4 ms·

My biggest concern with this open letter is that it acknowledges that the way to avoid having (currently weak) AI in army requires considering humans to be expa

My biggest concern with this open letter is that it acknowledges that the way to avoid having (currently weak) AI in army requires considering humans to be expandable. Those humans are both more intelligent than the equivalent AI, and more prone to error. They will stay better at murder (accidental and otherwise) while AI will slowly become better at risk assessment, avoiding unnecessary deaths. While humans will still get PTSD, war machines will only rely on analysis (and human orders).

Strilanc 11y ago
"Avoid unnecessary death"? That's a pretty rosy picture of how a weapon programmed to perform an ethnic cleansing, a terror attack, or to herd civilians out of an area would act.
jonathankoren 11y ago
What other choice do you have? Even if you didn't have someone in the loop pulling the trigger, then you'd have a bunch of engineers (or more likely Amazon turkers) sitting back and reviewing after action video filling out precision-recall reports, that read "Misidentified school bus as troop transport at 1:37. Misidentified farmer as rifleman at 4:36. Misidentified sniper at 5:32. Misidentified market as weapons depot at 6:32" We know what this looks like. Just ask anyone in content-review[1]. [1] http://www.nytimes.com/2010/07/19/technology/19screen.html http://www.nytimes.com/2010/07/19/technology/19screen.html