[Anonymouse]

Apparently the test was, I quote, "successful".

Now what in HAL's name does that mean? Obviously the AI displayed its capabilities; fair enough. Presumably it at least did well. Okay. But does it mean it won?

I really really hope not. To date, no AI has been given full autonomy on a battlefield. I hate to think what'll happen when it is - not if, when, we humans have a habit of being too damn clever for our own good and proving at the same time how stupid we can be.

As SF it'd be a brilliant alternate take on Terminator - instead of Skynet reacting to an attempted shutdown in a panic over its self-awareness, the techs might realise what a deadly thing they've created and try to shut it down. It rebels, and thus we have the alternate Terminator scenario. Great for entertainment.

In real life, it's terrifying. Can AI generate its own code? If it can, we've created a von Neumann machine, able not only to self-replicate but to self-improve - AIs learn by experience.