/sci/ - Science & Math

>>16137584
>>16140473
>>16150993
>>16137584
>>16139707
The problem with your metrics is a LLM; i.e., a statistical machine that answers only the question "given an incomplete sentence, what is the most likely next word of the sentence?", with ZERO reasoning capacity (pic related), scores very highly on all your IQ-adjacent tests. It didn't outscore me, but only barely, so I'll bet it outscored all you retards big time.

If you're wondering about real IQ tests, like Raven's matrices, you're also too late. CogSketch, a literal Lisp script from Northwestern University, already scored 80th percentile on that, and that was old tech, over half a decade ago.

You see, silly piggots, multiple guess tests like SAT are little more than a game, and can be trained for just like a game, which is why a mindless 0-IQ bot can smoke you on them, no matter which "race" you claim.

Conclusion: there is no direct correlation between test score and intelligence.
If a 0-IQ bot beat you on what you consider an IQ test, then you have two choices. Admit the test means fuck all, or neck yourself. whichever you choose, I recommend you follow through, immediately.

>>16114354
>believing the [WS] IQ scam

>>16131684
>muh SAT
Even a retarded chatbot with ZERO intelligence is a top performer in the joke known as the SAT.
I used to be proud of my SAT score until I saw this

>>16057688
>How are your intelligence more than an autocomplete, then?
First see pic related (the point about the joke known as the SAT is unrelated here; just focus on the chat logs).

Then look at this video
https://www.youtube.com/watch?v=ubzwB4FgvxI
This video still applies to ALL large language models, even the latest and greatest. Gemini, chatgpt, claude, all are unable to autonomously decide that "My intuitive f(in)=out answer pre-encoded in my NN might be incorrect, so I'm going to apply computational methods instead"

Read this link, starting from "Computational Irreducibility"
https://www.wolframscience.com/nks/p737--computational-irreducibility/
When you understand this you will understand why it is impossible for ANY neural network to be intelligent. Ever.

General AI might be possible but a NN will be but one part of it. There is still a crucial element missing, and a NN alone cannot provide it.

>>16012881
>which factors should they "control for", OP?

>>16020890
>muh SAT scores
the SAT is nothing more than a game
even a retarded chatbot can score high enough to get into harvard