk through my problem out loud - when alone. I tend to do this with philosophical problems, but it's unnatural and quite entertaining. The last time I did this must have been a few months ago when pondering Hume's problem of "necessary connection" or Kant's .. everything. I think this is the closest I get to "inner monologue" and it does have its uses, because language in and of itself has a computational character of sorts. By speaking the words out loud I sometimes trigger snippets of verbal knowledge that wouldn't be triggered otherwise.

If this qualifies me as "having inner monologue" already then we're done. By that definition I am in your club and I'm fine with that.

What would still remain however is that those role-playing sessions are deliberate and unnecessary. I can quite easily go without them. It's a trick I've picked up, but it's by no means the only way to make progress. Their nature is comparable to rubber duck debugging: by explaining your problem you work through it differently somehow than if you'd remain "stuck in your head". I can go without rubber ducking, but it's still a useful trick.

Another way of approaching this problem is through programming - assuming you code. You don't literally think "Ah, I see a linked list here and it gets initialized here.. [etc]"? Or maybe you do? I can tell you quite firmly that I don't. If you also don't then just assume I go through my life in the same abstract wordless mode.

> I don't buy that there is "another inner language of thoughts", of the same or greater complexity than our known language, which has not been discovered or described at all.

I find this quite interesting, because I don't understand it. Our language is made of rules and in and of itself has no content. It's literally made out of just rules. Any intelligence has to be brought into it. This was quite clear with GPT-2 era transformers spewing sentences like: "The bear and the snake are wearing stairs". You have to know quite a lot before you can explain why this is nonsense despite it being a valid sentence and it's not clear to me at all that this knowledge itself is purely lingual.

This "inner language" you speak of is just raw cogitation and surely you employ it just as much as me but we assign different words to things. This "inner language" is not to be put "besides" our natural languages, it's the core of it. Without it, there would be no "natural language". I view natural language as a communication convenience, built on top of and not besides raw wordless thought.

I'm not 100% sure how to handle your analogy because, as you've indicated, it's quite tricky, but I'll try something. Sorry for the mess again:

I agree that higher level abstractions enable cognitive moves that just won't be obvious from raw assembly.

Where we disagree, I think, is that these "higher level concepts" are somehow spontaneously lingual by nature. In my view of things these higher level concepts are still raw thought, but .. higher. More complex. Encoding these new more complex cogitations necessitates more complicated lingual structures.

Let's put it this way: do you first cogitate the "class concept" or does the "class syntax" enable the "class"-cogitation? Perhaps for a programmer it's the second, but for the one inventing the abstraction surely the cogitation must come first and, lacking existing lingual constructs, must be cogitated in a purer form.

My problem would thus be: how could you possibly cogitate something that doesn't yet exist in language if all you have is language? "class" does not "fall out of" assembly naturally - inside of it. You need to bring cogitation from outside and then realize it and be able to think of an encoding for that new concept to communicate the idea.

Thank you, these insights are extremely interesting!

Just FYI it’s pretty rude/insulting and not interesting to dismiss someone’s shared experience based on nothing but your own armchair feelings, and even moreso to concede if the person is telling the truth (how on earth could a person be lying or mistaken about this?) that they couldn’t be a functioning adult.

It’s honestly quite a narcissistic view that someone could not possibly think differently than you, and dismiss it out of hand. The other people in this comment thread sharing similar things make perfect sense to me.

My point is that "I hate that guy, but I must greet him friendly, or else I will be in trouble" is already being concluded by the brain at the semantic space way before running the language encoder and playing the inner voice. That voice already knows what is going to say before saying it.

The brain can extract those conclusions in mere milliseconds and act in consequence, and that doesn't mean being guided by feelings.

I'd argue that skipping the language lets you think in more complex scenarios where visualizing something is required, like when playing a sport, painting, or composing a melody.

Removing the noise of language also allows you to think in parallel and solve problems faster. In my case I'd say I can run a couple of "threads" only, but I wouldn't be surprised if top chess players could run a dozen.

I think you have locked on to the LLM-viewpoint quite strongly. I don't think I can agree with a lot of your presuppositions. Like "the rules are just there to make the pathfinding efficient" and "language is just an extremely efficient encoding of reality".

While very interesting I'd have a hell of time debating those issues in this kind of medium. If you take that as your starting point then I do understand why you have a hard time grasping the "lack of monologue"-thing. If language is the source of all raw cogitation then how could anyone possibly lack monologue and still function? Am I getting it? I think I get it and I'd agree if I'd take your framework as true.

> impossible to use without language.

Now I understand those types of statements. I am totally not in a position to say you are wrong, but I'd be careful with this because you seem to be a holding a flashlight and saying there is light everywhere you look.

All I can say is I just don't think language plays that large a role in cognition. Why not? That's an excellent question, but I cannot possibly do justice to that here - besides being a complete amateur. I can leave some clues or things that count to me as clues, perhaps. We are surrounded with enormous quantities of beings that utterly and completely lack language yet show tremendous intellectual sophistication. Our own children are completely non-verbal for a very long time and sure, while they won't be winning Nobel prizes at that stage, you can't possibly claim they are incapable of the very same cognitive maneuvering you showed in the coding example (abstraction, generalization, etc). To me it's just a matter of scale. We gained function, we added something new and fancy to the preexisting Mammalian machinery with language, but I don't think it's fundamental to raw cogitation.

> 2. There is another "hidden" code, capable of the same complexity and efficiency, that has not been discovered or really described by anybody, that can be translated immediately to spoken or written language.

3. All people use the same raw cognitive substrate but some feel the need / are wired to overlay it with running commentary. A running commentary they themselves can inspect, I might add. That's an interesting bit. If it is so foundational, what's inspecting it? Language again? Like an LLM-reasoning type of agent or agents?

I notice the question "What happens before the inner monologue? A deeper level of monologue?" didn't get addressed and I'm quite interested in it. What process produces the monologue you perceive as necessary for cogitation? It surely can't be a monologue again? At some point you have to admit these dialogues just "pop into existence" pre-baked, right? Out of what? What produced that if not some raw, unobservable, non-verbal cogitation? That non-verbal cogitation has to be capable of some kind of computation to make that happen or am I making some kind of error?

Honestly, that last question is rhetorical. I'm making all types of errors, but I hope you see I'm kind of on the fence on the language thing. I'm fully open to the fact that I just might not be perceiving the monologue.