A handful of bros in Silicon Valley and Wall Street have wagered much of the global economy on being able to hyperscale stochastic parrots into AGI super beings.
Many in the know have long known that is a losing bet as LLMs are mathematically guaranteed to produce plausible sounding errors (too often called “hallucinations”). Sucking up increasing amounts of water and energy to train on more stolen information only results in marginal improvements.
All this talk of ‘needing to slow down the super scary AI’ is just a tacit admission from the AI CEOs that they are running out of venture capitalist marks to con and are looking for a government bailout - it’s only fair to be compensated if they are tapping the breaks to save humanity, after all.
These AI companies have no path to profitability (unless LLM to AGI transition is, in fact, possible and around the corner, to which it’s not), but can find a ‘nice’ home within increasingly fascist surveillance states.