I'm not disputing your point vs cryonics, but 0.5 will only rarely be the best possible estimate for the probability of X. It's not possible to think about a statement about which literally nothing is known (in the sense of information potentially available to you). At the very least you either know how you became aware of X or that X suddenly came to your mind without any apparent reason. If you can understand X you will know how complex X is. If you don't you will at least know that and can guess at the complexity based on the information density you expect for such a statement and its length.
Example: If you hear someone whom you don't specifically suspect to have a reason to make it up say that Joachim Korchinsky will marry Abigail Medeiros on August 24 that statement probably should be assigned a probability quite a bit higher than 0.5 even if you don't know anything about the people involved. If you generate the same statement yourself by picking names and a date at random you probably should assign a probability very close to 0.
Basically it comes down to this: Most possible positive statements that carry more than one bit of information are false, but most methods of encountering statements are biased towards true statements.
I wonder what the average probability of truth is for every spoken statement made by the human populace on your average day, for various message lengths. Anybody wanna try some Fermi calculations?
I'm guessing it's rather high, as most statements are trivial observations about sensory data, performative utterances, or first-glance approximations of one's preferences. I would also predict sentence accuracy drops off extremely quickly the more words the sentence has, and especially so the more syllables there are per word in that sentence.
This thread is for the discussion of Less Wrong topics that have not appeared in recent posts. If a discussion gets unwieldy, celebrate by turning it into a top-level post.
July Part 1