I liked the list and I think it touches many fundamental aspects. I have two points which I think deserve more attention and are not explicitly stated here.
(1) As a warning shot is fired, how dependent will we already be upon the AI system which triggered it? If we find ourselves in a situation where stopping the AI system is locally too expensive and disadvantageous, there will be incentives to overlook the warning shots and to downplay the entity of the risk. Apart from explicitly hiding the results (see point 2 below), there will be incentives -- even ... (read more)
I liked the list and I think it touches many fundamental aspects. I have two points which I think deserve more attention and are not explicitly stated here.
(1) As a warning shot is fired, how dependent will we already be upon the AI system which triggered it? If we find ourselves in a situation where stopping the AI system is locally too expensive and disadvantageous, there will be incentives to overlook the warning shots and to downplay the entity of the risk. Apart from explicitly hiding the results (see point 2 below), there will be incentives -- even ... (read more)