A modern naval mine, for instance, is deployed and waits for an activation to autonomously engage targets. Does that meet your criteria?
While there is some room to nitpick his examples, they're largely relevant despite your dismissal of them -- and that's part of the problem. For example, an autonomous homing artillery shell might not fit your definition as it requires human interaction to initially deploy it, but once deployed, it chooses its own targets. The same is true for many other potential uses of autonomous weapon systems, but you seem hung up on novel new usages or extended periods between deployment and effect.
At best, the autonomous/non-autonomous weapon line is a blurred smudge on the road in our collective past. I'll agree that there's a larger potential for wider use going forward, but these tools are not new.