Earlier quoted context omitted.
The “how many R’s are in the word strawberry?” problem can’t be solved by LLMs specifically because they do not have access to the text directly. Before the model sees the user input it’s been tokenized by a preprocessing step. So instead of the string “strawberry”, the model just sees an integer token the word has been mapped to.
I've tried it out and found that some models can answer the question if it's phrased right. And pretty much all models get it right if you also spell it out letter by letter to solve the problem you pointed out.
Language is not essential for the cognitive processes that underlie thought
421–424 of 424 posts
Re: Language is not essential for the cognitive processes that underlie thought
#422Earlier quoted context omitted.
> As some who has a dis-harmonic intelligence profile, this has been obvious for a very long time. In the family of my mother there are several individuals struggling with language while excelling in the field of exact sciences. I very strongly suspect that my non-verbal (performal) IQ is much higher (around 130) than my verbal IQ (around 100) I used to rationalize to myself along similar lines for a long time, then…
I'm a brilliant genius according to IQ tests. Think me arrogant or conceited or whatever - that is literally the truth, fact - proven many times in the educational system (I was homeschooled and didn't follow any sort of curriculum and was allowed to do whatever I wanted bc I kept testing higher than almost everyone) and just for kicks also - the last time I took an IQ test I was in my late 20s and a friend and I had…
I’m dumb - I’ve skimmed through a few Wikipedia articles, my hobbies include forgetting passwords, misunderstanding instructions, and being confused by even simple math. It disturbs my friends how little I grasp most things.
One time I scored so low on a test that they didn’t even bother to tell me the number. I’ve always known I’m just not sharp – I see the world through this fog that most people seem to walk right through.
And honestly, all of that is true, but it’s a pretty bad way of measuring stupidity. The more you don’t know, the more free you are, but also trapped. Ignorance alone doesn’t make a person unsuccessful, broke, or confused, though I’m living proof that it sure doesn’t help. That’s not a good indicator of much, though. There are plenty of other things that are way more useful in life.
Every boss I’ve ever had? All smarter than me, without a doubt. They’ve taught me invaluable lessons, like how not to trip over my own feet. I’ve tried and failed at many things, and now I’m just an especially capable fool. Maybe something I’m messing up now will lead to something someday, but probably not. For many reasons. I could write books about how to fail spectacularly, though ;)
At the end of the day, the facts are what they are - being smarter than me doesn’t seem hard, but that’s not the point. What really matters is learning practical things, like how to do simple tasks without messing them up.
A Master Plumber is a genius plumber, and that’s genius you can learn. You could probably learn anything – heck, I could too, but I probably won’t. That’s life. I know more about ancient Roman sewers than I do about unclogging my own toilet, and that’s also life.
It’s not about what you know or how fast you can figure it out - it’s about what you actually manage to do. Even if everyone can do what you’re trying to do better than you, most won’t even try.
That’s my rant about how dumb I am.
Re: Language is not essential for the cognitive processes that underlie thought
#423Earlier quoted context omitted.
I've tried it out and found that some models can answer the question if it's phrased right. And pretty much all models get it right if you also spell it out letter by letter to solve the problem you pointed out.
Spelling it out works in my experience. So does asking it for a python program to solve the problem.
I think it's something like the counting parts of problems that current models are shaky with, and I imagine it's a training data problem.
Re: Language is not essential for the cognitive processes that underlie thought
#424Earlier quoted context omitted.
That's based on a well known fallacy, because analog models cannot exceed the computational power of Turing machines. The alternative position is Penrose who thinks quantum tubules are responsible for consciousness and thus somehow more powerful than TMs.
> analog models cannot exceed the computational power of Turing machines There is no reason to assume consciousness is Turing computable [1]. [1] https://en.m.wikipedia.org/wiki/Church%E2%80%93Turing_thesis
[1] Mark Bickford, Liron Cohen, Robert L. Constable, and Vincent Rahli. 2018. Computability Beyond Church-Turing via Choice Sequences. In Proceedings of the 33rd Annual ACM/IEEE Symposium on Logic in Computer Science (LICS '18). Association for Computing Machinery, New York, NY, USA, 245–254. https://doi.org/10.1145/3209108.3209200