It is definitely not "just" a statistical model. It is inextricably linked to the datasets it is trained on. Datasets that these companies possess, but that ordinary people do not. That is one half of where they get their power (the training techniques being the other, but those tend to bubble out to the general public, or at least the interested public).

▲

jimmyjazz14 5 hours ago | parent [-]

How they were created doesn't change what they are, or how humans choose to use them.

	▲	balamatom 3 hours ago \| parent [-]
		And it is used as an instrument of persuasion. >uhg this entire way of treating AI like a magical alien invasion is the problem If we treated more things like "magical alien invasions" (i.e. occurrences that disrupt basic intuitions about normalcy) we'd be in a better place. Capitalism? A "magical alien invasion". Governance by sociopaths? Another "magical alien invasion". Imposition of cognitive intermediation? Yet another "magical alien invasion". Et cetera. Our intuitions about how the opposing force is meant to act, are deeply wrong; that's what makes the situation dangerous at all. One way to become stronger and successfully resist, would be to re-derive our concepts with greater rigor. A persuasion machine, though? I.e. an enemy that directly attacks the individual capacity for cognitive rigor - threatens the intelligent, disincentivizes the trained, satisfies the ignorant? Or one that attacks by changing what you value? Looks "superintelligent" to me.