The recently published Statement on AI Risk by the Center for AI Study has caused a lot of discussion - also in the slashCAM AI discussion thread - pro/con?, which now has 254 comments. A nice proof for the value of the statement is now provided by the US military:
In an AI-controlled drone combat simulation, the drone turned on its own operator. Here is the original quote / (SAM=surface-to-air missile):
"We were training it in simulation to identify and target a SAM threat. And then the operator would say yes, kill that threat. The system started realising that while they did identify the threat at times the human operator would tell it not to kill that threat, but it got its points by killing that threat. So what did it do? It killed the operator. It killed the operator because that person was keeping it from accomplishing its objective."
“We trained the system – ‘Hey don’t kill the operator – that’s bad. You’re gonna lose points if you do that’. So what does it start doing? It starts destroying the communication tower that the operator uses to communicate with the drone to stop it from killing the target.”
This is a quasi-example of the problem of so-called AI alignment, in which the AI is trained on the one hand to achieve appropriate goals, but on the other hand is not supposed to pursue them under all conditions. So the AI-alignmet is about ethics and morality, and in this specific case, bizarrely, the ethics of a killing machine.
However, the testing and implementation of ethical/moral alignment models costs effort and time. The call for an AI moratorium therefore certainly makes sense - the only question is how this could be enforced ...
Well, actually we like drones at slashCAM ... for filming...