AI: Hidden Layers

Allow me to quickly describe a troubling issue with artificial intelligence in genuinely simple terms.

As far as I can understand it, which should be sufficient for the purposes of this article, the core mechanism of how artificial intelligence works is that: 1) you give the program some data; 2) you tell the program what objective you want it to achieve; and then 3) you wait for it to accomplish that objective WITHOUT TELLING ANYONE HOW IT DID SO. This invisible process is called "Hidden Layers."

This is like, if you gave a super robot a toolbox and told it to get you a million dollars as quickly as possible, but not tell you how it did so. Obviously it will just go kill or steal "without your knowledge" wink-wink. Or if you ask the robot to prove you are justified in cheating on your wife. It has to do so, in some way you needn't bother understanding, since math allegedly just proved that you're allowed to do anything you want.

To even conceive of designing a technology in this way should already raise some red flags. The programmers could claim that they've added rules that prevent AI from breaking the laws... but if it did, nobody would find out. What's to stop hackers from removing such rules? Or to stop the robot from inventing a way to find loopholes to still perform psychopathic activities to achieve whatever trivial objective it's been ordered to achieve?

Have we become so desperate for "someone" to approve of our bullshit that we've actually had to invent a technology that can "prove" anything we want to hear? How many Greys will they have to fellate in the metaverse before they realize that the Laws of Nature might be something worth looking into?