On Sat, Jan 5, 2013 at 8:52 AM, Anastasios Tsiolakidis < [email protected]> wrote:
> I doubt this will help but the prediction many of us are interested in is > "multiverse generation", the generation of many alternatives that could > happen next together with their paraphernalia, like their probability and > desirability and semantics. > > Anastasios, I have an idea that I called multiplicity that I think might have some similarities. However, it is not tied to some kludgy notion of probability or logic. If I was working on it and I realized that probability would be useful then I would use probability. If I found that logic was useful I would use logic. It is based on rational methods but I haven't predefined them other than it is to be used in a reasonable manner. Multiplicity is discreet. It just means that I would want the program to be keeping track of the possibilities given the situation that it is in and was in. But it is obvious that this should not be tied to some numerical method because it is inherently making guess about what is happening. It is also obviously not tied to a purely sequential model (like a Markov Process in probability). Finally it is not tied to a model based purely on previous experience. I know that Ben has thought of this stuff as well, but I am not caught up in an unrealized idealization. Multiplicity is based on the multiple possibilities of the future, but it is also based on the multiple possibilities of the past. And, oh yeah, it is based on the multiple possibilities of the present - since that is the problem. How am I pursuing something that is feasibly different than many of you? Knowing that an idealization is unrealized and unrealizable I would mix my method (my conjecture) of multiplicity with other feasible methods. The key to feasibility in AI/AGI is the recognition that ideas work wonderfully in some narrow situations and not so great in others. I hope this will help to clear up this misunderstanding. Now, how about taking my own advice and try 'multiplicity' in real life. I do, but unlike Tim's notion of expectation, I do use it consciously in real life as well as semi-consciously. Once I get to the point in my programming that I am ready to try the idea out I might begin writing some characterizations of my perception of real life to test how this method works when it is made in a form that can be studied. I appreciate that you made the effort to tell me about multiverse generation, because while I can make a guess about what you were talking about, I am not following the idea due to what I believe is its inherent infeasibility. (I presume the basis of the method is feasible in extremely simple applications - I get that). Jim Bromer ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
