Pentagon AI more ethical than adversaries’ because of ‘Judeo-Christian society,’ USAF general says::The path to ethical AI is a “very important discussion” being held at DOD’s “very highest levels,” says service’s programs chief.
Pentagon AI more ethical than adversaries’ because of ‘Judeo-Christian society,’ USAF general says::The path to ethical AI is a “very important discussion” being held at DOD’s “very highest levels,” says service’s programs chief.
This is what AI would like them to believe
A lot researchers a long time ago used to say that in order to judge AI, they would have to get it to pass a Turing Test in order for us to figure out if it just as intelligent or more intelligent than us.
I always enjoyed the idea that AI will quickly fly right over our heads and our ability to identify it that it will purposely make itself appear dumb or dumber than us while it figures out how to deal with us.
AI passes the Turing test long before it becomes AI. Are you thinking of the number of tests we have for AGI? (e. g. build flat-packed furniture from the instructions or clean and load a coffee machine to make coffee)
Malicious AI typically comes down to bad programming which includes deceptive AI, which is easy to do, given LLMs and the obscurity of natural language. This is also why AI is not really ready for public use (except as a toy or creative tool) since it’s very easy to give AI instructions that it will interpret to yield poor results (like AI attack drones killing their commander in simulation).