For folks who've been reading versions of this fun bit of AI military news, I just want to nudge you to Robert Sheckley's 1953 story "Watchbird", about anti-murder drones that go to... well, all-too-predictable extremes to carry out their mission. Long story short: SF has been all over the dangers of automating death machines for 70+ years. But apparently some folks thought those stories were instruction manuals? 🙃
I’d have to see the receipts on these stories.
Seems to me the drone would lose all its points if it killed itself without killing a drone.
I’ve read waaaaaay too much science fiction in my time. This sounds like too familiar a theme from sci-fi.
I'm confused by your statement. I literally referenced a 70-year-old SF story, so yes, I'm very much noting the fact that this is a familiar scifi theme.
The news item today is about a training simulation in which the "AI" found a workaround that involved damaging the operator to achieve its mission priority.
What exactly are you puzzled by?
If you want a more technical briefing, the story was originally reported among highlights from the RAeS Future Combat Air & Space Capabilities Summit, held in late May. Quick search for the subheading making a joke about "SkyNet":
@MLClark
Copy that! This is where I first read the report.