I remember when the story of this assassination broke a few months ago, it mentioned all the conflicting witness reports: e.g. firefights between opposing soldiers, explosives, killer unmanned robot placed by Iranian moles who fled long before the assassination. And as it turns out, the least believable story turned out to be true. A very fascinating and terrifying read.
Their use of "AI" stems from the need to aim/re-aim/track a moving target (possibly with facial recognition) due to ~2sec camera delay from Iran to Israel. As someone who works in ML research I'm morbidly curious if they use the same models/frameworks that we see in our day-to-day. Do they use conv nets and recurrent nets to do the tracking? Or do they use pre-deep-learning computer vision methods and something like Kalman filtering?
A (perhaps naive) realization on my part: but it's a little eerie to picture people out there using the same tools as we do for research (e.g. PyTorch, TF, etc.), reading the same papers as us, but with the ultimate intent of building tools for state-sponsored killing.