I'm by now past the point where I thought "this is a great tool that will help me programming" to the fear phase where I know it will destroy my job and career. But still I can't agree fully with the article. Hallucinations are not the issue. We can extract value from it even when it does hallucinate.
The only reason we're having this discussion is because AI is intelligent and therefore poses a threat to us.
That's what affecting me the most. Pure Claude hallucinated analysis that the poster has not really read; worse part is that I can ask Claude for the same analysis, so there's no benefit, only long questions that waste my time.
Carmack didn't invent BSP trees. He read about them in a paper and produced an efficient implementation that could render a 2D map in 3D using integer math on 386-class CPUs. He has always envied the mathematicians who actually come up with this stuff.
I think he did come up with "Carmack's Reverse" (a dynamic lighting algorithm) independently. But not even the "fast inverse square root" was his; that, oddly enough, came from what is currently probably the premiere AI company: NVIDIA.
But his implementations of these algorithms were indeed beautiful. I suppose AI could do something similar today, because the likes of Carmack had come before. But using AI for code is hobbling the creation of new Carmacks; and humans have the remarkable property that they can learn from other humans and still push the boundaries forward. When AIs learn from other AIs, model collapse occurs.
