The baseline here should be zero civilian casualties though. You could just as easily say "consider a new driver who got a car, and hasn't used it yet, vs someone who has killed an innocent bystander over 50% of the several hundred drives they've been on". I'd obviously rather take the new driver since the expected outcome in driving is closer to an accident every 20 years or so.
We don't have information from more than just one country, so we don't actually have a good accepted baseline to say a "new drone-having country" is good or bad... other than the fact I feel very comfortable saying that drones simply should not be used unless there can be greater than a 99.9% chance of killing zero civilians.
But there have always been civilian casualties in war. Why do drones need to be perfect in this regard, when nothing else in war ever has been? This feels like using airliner reliability as the standard that self-driving cars need to meet, instead of accepting that they're a net benefit as soon as they're safer than cars driven by humans.
More like a new driver versus a quintuple-DUI alcoholic who just got shitfaced again and mowed over a dozen members of a wedding party.
> China's track record is wildly better.
Consider a new driver who just got his license, versus someone who's been driving for 20 years and has been at fault in one fender bender. Would you consider the former to be the safer driver of the two?
Or consider an 18-year-old who has no credit history yet (and so no late payments), vs. a 50-year-old with a well-established credit history who made a credit card payment a few days late once 5 years ago. Which one of them has better credit?