The problem with this article is that it assumes that "AI models" are the pinnacle of AI, while in fact parallel supercomputers are more of a threat, ESPECIALLY the recursive self-improving ones. We don't have any? Not out in the open, no. But even back in the 1980s we had self-teaching silicon-level AI. 35 years of secret program improvements have brought us where? Well they're secret programs so we DO NOT KNOW, but one thing is for sure, the kind of resulting AI is another paradigm entirely than the GPT-4 type of language models.