Ground truth is the objective nature of the system that you are observing. If you are in a dream (like the characters in Twin Peaks), there won't be a ground truth that manifests as observable regularity; a dream universe is not mechanical, and so it does not yield to science.
-
-
an aside: I think we detoured from computationalism per se when I said No to “The theory that there are others with minds similar to mine has more explanatory and predictive power than its negation?”. That No is driven by …
1 reply 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
standardish criticisms of empiricism, namely sense-data sceptism and the problem of induction. The detour interests me, but please don't feel obliged to share my interest. I think what I'm trying to do on this thread then is update/restate those arguments in your vocabulary …
1 reply 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
1) Problem of Induction becames “There is no argument that regularities observed today describe the universe I will observe tomorrow except for the (circular) argument that the regularities I observed yesterday worked today.” …
2 replies 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
2) Sense-data scepticism maps to something like “there is no way to assign an initial p(H) to the hypothesis H=‘My observations are in fact observations of a real univsere’” –It seems relevant to add memory-scepticism: …
1 reply 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
3) Memory-scepticism “I don't have any evidence that I observed any regularities yesterday. Except my memory, and the only evidence I have that my memory is reliable is the memory of it being reliable yesterday.“ I don't see rejecting scepticism is anything except a step of faith
2 replies 0 retweets 1 like -
I don't think that you should reject skepticism but fully embrace it: reject the idea that you were ever entitled to belief without priors. Assign probabilities and confidence parameters to all of your statements, make them conditional, and you are good to go.
1 reply 0 retweets 0 likes -
Taking my sense-data scepticism as example then, do you mean something like: H=“My observations are in fact observations of a real universe” e=“I observe things” p(H)=The prior probability p(e)=Nearly 1 p(e|H)=Also nearly 1 p(H|e)=tobecalculated & plug into Bayes equation?
1 reply 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
I realise that where I was going with this is already written up at the final para of https://plato.stanford.edu/entries/induction-problem/#BaySub … : “The simple, and obvious, criticism of the Bayesian method is that the prior (before knowledge of any evidence) probabilities … are arbitrary. The Bayesian response…
1 reply 0 retweets 1 like -
Replying to @chrisfcarroll @Plinz and
…Bayesian response is that the Bayesian method of updating probabilities with successive outcomes progressively diminishes the effect of the initial priors. This updating uses the posterior probabilities from the first draw as the “prior” probabilities for the second draw…
2 replies 0 retweets 1 like
The Bayesian prior problem is different IMHO: are you guaranteed to get to the global optimum of models from an arbitrary prior? The general answer is no, but in practice self-consistent local optima are too rare to lock you in for more than a couple millennia or so.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.