Some People are Large Language Models
(twitter.com)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (95)
sorted by:
That second sentence is somehow scarier than the first.
I disagree that this has nothing to do with IQ. Language is only a component of IQ. I'm willing to bet that if you take a Nobel prize winning "LLM human" who speaks as if he has an IQ of 120 and test him, you'll find he has an IQ of 98. After which, much ado will be made of the unfairness of standardized tests and how "mathematical skill isn't intelligence!"
If you actually have an IQ of 125 (the hypothetical number in the tweet), I don't see how you can't develop a model of reality. While IQ tests don't check to see if you are capable of telling the truth, there is a heavy emphasis on visualizing and mentally manipulating complex three dimensional shapes. One of those "I can't picture an apple in my mind's eye" fuckers would score abysmally, no matter how nice they talk.
The balancing of aggregate test scores between fields (verbal reasoning, abstract reasoning, etc.) to create a singular IQ, is basically just arbitrary. The theoretical general intelligence factor, g, is a significant portion of all those scores, so it still generally holds true, it's just too fuzzy on the details. Say you switch the weighting from 0.5 & 0.5 of two tests to 0.3 & 0.7. You're probably going to end up with a similarish number of >125 IQ results, but who is above and below that threshold is going to switch around dramatically.
Personally I think the weighting that gives the greatest possible contribution of g to results, on a population level, is the correct weighting in terms of quantifying the physical performance of someone's brain as a computational organ, and thus an individual's actual mental potential independent of training. But if you apply that to IQ test results the disproportionately high verbal IQ people are going to be very disappointed with their results, and they also just so happen to govern all the institutions that legitimize anything social "science" related to normies.
I remember my test scores from when I was younger. They showed a graph of what I was good vs very good at. Added up at the bottom for the final score. Idk what method that was, but there was no weight given in scoring. The test was then used to improve my skills in the areas I didn't test very well in.
That's not what my children experienced in school AT ALL. The sole purpose of these tests in the MD school systems is to earn the schools money. That's where teaching to the test comes in. That's where weighted test scores come in. The tests aren't being used for the benefit of the child. It's monetization.
I remember only one question from the first one I did in kindergarten. I was asked how and airplane flew I distinctly remember saying it was powered by an engine that let it stay in the air because of it's wings. The last one I took was in the 4th grade and I scored a 132 while my best friend and rival scored a 120 something.
I think the fundamental thing that IQ test measure is a person's ability to comprehend, infer, and break down principles and rules. Can you think for yourself essentially
Do you remember if those results were used to benefit you? I was very surprised between the different purpose of tests between when I was young, and when my kids were young.
The change is my opinion on why cities like Baltimore can't pass a single child. The tests serve the child no purpose.
That whole tweet could have just said, that person would fail a new and improved Touring Test.