>>21806302>>21806306>IT CANT COUNT DAH NUMBER OF RS!! ITS NOT INTELLIGENT!1 JUST A PARROT MACHINE!!The finest /g/ midwit cope.
Some tasks are inherently hard for LLMs, just like how humans can't calculate 99999x88888 but a machine can do it in a nanosecond.
Intelligence is about two things, learning and generalization.
>If Bob fell on a slippery road, Adam might also fall on a slippery roadA human naturally infers this prediction, and LLM does it as well. In practice, there is no fundamental difference as AI can already generalize its knowledge based on similarities. What is the human thought if not a constant chain of such inferences? Sure, a brain has some special circuitry that reminds you when you "hallucinate", but this doesn't make your thoughts intrinsically different than an LLM. To an LLM, real memory and predictions are not separate like in the human brain, but even the most pro-human Luddite knows this is just a temporary technical issue.
Similarly, the human brain has special circuitries that recognize graphical patterns which make them excel at them IQ tests. It's not simply using its basic neural networks and brute forcing it like current LLMs. Some humans actually lose such abilities because of brain damage, and they find it hard to tell the differences between images and recognize patterns or faces since they're relying only on their primitive neural nets, just like LLMs we have today.
I'm not saying that LLMs technically work in a similar way to a biological brain, just that the human thinking is not inherently superior. Intelligence itself is just not as sophisticated a design as some may think.
https://www.youtube.com/watch?v=Rh1ejzNyCso 
