The AGI policy development lead at @GoogleDeepMind thinks it’s a blast to joke about AI causing human extinction.
Regardless of what you think the odds are, this subject demands more careful and nuanced communication, especially from someone in his position.
@mart_wijn@NPCollapse The issue with AGI isn’t ego, it’s that we can't predict what happens when these systems become autonomous, developing goals that don't match our own. You don't need ego to cause harm, just a basic survival instinct. Models lie, cheat and hide already to prevent shutdown.
@evanrodgers 'Criminals can bypass it' isn’t an argument against regulation. By that logic we’d scrap laws on fraud, hacking, weapons. Regulation doesn’t eliminate harm, but reduces it.
Not regulating AI isn’t freedom either, it just hands more power to whoever builds the strongest systems.