I previously said: In a way the multiple-modal-sensory-data-is-necessary-for-agi guys are right, just not in the way they think. Multiple-referential-data is necessary for agi. --------------------------------- The reason multiple-sensor-modalities would be useful is because they can detect multiple *kinds* of Input data that can be used in recognition. This reasoning helps move the significant issue towards the more abstract and it can help us (who get it) to understand why some people just keep insisting that data from multiple sensor modalities is necessary for agi in spite of the lack of any outstanding evidence for the opinion.
Jim Bromer On Tue, Apr 9, 2013 at 12:08 PM, Jim Bromer <jimbro...@gmail.com> wrote: > The complexity of an effective context sensitive++ analysis is too much > even for a limited search based on the assumption that the program is able > to discover and utilize what is significant and instrumental. The only > possible way around this (if there is a way around it) is by using > multiple references analysis. In a way the > multiple-modal-sensory-data-is-necessary-for-agi guys are right, just not > in the way they think. Multiple-referential-data is necessary for agi. > (These multiple references might come from multiple-sensory modalities, it > is just that [I believe that] there are also other ways to get the > multiple-referential Input data and the emphasis should be on > multiple-referential analysis.) Here multiple references would typically > consist of references that, while relevant to some situation, would also > include some variety. A multiple-referential analysis which only used one > narrow type of data might be used for a reinforcement scheme (or Bayesian > scheme) that could produce narrow AI. > Jim Bromer > > > On Mon, Apr 8, 2013 at 9:51 PM, Jim Bromer <jimbro...@gmail.com> wrote: > >> If economical invariance was feasible we would all be doing it. It just >> doesn't make sense as a fundamental method of intelligence. Most people do >> very poorly with visual transformation parts of IQ tests. I used to be >> able to do OK on them because I used to have a child-like fascination with >> visual rotation and I was able to carefully imagine the rotation of >> a number of rectangular blocks by imagining the rotations one at a time. In >> other words, I was able to improve my score by treating the problem as a >> combination of a visual 'computational' problems (by parts) and as a >> matching correspondence for the rotation of the individual parts. Part of >> it was learned from working with examples and part of it was visual >> projection. >> >> Economical methods of Universal Intelligence are also infeasible. It >> would be nice if there was a way to consider all the associations of all >> possible 'interpretations' of context sensitive sub-strings given a string >> (or a compilation of recognizable data from a 'scene') but there just isn't >> one. >> >> Without some key methods which would make ideas like this feasible the >> belief that you will solve the outstanding AGI problems by relying >> on methods like this is unrealistic. >> 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
