The arguments going back and forth in these threads are truly a sight to behold. I don’t want to lean to any one side, but in 2025 I‘ve begun to respond to everyone who still argues that LLMs are only plagiarism machines, or are only better autocompletes, or are only good at remixing the past: Yes, correct!
And CPUs can only move zeros and ones.
This is likewise a very true statement. But look where having 0s and 1s shuffled around has brought us.
The ripple effects of a machine doing something very simple and near-meaningless, but doing it at high speed and again and again without getting tired, cannot be underestimated.
At the same time, here is Nobel Laureate Robert Solow, who famously, and at the time correctly, stated that "You can see the computer age everywhere but in the productivity statistics."
It took a while, but eventually, his statement became false.