4 ms·
Defining "know" something to mean "know the probability of something is exactly 1", then yeah, you can't do that. No amount of Bayesian evidence can raise a pro
by JulianMorrison 13y ago
Defining "know" something to mean "know the probability of something is exactly 1", then yeah, you can't do that. No amount of Bayesian evidence can raise a probability from uncertain to infinitely certain.
So what you need to do, is tear up that useless definition of "know" and replace it with something sensible like "I know X if the evidence for it is very strong."
- jumbled 13y agoThen how do you know that the evidence is very strong? You'd have to have strong evidence that the evidence is strong, ad infinitum. That was essentially Plato's definition of knowledge, "justified true belief." Then a guy named Gettier came around and showed that you can have strong evidence supporting a belief, and still be wrong. Then a guy named Nozick[1] came along and provided his own definition of knowledge, and that brings us up to the early 1980's. There's an entire branch of philosophy dedicated to figuring out exactly what "knowledge" is, Epistemology. The "brain in the vat" problem has be considered in context[2]. [1]http://www.iep.utm.edu/nozick/#H3 http://www.iep.utm.edu/nozick/#H3 [2]http://plato.stanford.edu/entries/epistemology/ http://plato.stanford.edu/entries/epistemology/
- JulianMorrison 13y agohttp://en.wikipedia.org/wiki/Solomonoff's_theory_of_inductive_inference http://en.wikipedia.org/wiki/Solomonoff's_theory_of_inductiv... Crudely, you evolve a simplicity prior over every possible universe-predicting theory according to Bayesian updates on input evidence.