Depending on what structure of belief we have, we might make opposite conclusions from the same data.
For example, if we believe strongly that the number of heads and tails from a coin flip will average out to 50/50 in the short run, then if we see "h h h" we will predict "t" next. This is common in humans, which suggests that although it's a bad predictor for coin flips, it might be good for other things humans deal with. Lifespan is like this: the longer someone is alive, the sooner we expect them to die. If someone is 120, we expect them to die any day! On the other hand, if we are suspicious that the coin might be weighted towards one side or the other (or even be a trick coin, with the same image on both sides), then we will predict "h" next instead. The more heads we see, the stronger this prediction. Philosophy of science tends to think the heuristic "expect future observations will be like past observations" is more general-purpose than "expect future observations to differ". But we have to formalize this somehow to get AIs to do it! :) On Tue, Jul 24, 2012 at 1:02 PM, Piaget Modeler <[email protected]>wrote: > > It's not the endowment (biological or algorithmic) but the interaction > with the environment > that's most important: especially for developmental cognitive systems. > > > > > Date: Tue, 24 Jul 2012 12:11:17 -0400 > > Subject: [agi] A Program Is Not Completely Predefined Until It Has Been > Run For The Last Time > > From: [email protected] > > To: [email protected] > > > > > When someone writes a word processing program he cannot know, as he > > writes it, everything that the users will write as they use it. This > > proves that a program, even though it may be preplanned, is not > > limited to precisely producing only something that is totally > > preplanned. I would call the parts of the word processor algorithms > > even though the term is not exactly the same as the mathematical > > definition that is given in Wikipedia. My guess is that the dominant > > mathematical definition of the word “algorithm” proffered in Wikipedia > > was actually formulated in the twentieth century. I certainly do not > > use the word in only this way, although I have to admit that I > > sometimes have felt that my use of the term was somewhat awkward. > > > > Now, a programmer could write his word processor program so that it > > did not take every single letter that a user typed in. In this case, > > neither the programmer, nor the user could predict beforehand what > > would be appended to the text. In fact, the programmer could design > > the program so that no one could predict which letters would be > > appended regardless of their knowledge of the program and of computer > > science. This shows that a computer program is not necessarily fully > > predefined by the program itself but may be dependent on how it is > > used. If a program can react to Input then it can react in creative > > and unpredictable ways. (The fascination for “prediction” that some > > AGIers worship seems slightly ironic given that some unpredictability > > seems like a requisite for independent thinking.) > > Jim Bromer > > > > > > ------------------------------------------- > > AGI > > Archives: https://www.listbox.com/member/archive/303/=now > > RSS Feed: > https://www.listbox.com/member/archive/rss/303/19999924-5cfde295 > > Modify Your Subscription: https://www.listbox.com/member/?&; > > > Powered by Listbox: http://www.listbox.com > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/7190161-766c6f07> | > Modify<https://www.listbox.com/member/?&;>Your Subscription > <http://www.listbox.com> > -- Abram Demski http://lo-tho.blogspot.com/ ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
