RE: [agi] The Grounding of MathsEdward, As a v. quick reply to start with, grounding means "making sense of" - using ALL senses not just visual.
"Did you think about it?" "Yes I did. A lot" Your ability to reply here is based on your sensory experience of having thought - that is not a visual.sensory experience. "I felt sad" - is a grounded statement - grounded in your internal kinaesthetic experience of your emotions. Would you like to rephrase your question in the light of this - the common sense nature of grounding, which I think obvious and beyond dispute? One further and huge point. My impression is that you may be like most AI guys - somewhat lacking in both an evolutionary and developmental perspective (perhaps it's to do with being so future-oriented). Consequently, you are making another mistake as drastic as thinking that grounding is just one sense. You are leaving out the developmental history of human understanding altogether. No child will be able to understand this legal document. Many adults won't be able to understand it either. Why not? Because human understanding and the human model of the world have to go through a great number of stages. It takes many stages of intellectual development to go from something like: "Cat bite Lucy" to "Animals eat people" to "Human-animal relationships are fraught with conflict" to "This Darwinian picture of evolution presupposes arms races as an important factor." Ben a mathematician looked at the immense complexity of the numbers and maths he deals with and failed to appreciate that they are composite affairs - which can only be mastered, psychologically, stage by stage, building from very directly grounded numbers like ten's, to very complexly and indirectly grounded numbers like trillions, "very large numbers", irrational numbers etc. For maths this is actually rather obvious WHEN you look at things developmentally. You are making the same mistake in jumping to the most complex forms of language and concepts and asking: how can these immensely complex and abstract concepts possibly be grounded? It's a good question. But,as with maths, the broad answer is: only developmentally, grounded stage by grounded stage. If human reasoning were as you think it, based only (or only in certain areas) on manipulation of symbolic networks, you and other humans would have no problem jumping to an understanding of that legal document at the age of 5. In fact to understand it, you have had to built up, stage by stage, a vast GROUNDED model of the world - you have had to learn what "courts" are, what a "justice" is (and you had to SEE courts and watch movies to do that), you had to look at several machines before you could understand what a general concept like "mechanism" meant, you have had to LOOK at patents and then physically compare them with actual machines, and then physically compare those machines with other machines to see whether their parts are indeed new or essentially copies. You had to SEE books of logic etc etc And great sections of that immensely complex grounded model will be invoked - UNCONSCIOUSLY - as you read the document. And even so as you read sections like: Claim 4 of the Engelgau patent describes a mechanism for combining an electronic sensor with an adjustable automobile pedal so the pedal's position can be transmitted to a computer that controls the throttle in the vehicle's engine. When Teleflex accused KSR of infringing the Engelgau patent by adding an electronic sensor to one of KSR's previously designed pedals, KSR countered that claim 4 was invalid under the Patent Act, 35 U. S. C. ยง103, because its subject matter was obvious. you will repeatedly have momentary if not extensive difficulties understanding which parts and which machine is being referred to at which point. Why? Because your brain is continually trying to MAKE SENSE of those damn words and SEE where the pedal is in relation to the throttle, and which pedal is which etc. As with all legal documents where "the party of the third part believes that the party of the second part attacked the party of the first part.." your unconscious mind goes blank trying to GROUND and physically connect all these different parts. And that's partly why we pay you lawyers such outrageous sums of money, because you are prepared to so torture your brain's grounding/ sense-making faculties - and most of us can't be bothered! Does all this make things a little clearer? (Note that you are trying to SEE how difficult prose is grounded!) (There's a lot more to explain but I hope I have got you to start asking questions. The key is to understand that high-level conceptual language like: "This Darwinian picture of evolution presupposes arms races as an important factor" is very complexly and hierarchically grounded, just as but even more so, than numbers like 10 to the 150 describing the number of chess moves, are grounded. You are having natural difficulties because you want something like that Darwininan sentence to be as directly grounded as "Cat bite child" is. I and psychology need to set out the stages and levels of grounding and processing that sentence - & I'm working on it occasionally]. P.S. I can add one useful definition here - i.e. that the grounding/processing of complex general and abstract language (especially "intellectual prose") can and does involve SEVERAL levels of conceptual AND image processing simultaneously. It's awesome stuff. ----- Original Message ----- From: Edward W. Porter To: agi@v2.listbox.com Sent: Saturday, October 13, 2007 1:51 PM Subject: RE: [agi] The Grounding of Maths I have attached an MSWord document of a recent important patent law case.(the highlighting is mine.) Read it and tell me what percent of it -- other than its description on the invention and alleged infringing device -- has visual imagery as its major direct source of primary inference. I would contend it is virtually impossible to understand such a document though visual grounding and reasoning alone. Although visual experience may provide some grounding for most of the words and relationships described, at the level of abstraction commonly used in this documents the key associations and implications are of generalizations that are quite removed from the specifics -- or even of any meaningful low level generalizations -- of any such visual memories. Yes, you could stop and apply a mental visual image to almost every word or phrase in this text, and such visualizations do occasionally occur to me when reading such a case, but, at least consciously, what I am aware of most are thoughts of words, of legal concepts defined largely in terms of words, of relationships between such