If given massive numerical linear algebra coprocessor resources like all the GPUs going round, I think it would make more sense to use them for optimal control affairs to reduce power consumption, system-internal statistical modelling and optimal control e.g. sequential Monte Carlo / particle filters for network auto-tuning and higher-level SMP load balancing and system power management, and perhaps even tasks of the same kind but less frivolous than graphical rendering for video games and local statistical text generation kept naïve to a priori NLP and other concerns out of unscientific commitments to crude and misconstrued connexionism and motivated misreadings of Patricia & Paul Churchland's eliminative materialism as vindications of their beliefs that the vast majority of humanity are NPCs i.e. philosophical zombies, but not themselves. To the extent they're in big compute centres, things like physical simulations of the climate, the Big Bang, engineering optimisation problems like circuit layouts etc. could make sense. However, it appears that the AI/LLM majors are, in fact, too incompetent to construct the computing centres they claim to be. At this point questions should be arising as to where the money is actually going, as well as why the money managers themselves didn't ask that themselves before the scammers could fritter it all away on embezzlement, drugs, corruption or whatever they're actually doing with it all.
Or such were my musings inspired by
https://youtu.be/or8butOTUp8
I have no idea what the relevant hashtags would be. The most obvious ones are overrun.
