> > Tinter > Patterns have RULES, patchworks have NO RULES. >
> PM > Is it the case that this distinction between pattern and patchwork > only exists in Mike T.'s mind...? ... Some may see pattern and patchwork > as the same, some may see patchwork as a subset of pattern, or vice versa. > Without coming down on either side of this debate, here are some observations: Tintner's distinction is reminiscent of the dual process theory's account of systems ( http://en.wikipedia.org/wiki/Dual_process_theory#Systems ), namely recognition, perception, orientation versus rule following, comparisons, weighing of options. This then leads to fast, unconscious reasoning and intuition versus slow, conscious reasoning and critical examination. A case in point is chicken sexing ( http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.278.7238 ), where this abstract describes the task: *Sexing day-old chicks: A case study and expert systems analysis of a > difficult perceptual-learning task.* > The sexing of day-old chicks has been regarded as an extraordinarily > difficult perceptual task requiring years of extensive practice for its > mastery. Experts can sex chicks at over 98 % accuracy at a rate of 1,000 > chicks per hour spending less than a half second viewing the cloacal > region. Naive subjects were shown 18 pictures of cloacal regions of male > and female chicks (in random appearing arrangement) and asked to judge the > sex of each chick. The pictures included a number of rare and difficult > configurations. The subjects were then instructed as to the location of a > critical cloacal structure for which a simple contrast in shape (convex vs. > concave or flat) could serve as an indicant of sex. When the subjects > judged the pictures again (in a different order), accuracy increased from > slightly above chance to a level comparable to that achieved by a sample of > experts. The correlation (over items) between the naive subjects and the > experts before instruction was.21; after instruction,.82. The instructions > were based on an interview and observation of an expert who had spent 50 > years sexing 55 million chicks. Much of the reported difficulty in > developing perceptual expertise in this task may stem from the need to > classify extremely rare configurations in which the convexity of the > structure is not apparent. The rate of learning of these instances could be > greatly increased through the use of simple instructions that specified the > location of diagnostic contour contrasts. Continuing this line of reasoning, PM's subset argument may be justifiable on the basis of a shift from one system to the other: conscious competence (with rules) to unconscious competence (no rules). Considering to natural world perceptual learning in general ( http://en.wikipedia.org/wiki/Perceptual_learning#In_the_natural_world ), one can revisit a previous Tinter patchwork example of stone masonry (google image search link: https://www.google.com/search?tbs=simg:CAESUQnbzBKMFPx2fRo9CxCwjKcIGjQKMggBEgxUzgeLBs0HIvEF8wUaIKX6pvVsiGkF5ZjW-maIzVq_1eMmVVgarosT6psK7NvkTDCEVtwqky_1ukQw&q=textured+wall+stone&tbm=isch&sa=X&ei=jNiGUuK8B-OO2gWNsYD4Aw&ved=0CDQQsw4&biw=1615&bih=937 ). As with chicken sexing expertise, a rule can possibly prune the search space. However, it still remains, at worst, unsolvable by algorithm and at best, an NP-Hard optimization problem. The question is posed as to whether one can abstract rules/patterns from such configurations of stones, thus moving from the system where rules do not exist, to the system where they do. One of Kurzeil's first accomplishments was a pattern-recognition software program that analyzed the works of classical composers, and then synthesized its own songs in similar styles ( http://en.wikipedia.org/wiki/Ray_Kurzweil#cite_ref-16 ). This seems analogous with the Tinter's magazine cover example, but in a different modality. Does the infinite variability of audio/music sequences that can be generated from such a program produce a patchwork, albeit generated from a pattern-recognition program? One can capture the raw input data and the observed output and then data mine with deep learning algorithms, be it a pile of stones that get composed into a wall or the sounds produced by a piano that are composed into sonata. Within the context of this discussion, it remains to be resolved as to what the output of such data mining produces and what to call it. >From the definitions and features provided, perhaps the argument can be made that patterns are capable of producing patchworks. It is not clear from the history of this discussion that a pattern can be abstracted from a collection of patchworks. However, it is reasonable to expect that exemplars can be selected from among a class of patchworks and that a prototype can be synthesized from a combination of features from different exemplars. If one can agree that probabilistic regularities can be applied to a prototype to produce patchworks indistinguishable from the sample population (as judged by an expert), then perhaps the rules or no rules issue is moot. ------------------------------------------- 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
