Interesting, Jim. I was just describing to my brother earlier an observation I made about recipes, that different styles of food have different "grammars" for ingredients. Certain combinations of ingredients, textures, and flavors make sense for the foods of one culture but not in another's. And then there are certain shared patterns that result not from syntactical constraints, but from pragmatic utility.
On Fri, Dec 27, 2013 at 10:16 PM, Jim Bromer <jimbro...@gmail.com> wrote: > The view that an insight is a system based on observations and a lot > of creative explanations is a little problematic. > > But, just because a part of an insight is imaginative does not mean > that it is not a rational bridge in the insight (of course). > > So when we can come up with a creative explanation to fill in a gap of > an insight we would like to make the explanation utilize some > observations of effects in a way as to provide the explanation with > more structure. So it is not just an observation correlation but an > rational explanation that correlates with some effective observation > points. Observation points are often used in definitions and the > rational explanations needed to fill in the gaps are often based > explanations for similar kinds of things. > > For example: > A programming language is based mostly on using a context-free > grammar. (Some of the observation points here are the programmer's > recall of first realizing that he is using syntactic grammars to write > programs.) > > So a computer program that is designed to learn can be said to be > using a syntactic grammar. Even if an AI program that is designed to > learn a natural language grammar through trial and error does not > start with a base of a natural language grammar, it still cannot be > said to use no grammar at all. It is using a computational grammar of > some sort even if the programmer does not consciously think of it in > that way. (Here, for example, the programmer might recall his > recognition that computer programs are inputting, rearranging and > outputting strings of individual values that are similar to or are > characters in a syntactic string. > > A computer could learn a very simple context-free grammar through > trial and error alone. (We have all seen programs that were able to > 'learn' something incrementally and most of us are familiar with > reinforcement methods so this does not require a lot of fantasizing to > arrive at the conclusion that this may be feasible. And when you > realize that what I am talking about is that simple context free > grammars only have to be treated as worded input 'commands' -that are > followed at least some of the time- then this looks very feasible. In > fact, it seems so feasible that almost any experienced programmer who > has some sense of what I am talking about could try it.) > > Finally, the acquired (not pre-programmed) simple context-free > grammars (using words) could be used to teach the AI program some > simple natural language structure that use context-sensitive and other > natural language grammars. (This is the conjecture which seems > feasible if you accept the other steps. But this step absolutely > requires experimentation to confirm. The skeptics try to point out > that learning to use natural language requires some fundamental > knowledge of what the words represent but that is what can be taught > when the program is learning to react to simple worded commands and > later higher level explanations.) (There were few observation points > in this last part but it is really the rearrangement of familiar > definitions that are serving as rational bridges over the spans that > the incredulous skeptics of the conjecture object to. So even though > no one has observed an AI/AGI program that can do this, it really does > make sense. If there is a problem then, it probably must be due to > the complexity of the knowledge that would be required to make this an > effective AGI paradigm. > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/23050605-2da819ff > 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
