I occasionally see people here repeatedly making the same statement, a statement which appears to be unique to them, and rarely giving any justification for it. Examples of such statements are “Bayes’ law is not the fundamental method of reasoning; analogy is” and “timeless decision is the way to go”. (These statements may have been originally articulated more precisely than I just articulated them.)
I’m at risk of having such a statement myself, so here, I will make this statement for hopefully the last time, and justify it.
It’s often said around here that Bayesian priors and Solomonoff induction and such things describe the laws of physics of the universe. The simpler the description, the more likely that laws-of-physics is. This is more or less true, but it is not the truth that we want to be saying. What we’re trying to describe is our observations. If I had a theory stating that every computable event happens, sure, that explains all phenomena, but in order for it to describe our observations, you need to add a string specifying which of these computable events are the ones we observe, which makes this theory completely useless.
In theory, this provides a solution to anthropic reasoning: simply figure out which paths through the universe are the simplest, and assign those the highest probability. Again, in theory, this provides a solution to quantum suicide. But please don’t ask me what these solutions are.
Does anyone understand the last two paragraphs of the comment that I’m responding to? I’m having trouble figuring out whether Warrigal has a real insight that I’m failing to grasp, or if he is just confused.
I occasionally see people here repeatedly making the same statement, a statement which appears to be unique to them, and rarely giving any justification for it. Examples of such statements are “Bayes’ law is not the fundamental method of reasoning; analogy is” and “timeless decision is the way to go”. (These statements may have been originally articulated more precisely than I just articulated them.)
I’m at risk of having such a statement myself, so here, I will make this statement for hopefully the last time, and justify it.
It’s often said around here that Bayesian priors and Solomonoff induction and such things describe the laws of physics of the universe. The simpler the description, the more likely that laws-of-physics is. This is more or less true, but it is not the truth that we want to be saying. What we’re trying to describe is our observations. If I had a theory stating that every computable event happens, sure, that explains all phenomena, but in order for it to describe our observations, you need to add a string specifying which of these computable events are the ones we observe, which makes this theory completely useless.
In theory, this provides a solution to anthropic reasoning: simply figure out which paths through the universe are the simplest, and assign those the highest probability. Again, in theory, this provides a solution to quantum suicide. But please don’t ask me what these solutions are.
Does anyone understand the last two paragraphs of the comment that I’m responding to? I’m having trouble figuring out whether Warrigal has a real insight that I’m failing to grasp, or if he is just confused.