sokratis.dk said: > The technical issues surrounding statistics, symbols, production systems and > their intersection are under constant investigation and innovation therefore > yes, I expect more and more AGI goodies to come out of them. I think I have > already stated about 20 times "in print" that most no one has attempted a > really ambitious general synthesis... If you want to leave it at that, ok. If you can use some method effectively in an AGI program then you should. The problem with your reason is that the application of statistics to AGI is not the same as the scientific application of statistics in scientific studies. This can produce serious problems. The same kind of thing occurred when logic was applied to AI. A fundamental form of logic that was used in AI was reasoning from general to specific which can work in some cases but not in all cases. The value was that the generalization represented a statement that could cover a vast number of cases and it acted as a compression. If it was known that a concept object was a member of a collection then a generalization that had been assigned to the collection could be applied to the object. The problem is that in our thinking the characteristics that generally apply to the members of a collection do not necessarily apply to every member of the collection. I have a number of criticisms about statistics which I started to discuss but the primary criticism is that I do not consider AGI-statistics to be a good candidate as a fundamental AGI method. I feel that using concepts in relational structures where, for example, a concept may be considered to be typed and to play a kind of role in an idea, and using reason-based reasoning are much better candidates for use as a fundamental objects for an AGI program. However, this is not a simple solution since a particular concept can play different roles in different kinds of ideas. This is an example of what I meant when I explained that knowledge is relativistic. Jim Bromer > Subject: Re: [agi] A Very Simple AGI Project > From: [email protected] > Date: Thu, 1 Aug 2013 13:41:26 +0200 > To: [email protected] > > As I said, and was somewhat dubiously annotated by Matt, even if you don't > have statistics you'd be wise to pick it up rather quickly. I don't very much > get the discussion over belief or not, if it represents a belief so what, did > anyone say beliefs are incompatible with artificial intelligence? If it is > not a belief, did someone claim beliefs are necessary? > > > The technical issues surrounding statistics, symbols, production systems and > their intersection are under constant investigation and innovation therefore > yes, I expect more and more AGI goodies to come out of them. I think I have > already stated about 20 times "in print" that most no one has attempted a > really ambitious general synthesis (assuming complete architectures like > OpenCog don't count as mere synthesis). It's the kind of thing you'd rather > do on big hardware anyway, even though I am working on a BOINC like concept. > > AT > > On 31.07.2013, at 21:09, Jim Bromer <[email protected]> wrote: > > > Me: I may be misinterpreting what you said, but it looks like you said that > > since you can see distributions everywhere they should be used as the basis > > for AGI. > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/24379807-f5817f28 > 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 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
