This thread by @fchollet, applied to Bayesianism, is the basic counterargument to the last ten years of “AGI as existential threat”.https://twitter.com/fchollet/status/1010988618993655808 …
-
-
Either an equivocation on the word prediction, or atemporal (i.e., inconsistent with the assumption of updating based on new information). Lots of fun to propose to students, however!—a nice gauge of how religious they are about ML.
-
I get that it is a theoretical game, but how is it inconsistent with updating based on new information? Do you just mean it’s a single play game?
-
Generically: there's an arrow of time in Bayes. You can learn (i.e., reduce your entropy over theories), but you can't consistently "forget". Newcomb's paradox doesn't have a consistent arrow of time.
-
But what about inconsistencies with updating information if this is single instance game? Are you talking about the predictor being thermodynamically limited now?
End of conversation
New conversation -
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.