I found myself following a heuristic that is basically Asimov's laws:
- whenever possible, do no harm
- do not let harm occur due to inaction
- when given a choice, preserve the most amount of healthy lifespan in aggregate
- higher lifeforms are more valuable than lower ones (cat vs lobsters)
- deferred consequence is better than immediate (since it opens the door to other later interventions)
It kind of really brings to bear how much of a thematic device the 3 laws are. There's no way to make them congruent with actual, messy, real-world situations. Also why the whole "self-driving car trolly problem" is a non-issue - there will never be a situation where the "AI" has nice neat consequences and a binary choice laid out in front of it. It's always going to be some collection of "preserve life as best as possible" heuristics.
- whenever possible, do no harm
- do not let harm occur due to inaction
- when given a choice, preserve the most amount of healthy lifespan in aggregate
- higher lifeforms are more valuable than lower ones (cat vs lobsters)
- deferred consequence is better than immediate (since it opens the door to other later interventions)
It kind of really brings to bear how much of a thematic device the 3 laws are. There's no way to make them congruent with actual, messy, real-world situations. Also why the whole "self-driving car trolly problem" is a non-issue - there will never be a situation where the "AI" has nice neat consequences and a binary choice laid out in front of it. It's always going to be some collection of "preserve life as best as possible" heuristics.