concepts, mostly at a non-visual level. I am not denying that visual memories and associations may be playing a some role in my subconscious during such thinking, but for many of the types of things being described I have no specifically visual memories, they are mental constructs I have learned largely through reading. And to the extent that I do have grounding for them from my own personal experience, they are often of experience that involve letters, phone conversations, and emotions as much as visual memories. In the absence of evidence to the contrary why is that we are to believe that that virtually all relevant inference is from visual memories, models, and associations, and not from the memories, models, and associations represented in much of the rest of our brain. It is widely believed that the cortex has a surprisingly uniform architecture. Why then assume that inferencing only takes place from information in the visual part of the brain, and not in any of its other parts. We know that meaningful reasoning can be done from systems representing information in non-visual forms. Bayesian nets, Copycat, Shruiti, Fair Isaac, and CYC, can all do some such reasoning, and some of them can do reasoning sufficiently well as to be commercially extremely useful. Why can we thus not imagine that with a really powerful AGI like Novamente -- that should have much more powerful methods of inferencing and learning -- will be able to do reasoning of the type required to understand the above paper, based largely on non-visual information, such as that which would be included in a good university law school library, that has been added on top of a common sense world-knowledge base that would preferably include visual grounding. Why do some people seem to assume that only visual memories, patterns, and associations are valuable for human reasoning? Are then only meaningful generalities in life those caught in visual memories? The "G" in "AGI" stand for "general" not "visual". <<...>> Edward W. Porter Porter & Associates 24 String Bridge S12 Exeter, NH 03833 (617) 494-1722 Fax (617) 494-1822 [EMAIL PROTECTED] -----Original Message----- From: a [mailto:[EMAIL PROTECTED] Sent: Friday, October 12, 2007 10:20 PM To: agi@v2.listbox.com Subject: Re: [agi] The Grounding of Maths Edward W. Porter wrote: > In response to Charles Hixson's 10/12/2007 7:56 PM post: > > Different people's minds probably work differently. For me dredging > up of memories, including verbal memories, is an important part of my > mental processes. Maybe that is because I have been trained as a > lawyer. > > I am not arguing against the fact that visual memories play an > important role in human thinking. They do. I often do a lot of my > best thinking in terms of images. > > What I am arguing is that other types of grounding play an important > part as well. I am arguing that visual grounding is not necessarily > the largest force in each and every mathematical thought. Yes, the > human brain dedicates a lot of real estate to visual processing, but > if you take all of the language, behavioral, emotional and higher > level association areas, you have a lot of brain real estate dedicated > to concepts that are either non-visual or only partially visual. We > should not assume that all that brain real estate plays little or no > role in most thinking. > > Of course, I wouldn't be surprised if visual memories and patterns are > taking at least some part in the massively parallel spreading > activation and inferencing in the sub-conscious that helps pop most > thoughts up to consciousness -- without me even knowing it. But by > similar reasoning I would also assume a lot of non-visual memories and > patterns would also be taking part in such massive parallel > inferencing. > > In many types of thinking I am consciously aware of words in my head > much more than I am of images. Perhaps this is because I am a patent > lawyer, and I have spent thousands of hours reading text in which many > of the words have only loose association to concrete visual memories. > And as a lawyer when I read such abstract texts, to the extent that I > can sense what is in my consciousness and near consciousness, many of > the words I read seem to derive their meaning largely from other > concepts and memories that also seem to be largely defined in terms of > words, although occasionally visual memories pop out. > > When I read "The plaintiff is an Illinois corporation selling services > for the maintenance of photocopiers" it is probably not until I get to > "photocopiers" than anything approaching a concrete image pops into my > mind. > > Thus, at least from my personal experience, it seems that many > concepts learned largely through words can be grounded to a > significant degree in other concepts defined largely through words. > Yes, at some level in the gen/comp pattern hierarchy and in episodic > memory all of these concepts derive at least some of their meaning > from visual memories. But for seconds at a time that does not seem to > be the level of representation my consciousness is aware of. > > Does any body else on this list have similar episodes of what appears > to be largely verbal conscious thought, or am I (a) out of touch with > my own conscious processes, and/or (b) weird? > > I also do not visualize your plaintiff example except for the photocopier at first, but eventually I have to visualize to remember as a mnemonic technique. I do not think that your plaintiff example is a "verbal conscious thought," but a verbal memory. I first just memorized the words and sounds of the sentence without visualizing it. I can easily memorize the example by selectively memorizing the important words by rehearsal. Then I visualized the shape of the state Illinois from a map. For a corporation, I visualized a big building. For "selling services" I visualized a grocery store. I visualized the phrase "for the maintenance" to a picture of someone fixing a computer. I memorized a picture of a photocopier. For reconstruction of the sentence, I converted the images back to words. I will eventually forget the "selling services" part because I can still reconstruct the sentence without it. ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?&; <<...>> ------------------------------------------------------------------------------ This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?&; ------------------------------------------------------------------------------ No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.488 / Virus Database: 269.14.9/1067 - Release Date: 12/10/2007 18:02 ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=53245519-86b3df
