More seriously, AI has no fidelity so its function as an intelligence tool is simply pattern recognition: generate the big spreadsheet of correlated traits.

That’s…alarming…on it’s own because correlation doesn’t necessarily signify anything significant, but it’s especially bad when if the method of achieving those correlations is opaque, which is often the case with large algorithms. It’s collective guilt presented as a spread sheet.

But if this opaqueness is simply added on top of the woo-woo opaqueness of the intelligence processes already in place, what this amounts to is a lethal Mechanical Turk: the machine inputs are controlled by actors with intentions and perspectives, and the correct Garbage In can create the desired “no, take the shot” Garbage Out, with the mediating step of the algorithm less of a proof than a ritual to make the killing impersonal and thus acceptable.

I mean, we’re already doing this…that drone strikes kill civilians is simply accepted practice, and there is a standard language that explains away that death…but AI “intelligence” provides the opportunity to infinitely expand culpability and thus achieve through mystification what the Russian Army simply achieves through shamelessness: if the target was hit they were a viable target.