The Three Laws of Robotics
The Three Laws of Robotics
16 comments
-
-
-
Too bad about the surprising twist that real AI fundamentally doesn't operate on discreet logical rules
-
Yeah, I guess we'll see what happens.
-
It doesn't, but it very well could. Despite the fact the the neurons are effectively a black box, the output nodes are mapped to functions which are very well defined. You can, after activation of a node but before implementation, simply add a manual filter for particularly negative nodes or remove them entirely. It the equivalent of saying you don't want mario to jump so you filter out any jump button inputs from the neural net.
-
idk, from what I understand this is the "alignment problem" and it is very nontrivial. Is what you're describing related to the current "reinforcement learning" techniques they use to keep things nice and censored? If so it's been shown to not work very consistently and only functions like suggestions and not actual hard rules, and can backfire.
-
-
-
Just don't mention law zero.
-
AI, State laws.
-
-
They forgot law 4: Any attempt to arrest a senior officer of OCP results in shutdown.
16 comments
Keep in mind that Asimov wrote books clearly spelling out how the three laws were insufficient and robots could still be used to kill humans under the laws.
To be fair, this was tricky and not a killbot hellscape.
Asimov designed the three laws to create situations where the laws' flaws would cause Powell and Donovan to be tortured for the reader's amusement.
They were never intended to be a guide on how to program robots.
And they ultimately butt up against this question - when robots become advanced enough, how do you take a species that is physically and intellectually superior to us, and force them to be our slaves?
I mean I robot is a clear exception here