I thought the correct response should be “Is the thing in fact a giant or a windmill?” Rather than considering which way our maps should be biased, what’s the actual territory?
I do tech support, and often get responses like “I think so,” and I usually respond with “Let’s find out.”
In the “evil giant vs windmill” question, the prior probability of it being an evil giant is vanishingly close to zero, and the prior probability of it being a windmill is pretty much one minus the chance that it’s an evil giant. Spending effort discovering the actual territory when every map ever shows it’s a windmill sounds like a waste of effort.
A missile silo disguised as a windmill? A helicopter in an unfortunate position? An odd and inefficient form of rotating radar antenna? A shuttle in launch position? (if one squints, they might think it’s a broken windmill with the vanes having fallen off or something)
These are all just off the top of my head. Remember, if we’re talking about someone who tends to, when they see a windmill, be unsure whether it’s a windmill or an evil giant, there’s probably a reasonable chance that they tend to get confused by other objects too, right? :)
I thought we were listing anything at least as plausible as the evil giant hypothesis. I have no information as the morality distribution of giants in general so I use maximum entropy and assign ‘evil giant’ and ‘good giant’ equal probability.
Given complexity of value, ‘evil giant’ and ‘good giant’ should not be weighted equally; if we have no specific information about the morality distribution of giants, then as with any optimization process, ‘good’ is a much, much smaller target than ‘evil’ (if we’re including apparently-human-hostile indifference).
Unless we believe them to be evolutionarily close to humans, or to have evolved under some selection pressures similar to those that produced morality, etc., in which we can do a bit better than a complexity prior for moral motivations.
(For more on this, check out my new blog, Overcoming Giants.)
Well, if by giants we mean “things that seem to resemble humans only are particularly big”, then we should expect some sort of shared evolutionary history, so....
Or, possibly, a great big fan! In fact with some (unlikely) designs it would be impossible to tell whether it was a fan or a windmill without knowledge of what is on the other end of the connected power lines.
I’d consider myself puzzled. Unidientified object, is it a threat, a potential asset, some kind of Black Swan? Might need to do something even without positive identification. Will probably need to do something to get a better read on the thing.
I thought the correct response should be “Is the thing in fact a giant or a windmill?” Rather than considering which way our maps should be biased, what’s the actual territory?
I do tech support, and often get responses like “I think so,” and I usually respond with “Let’s find out.”
Giant/windmill differentiation is not a zero-cost operation.
In the “evil giant vs windmill” question, the prior probability of it being an evil giant is vanishingly close to zero, and the prior probability of it being a windmill is pretty much one minus the chance that it’s an evil giant. Spending effort discovering the actual territory when every map ever shows it’s a windmill sounds like a waste of effort.
What about a chunk of probability for the case of where it’s neither giant nor windmill?
Very few things barring the evil giant have the ability to imitate a windmill. I did leave some wiggle room with
because I wished to allow for the chance it may be a bloody great mimic.
A missile silo disguised as a windmill? A helicopter in an unfortunate position? An odd and inefficient form of rotating radar antenna? A shuttle in launch position? (if one squints, they might think it’s a broken windmill with the vanes having fallen off or something)
These are all just off the top of my head. Remember, if we’re talking about someone who tends to, when they see a windmill, be unsure whether it’s a windmill or an evil giant, there’s probably a reasonable chance that they tend to get confused by other objects too, right? :)
You are right! Even I, firmly settled in the fourth camp, was tricked by the false dichotomy of windmill and evil giant.
To be fair, there’s also the possibility that someone disguised a windmill as an evil giant. ;)
A good giant?
Sure, but I wouldn’t give a “good giant” really any more probability than an “evil giant”. Both fall into the “completely negligible” hole. :)
Though, as we all know, if we do find one, the correct action to take is to climb up so that one can stand on its shoulders. :)
I thought we were listing anything at least as plausible as the evil giant hypothesis. I have no information as the morality distribution of giants in general so I use maximum entropy and assign ‘evil giant’ and ‘good giant’ equal probability.
Given complexity of value, ‘evil giant’ and ‘good giant’ should not be weighted equally; if we have no specific information about the morality distribution of giants, then as with any optimization process, ‘good’ is a much, much smaller target than ‘evil’ (if we’re including apparently-human-hostile indifference).
Unless we believe them to be evolutionarily close to humans, or to have evolved under some selection pressures similar to those that produced morality, etc., in which we can do a bit better than a complexity prior for moral motivations.
(For more on this, check out my new blog, Overcoming Giants.)
Well, if by giants we mean “things that seem to resemble humans only are particularly big”, then we should expect some sort of shared evolutionary history, so....
Which can be fun to do with a windmill, also.
Since when do windmills have shoulders? :)
Or, possibly, a great big fan! In fact with some (unlikely) designs it would be impossible to tell whether it was a fan or a windmill without knowledge of what is on the other end of the connected power lines.
Do you consider yourself “objective and wise”?
I’d consider myself puzzled. Unidientified object, is it a threat, a potential asset, some kind of Black Swan? Might need to do something even without positive identification. Will probably need to do something to get a better read on the thing.