>If what you mean by SIA is something more along the lines of “Constantly update on all computable hypotheses ranked by Kolmogorov Complexity”, then our definitions have desynced.
No, that’s what I mean by Bayesianism—SIA is literally just one form of interpreting the universal prior. SSA is a different way of interpreting that prior.
>Also, remember: you need to select your priors based on inferences in real life. You’re a neural network that developed from scatted particles- your priors need to have actually entered into your brain at some point.
The bootstrap problem doesn’t mean you apply your priors as an inference. I explained which prior I selected. Yes, if I had never learned about Bayes or Solomonoff or Occam I wouldn’t be using those priors, but that seems irrelevant here.
>SIA has you reason as if you were randomly selected from the set of all possible observers.
Yes, this is literally describing a prior—you have a certain, equal, prior probability of “being” any member of that set (up to weighting and other complications).
>If you think you’re randomly drawn from the set of all possible observers, you can draw conclusions about what the set of all possible observers looks like
As I’ve repeatedly stated, this is a prior. The set of possible observers is fully specified by Solomonoff induction. This is how you reason regardless of if you send off probes or not. It’s still unclear what you think is impermissible in a prior—do you really think one can’t have a prior over what the set of possible observers looks like? If so, you’ll have some questions about the future end up unanswerable, which seems problematic. If you specify your model I can construct a scenario that’s paradoxical for you or dutchbookable if you indeed reject Bayes as I think you’re doing.
Once you confirm that my fully specified model captures what you’re looking for, I’ll go through the math and show how one applies SIA in detail, in my terms.
>If what you mean by SIA is something more along the lines of “Constantly update on all computable hypotheses ranked by Kolmogorov Complexity”, then our definitions have desynced.
No, that’s what I mean by Bayesianism—SIA is literally just one form of interpreting the universal prior. SSA is a different way of interpreting that prior.
>Also, remember: you need to select your priors based on inferences in real life. You’re a neural network that developed from scatted particles- your priors need to have actually entered into your brain at some point.
The bootstrap problem doesn’t mean you apply your priors as an inference. I explained which prior I selected. Yes, if I had never learned about Bayes or Solomonoff or Occam I wouldn’t be using those priors, but that seems irrelevant here.
>SIA has you reason as if you were randomly selected from the set of all possible observers.
Yes, this is literally describing a prior—you have a certain, equal, prior probability of “being” any member of that set (up to weighting and other complications).
>If you think you’re randomly drawn from the set of all possible observers, you can draw conclusions about what the set of all possible observers looks like
As I’ve repeatedly stated, this is a prior. The set of possible observers is fully specified by Solomonoff induction. This is how you reason regardless of if you send off probes or not. It’s still unclear what you think is impermissible in a prior—do you really think one can’t have a prior over what the set of possible observers looks like? If so, you’ll have some questions about the future end up unanswerable, which seems problematic. If you specify your model I can construct a scenario that’s paradoxical for you or dutchbookable if you indeed reject Bayes as I think you’re doing.
Once you confirm that my fully specified model captures what you’re looking for, I’ll go through the math and show how one applies SIA in detail, in my terms.