

LLMs are just massively-multidimensional maps of human language use. It is academically interesting to have developed both the map and a method for plotting a course through language-space using a prompt as an initial vector, but human intellience is not in language. Rather, language is part of human intelligence, and mapping it to ever more computationally-expensive distances is never going to chart a path to the digital mind that all the tech billionaires are so desperate to enslave.
You know a good heist always needs that genius mechanic who can bodge a mechanical solution to such practical problems, or just have a flatbed truck with a crane or something.
If this were a movie, it’d end up with an 18-wheeler and whatever nonsense would be required to steal a whole train car. XD