3 ms·

Already implemented and being tested.

simonh 11y ago
I don't really like it when people demand references to support opinions, it's usually a jerk move, but you're making a statement of fact. Can you point me to any information about these tests?
- atemerev 11y ago
  One example: http://www.techinsider.io/british-taranis-drone-first-autonomous-weapon-2015-9 http://www.techinsider.io/british-taranis-drone-first-autono... Short googling for "autonomous weapons" will yield many more results.
  - simonh 11y ago
    That does exactly what I describe. It flies automatically, but target selection and weapons deployment is done manually. The article is quite clear on this. It doesn't recognize terrorists automatically, decide whether they're on a kill list by itself and decide where and when to attack them without human intervention. I can't reply to atemerev, so I'll do so here. Are you even reading my posts? Autonomous flight is largely a solved problem. Recognizing a gun in a silhouette isn't even close to good enough. What if he's a hunter, or a Pakistani soldier, or just wants to defend himself? The kinds of strikes drones perform are driven by intelligence., e.g. form the article, 4 people on camels traversing a given valley. What if there is another group of 4 people? What if one of them splits off before the drone is in strike range? What time window is reasonable given how fast they could traverse the terrain? What if the image now shows that actually there's a child with them that wasn't in the intelligence report? Humans can interpret the intelligence report, evaluate incidental factors such as travel times and terrain, recognize new data points and come to a reasoned conclusion based on hundreds of factors about whether or not to engage. Computers simply cannot cope with these sorts of problems.
    - atemerev 11y ago
      OK, consider this: Driving a car in San Francisco is harder than flying a drone. Google had built autonomous cars that can do that. Hence, there are autonomous drones. Finding your face among millions poorly made pictures taken from mobile cams is _much_ harder than determining whether somebody's silhouette in a video feed contains the profile of a gun. _I_ can take Python and write this kind of program (it will not be military-grade, but will work most of the time). Hence, military also tried and tested this. Especially given the telemetry and infrared data available. Image processing and recognition is now at least on par with humans. See clarifai.com demos for examples.

I don't really like it when people demand references to support opinions, it's usually a jerk move, but you're making a statement of fact. Can you point me to any information about these tests?

atemerev 11y ago
One example: http://www.techinsider.io/british-taranis-drone-first-autonomous-weapon-2015-9 http://www.techinsider.io/british-taranis-drone-first-autono... Short googling for "autonomous weapons" will yield many more results.
- simonh 11y ago
  That does exactly what I describe. It flies automatically, but target selection and weapons deployment is done manually. The article is quite clear on this. It doesn't recognize terrorists automatically, decide whether they're on a kill list by itself and decide where and when to attack them without human intervention. I can't reply to atemerev, so I'll do so here. Are you even reading my posts? Autonomous flight is largely a solved problem. Recognizing a gun in a silhouette isn't even close to good enough. What if he's a hunter, or a Pakistani soldier, or just wants to defend himself? The kinds of strikes drones perform are driven by intelligence., e.g. form the article, 4 people on camels traversing a given valley. What if there is another group of 4 people? What if one of them splits off before the drone is in strike range? What time window is reasonable given how fast they could traverse the terrain? What if the image now shows that actually there's a child with them that wasn't in the intelligence report? Humans can interpret the intelligence report, evaluate incidental factors such as travel times and terrain, recognize new data points and come to a reasoned conclusion based on hundreds of factors about whether or not to engage. Computers simply cannot cope with these sorts of problems.
  - atemerev 11y ago
    OK, consider this: Driving a car in San Francisco is harder than flying a drone. Google had built autonomous cars that can do that. Hence, there are autonomous drones. Finding your face among millions poorly made pictures taken from mobile cams is _much_ harder than determining whether somebody's silhouette in a video feed contains the profile of a gun. _I_ can take Python and write this kind of program (it will not be military-grade, but will work most of the time). Hence, military also tried and tested this. Especially given the telemetry and infrared data available. Image processing and recognition is now at least on par with humans. See clarifai.com demos for examples.