> Perhaps we have yet to find the ‘periodic table of moral elements’ because we are trying to write an algorithm to compute a function which is uncomputable?
Its not uncomputable. Its just that humans are bad at philosophy. If something feels vauge and hard to understand, its not uncomputable, nor is it down to the heisingberg uncertainty principle.
If it doesn't involve infinite amounts of information, or telling if arbitrary turing machines halt, its not uncomputable.
Good != long term survival. Its possible to live for aeons and hate every minute of it. Also, whose long term survival? Yours? All humans? All mammals? All humans + some sufficiently smart AI's?
> If it doesn't involve infinite amounts of information
I think all kinds of moral decisions do, though.
For a simple example, is it ever immoral to run a computer program? If the morality of running a computer program depends on what the program does, then yes, it’s uncomputable. Just replace “will it halt” with “will it kill everyone.”
> A particular human mind deciding whether or not to run the computation is computable.
Sure, just as deciding whether _any single_ computer program will halt is computable. The problem of uncomputability comes into play when attempting to developing a general rule that will work in all cases.
Sure, and a general simulation of a human mind is computable. A simulation of the universe so far is computable. Basically the only "moral problem" that is uncomputable is when you take a magic source of uncomputability, and set it to blow up and hurt someone, or not, in an uncomputable way.
"Morality is random, because trolly problem, left track 1 person, right track randomly 0 or 2 people. Whether or not its moral to pull the lever is random".
Do you think the existence of uncomputable numbers has no practical implications then?
Why can’t someone sell a computer program that analyzes software and says whether or not it will crash? Why do apple and google release code with bugs? Shouldn’t they just write code that will prove their systems never crash?
Formal verification is done on some code, especially small, safety critical code. Its slow, and takes a lot of time and effort. So why is buggy code released, on some level its economics. Provably bug free code is not impossible, but it is expensive in programmer time or computer time.
An analysis tool that answers "will it crash" always, for hypothetical computer processes, is impossible. One that works for all computers that fit in the observable universe is possible (but it itself would need more memory)
Alternate reality hyperbeings with a Grahams number flop computer could answer any question about the behavior of programs in this universe.
(Well its possible that there is some trick to solving the halting problem with a black hole or something, but in that case, that still isn't of practical relevance right now, with our computers. )
> Perhaps we have yet to find the ‘periodic table of moral elements’ because we are trying to write an algorithm to compute a function which is uncomputable?
Its not uncomputable. Its just that humans are bad at philosophy. If something feels vauge and hard to understand, its not uncomputable, nor is it down to the heisingberg uncertainty principle.
If it doesn't involve infinite amounts of information, or telling if arbitrary turing machines halt, its not uncomputable.
Good != long term survival. Its possible to live for aeons and hate every minute of it. Also, whose long term survival? Yours? All humans? All mammals? All humans + some sufficiently smart AI's?
> If it doesn't involve infinite amounts of information
I think all kinds of moral decisions do, though.
For a simple example, is it ever immoral to run a computer program? If the morality of running a computer program depends on what the program does, then yes, it’s uncomputable. Just replace “will it halt” with “will it kill everyone.”
Will it kill anyone in the finite amount of time before the heat death of the universe? is computable.
A particular human mind deciding whether or not to run the computation is computable.
> A particular human mind deciding whether or not to run the computation is computable.
Sure, just as deciding whether _any single_ computer program will halt is computable. The problem of uncomputability comes into play when attempting to developing a general rule that will work in all cases.
Sure, and a general simulation of a human mind is computable. A simulation of the universe so far is computable. Basically the only "moral problem" that is uncomputable is when you take a magic source of uncomputability, and set it to blow up and hurt someone, or not, in an uncomputable way.
"Morality is random, because trolly problem, left track 1 person, right track randomly 0 or 2 people. Whether or not its moral to pull the lever is random".
Do you think the existence of uncomputable numbers has no practical implications then?
Why can’t someone sell a computer program that analyzes software and says whether or not it will crash? Why do apple and google release code with bugs? Shouldn’t they just write code that will prove their systems never crash?
Formal verification is done on some code, especially small, safety critical code. Its slow, and takes a lot of time and effort. So why is buggy code released, on some level its economics. Provably bug free code is not impossible, but it is expensive in programmer time or computer time.
An analysis tool that answers "will it crash" always, for hypothetical computer processes, is impossible. One that works for all computers that fit in the observable universe is possible (but it itself would need more memory)
Alternate reality hyperbeings with a Grahams number flop computer could answer any question about the behavior of programs in this universe.
(Well its possible that there is some trick to solving the halting problem with a black hole or something, but in that case, that still isn't of practical relevance right now, with our computers. )