Re: [agi] Artificial [Humor ] vs Real Approaches to Information
Jiri and Matt et al, I'm getting v. confident about the approach I've just barely begun to outline. Let's call it realistics - the title for a new, foundational branch of metacognition, that will oversee all forms of information, incl. esp. language, logic, and maths, and also all image forms, and the whole sphere of semiotics. The basic premise: to understand a piece of information and its information objects, (eg words) , is to realise (or know) how they refer to real objects in the real world, (and, ideally, and often necessarily, to be able to point to and engage with those real objects). - this includes understanding/realising when they are unreal - when they do NOT refer directly to real objects, but for example to sur-real or metaphorical or abstract or non-existent objects Realistics recognizes that understanding involves, you could say, object-ivity. Complementarily, to 'disunderstand is to fail to see how information objects refer to real objects. to be confused is not only to fail to see, but to be unsure *which* of the information objects in a piece of information do not refer to real objects (it's all a bit of a blur) Bear in mind that human information-processing involves an ENORMOUS amount of disunderstanding and confusion. And a *major point* of this approach (to be explained on another occasion) is precisely that a great deal of the time people do not understand/realise *why* they do not understand/ are confused - *why* they have such difficulty understanding genetics, atomic physics, philosophy, logic, maths, ethics, neuroscience etc. etc - just about every subject in the curriculum, academic or social - because, like virtual AGI-ers they fall into the trap of FAILING to refer the information to real objects. They do not try to realise what on earth is being talked about. And they even end up concluding (completely wrongly) that there is something wrong with their brain and its information-processing capacity, ending up with a totally unecessary inferiority complex. (There will probably be v. few here, even at this exalted level of intelligence, who are not so affected). (Realistics should enormously improve human understanding, and holds out the promise that no one will ever fail to understand any information/subject ever again for want of anything other than time and effort). Now there is a LOT more to expand here [later]. But for now it immediately raises the obvious, and inevitable object-ion to any contradictory, unreal /artificial approach to information and esp language processing/NLP such as you and many other AGIers are outlining. How will you understand, and recognize when information objects/ e.g language/words are unreal ? e.g. Turn yourself inside out. Turn that block of wood inside out. Turn around in a straight line. What's inside is not more beautiful than what's on the outside Drill down into Steve's logic. Cars can hover just above the ground The car flew into the wall. The wall flew away. Bush wants to liberalise sexual mores. Truth and beauty are incompatible. [all such statements obviously real/unreal/untrue/metaphorical in different and sometimes multiple simultaneous ways] You might also ask yourself how you will, if your approach extends beyond language, know that any image or photo is unreal. IOW how is any unreal approach to information processing (contradictory to mine) different from a putative logic that does *not* recognize truth or a maths that does *not* recognize equality/equations? Mike, The plane flew over the hill The play is over Using a formal language can help to avoid many of these issues. But then the program must be able to tell what is in what or outside, what is behind/over etc. The communication module in my experimental AGI design includes several specialized editors, one of which is a Space Editor which allows to use simple objects in a small nD sample-space to define the meaning of terms like in, outside, above, under etc. The goal is to define the meaning as simply as possible and the knowledge can then be used in more complex scenes generated for problem solving purposes. Other editors: Script Editor - for writing stories the system learns from. Action Concept Editor - for learning about actions/verbs related roles/phases/changes. Category Editor - for general categorization/grouping concepts. Formula Editor - math stuff. Interface Mapper - for teaching how to use tools (e.g. external software) ... Some of those editors (probably including the Space Editor) will be available only to privileged users. It's all RBAC-based. Only lightweight 3D imagination - for performance reasons (our brains cheat too), and no embodiment.. BTW I still have a lot to code before making the system publicly accessible. To understand is .. in principle, ..to be able to go into the real world and point to the real objects/actions being referred to.. Not from my perspective.
Re: [agi] Artificial [Humor ] vs Real Approaches to Information
On Friday 12 September 2008, Mike Tintner wrote: to understand a piece of information and its information objects, (eg words) , is to realise (or know) how they refer to real objects in the real world, (and, ideally, and often necessarily, to be able to point to and engage with those real objects). This is usually called sourcing and citations, and so on. It's not enough to have a citation though, it's not enough to just have a symbolic representation of some part of the world beyond you within your system, you always have to be able to functionally and competently use those references, citations, or links in some useful manner, otherwise you're not grounded and you're off in la-la land. Computers have offered us the chance to encapsulate and manage all of these citations (and so on) but in many cases they are citations that are limited and crude. Look at the difference between these two citations: Tseng, A. A., Notargiacomo A. Chen T. P. Nanofabrication by scanning probe microscope lithography: A review. J. Vac. Sci. Tech. B 23, 877– 894 (2005). Compared to: http://heybryan.org/graphene.html Both would seem cryptic to any outsider to scientific literature or to the web. The first one is generally variablized across the literature, making OCR very difficult, and making it generally a challenge to always fetch the citations and refs in papers for researchers. Take a look at my attempts at OCR of bibliographies: http://heybryan.org/projects/autoscholar/ Not good is an accurate summarization. With the HTTP string, it's not any better at all, *except* the fact that DNS servers are widely implemented, here's how to implement one, here's how the DNS root servers for the internet work, here's why you can (usually) type in any URL on the planet and get to the same site (unless you're on some other NIC of course - but this is very rare). There's a social context surprisingly involved for DNS .. which I guess is what you consider to be the realistics that everyone overlooks when they just assign symbols to many different things; for instance, I bet you don't know what DNS is, but you know what a dictionary is, even though they refer to more or less the same functional things (uh, sort of). Anyway, it's context that matters when it comes to groundtruthing citations and traces in information ecologies, and not so much the symbolic manipulation thereof. It's the overall groundtruthed process, the instantiated exploding von Neumann probe phylum that will ultimately (not) grey goo you. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Artificial [Humor ] vs Real Approaches to Information
Mike, How will you understand, and recognize when information objects/ e.g language/words are unreal ? e.g. Turn yourself inside out. ... unreal/untrue/metaphorical in different and sometimes multiple simultaneous ways It's like teaching a baby. You don't want to use confusing language/metaphors.. I expect my users to understand the GIGO effect. But GINA (=my AGI experiment) does have some features for dealing with unreal / confusing concepts. As I mentioned before, it learns from stories (written in a formal language). Each story can be marked Real, Unreal, or Abstract. The Real means real world, the Unreal means fairy tale kind of stuff (animals talking etc), and the Abstract covers things like math and other very formal worlds (e.g. chess rules etc). When a user submits a problem-to-solve, he/she can also specify if the scope of the solution search should include the Unreal domain. Another relevant feature is support of phrase concepts. It allows to teach the system about the impact of saying something particular in particular scenarios (e.g. Good night, WTF, I love you, H or possibly your Turn yourself inside out). The description of what it literally means is optional (unlike the impact descriptions). There are also some automated evaluation procedures applied to new knowledge before it's approved as a knowledge useful for problem solving. Another thing is that the confusing input (assuming it will make it to the knowledge used for problem solving) will have the tendency to be eliminated because users will be rejecting solutions that were based on it. There is a lot more but I cannot explain it well in short. You might also ask yourself how you will, if your approach extends beyond language, know that any image or photo is unreal. GINA just stores URLs for images and users describe it using system's formal language (which I named GSL by the way - General Scripting Language). GINA deals with images in similar way as with above mentioned phrases. Regards, Jiri Jelinek --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com