heres a rather amateurish attempt;
god = argmin_h L(h), the hypothesis with minimum description cost over the
set of all possible hypotheses.
L is the description length of h -- the number of bits required to explain different
observations with hypothesis h.
L(h) = sum_o {l_h(o)}, where o ranges of the set of observations you want to explain, i.e, each observation o has some description cost l_h(o) to be "adequately" explained by the hypothesis h.
Now, the hypothesis h which requires the smallest effort to explain the observations is 'god'
does this seem a reasonably meaningful description of god?
Where I see it failing:
1. modeling appropriate cost functions
2. handling observations that may contradict each other
{modeling contradictions would incur infinite cost, implying that either god doesnt exist or everything is god}
p.s. its a bloody pain to typeset math!
Subscribe to:
Post Comments (Atom)
5 comments:
I suppose when you use the word "cost", you mean to assign a numerical value to the faith one places - i.e a valuation of the faith over and above what can be rationally explained.
Right?
In that case, your definition of God would include valuation of the faith in assuming that P!=NP etc. Any conjecture/hypothesis will be part of God! I had not thought of God this way. Interesting!
To add some thoughts on your theory: Every individual has a set of observations that he will use to define his own God. A religion usually strives to define this set for all its members - so that all of them see God the way the religion sees it.
alternatively you can view it as each religion/philosphical branch has its own cost model. its easier to set a common ground set of observations and vary the costs associated with the observations, within a rational framework. allowing different people to define different sets of observations just allows for one more dimension of irrationality!
btw, my god is only a variable and hence not capitalized to imply any semantic associations.
Can u explain in English pls ?
point5:
assuming you are serious and not joking, heres an attempt at rephrasing it without symbols.
god is an attempt to answer questions around us. from a rational perspective, answers which are more convincing are better ,i.e, answers which take much less description to adequately explain the observations are desirable.
trouble is there are too many questions and observations, and we need to choose the best available god (i.e hypothesis).
my definition is that god (i.e., hypothesis) is the one that can most satisfactorily explain the set of questions you have. the way it finds this optimal hypothesis is to solve an optimization using costs for explaining each observation using a specific hypothesis.
much better...now u r speaking in laymans term
Post a Comment