Nano: I thought that you might be thinking of something like that. As I tried to form a response I finally have started to make sense of it. A discrete mathematical algorithm will suffer from the combinatorial complications. A deep learning net seems to be able to deal with those kinds of problems - as long as there are good approximate solutions available. So then, is there anyway a symbol-net could outperform neural nets (if rapidity was not a fundamental problem)? It seems to me that you may be approaching a possible solution to that kind of situation. Does that analysis make sense to you? Jim Bromer
On Sat, Feb 23, 2019 at 3:39 PM Nanograte Knowledge Technologies < [email protected]> wrote: > <https://mathinsight.org/definition/network>In other words...the purpose > should be functional RNA. > > Now, is an AGI blueprint justified? > > https://en.wikipedia.org/wiki/Gene_regulatory_network > <https://en.wikipedia.org/wiki/Gene_regulatory_network> > Gene regulatory network - Wikipedia > <https://en.wikipedia.org/wiki/Gene_regulatory_network> > A gene (or genetic) regulatory network (GRN) is a collection of molecular > regulators that interact with each other and with other substances in the > cell to govern the gene expression levels of mRNA and proteins. These play > a central role in morphogenesis, the creation of body structures, which in > turn is central to evolutionary developmental biology (evo-devo). > en.wikipedia.org > > > > > > ------------------------------ > *From:* Jim Bromer <[email protected]> > *Sent:* Saturday, 23 February 2019 6:31 PM > *To:* AGI > *Subject:* Re: [agi] Some thoughts about Symbols and Symbol Nets > > So now I am beginning to wonder if nano was talking about some kind of > syntactic rules of combination for symbols and that all knowledge would > have to 'emerge' by selection of which combination, within a network, were > selected by learning. > Jim Bromer > > > On Fri, Feb 22, 2019 at 1:54 AM Nanograte Knowledge Technologies < > [email protected]> wrote: > > If this were the case, I'd agree with you. What I'm proposing is content > independent and context dependent. It is suitable for CAS applications. It > is not "designed"to be constrained, but to identify and normalize a > primary, contextual constraint in order to resolve it in an adaptive sense. > Meaning, humans do not resolve it, but the contextually-bound instance of > the system does. By implication, all possible meanings of the symbol are > always resident and latent. However, the decisive meaning for a particular > context is alive for the duration of that contextual reference in the > greater schema of information. In other words, the correct answer is always > possible within a particular context. Such is the basis of critical > thinking, to derive the correct answer to every situation. Yes, there is an > underlying assumption, which is that a correct answer exists for every > context, but this could be proven scientifically. > > Previously, mention was made in the forum about hierarchy (meaning > control). Having hierarchy within a systems constructs provides a system > with robustness and integrity, which translates into informational > reliability. Now, it seems the question of validity has been settled, but > not the one on reliability. What I'm proposing already has embedded into > its language what could be termed validity and reliability, at scale. > > That is where the analogy of the tyre hitting the tar has relevance, or > the point in project and program management where the the essential truth > hits home. It is where the absolute impact on a situation has most effect. > We could also argue how it resembles the point of effective complexity, > which is the point-of-reasoning we are all desire within an AGI entity. > > You stated: "The term 'context-free' refers to the syntactic context, not > the greater global context (of variable type definitions or redefinitions > and so on)." > > >>I strongly disagree with this view. In a semantic system, which I > contend is required for a symbolic system to become operational, syntax > lends itself to context specificality. I think that point was born out via > recent discussions on the forum. > > I think no designer should (be allowed) arbitrarily decide local and > global boundaries. That's a recipe for disaster. Boundaries are outcomes of > the inherent (natural) design resident within components and collective > contexts. In addition to a specified, context boundary, the underlying > methodology should allow for scalability, which is not only an issue of > size, but also of adaptive scope (implying boundary adaptiveness). In this > sense, a contextual/systems boundary would be structured/unstructured in a > construct of thesis/antithesis - 2 parts of the same coin. Perhaps in using > this approach, we would achieve Haramein's et al's perspective on a > finite-infinity in a computational model. > > When looked at via a report, or a snapshot view, such a system would > appear to be structured (which it also is). However, if you could view it > as a continuous value stream, as a movie, it would be possible to watch > (and trace) how it morphed relative to its adaptive algorithm - as an > unstructured system. In time, for each specific context, it should become > possible to identify the patterns of such morphing, and apply predictive > algorithms. > > I think one outcome (there are multiple outcomes) of such a system would > resemble a Symbol Net. It should theoretically be possible to extract such > nets from the "live" system. I think this is rather similar to how we do it > within society today. > > Robert Benjamin > ------------------------------ > *From:* Jim Bromer <[email protected]> > *Sent:* Thursday, 21 February 2019 11:46 PM > *To:* AGI > *Subject:* Re: [agi] Some thoughts about Symbols and Symbol Nets > > A contextual reference framework, designed to limit the meaning of a > symbol to one meaning within a particular context, would only displace the > ambiguity - unless the language was artificially designed to be that way. > So called 'context-free' languages, ironically enough, do just that. They > have some value in AI, but it is difficult to see how it could be used as > an effective basis for stronger AI. The term 'context-free' refers to the > syntactic context, not the greater global context (of variable type > definitions or redefinitions and so on). Perhaps the term is misunderstood > or misused in compiler design, but, a lot like applied logic, its > application is useful because it can be limited to 'a framework' (like a > local function and so on). So perhaps industry did develop a way to limit > ambiguity within a contextual framework, but so far it has not proven to be > very useful in stronger AI. The nature of *limiting* ambiguity of a symbol > (or possible referential signification) does not seem to be a very powerful > tool to rely on when you are trying to stretch the reach of current (or 30 > year old) ideas to attain greater powers of 'understanding'. > Jim Bromer > > > On Thu, Feb 21, 2019 at 2:49 PM Nanograte Knowledge Technologies < > [email protected]> wrote: > > If one had a contextual reference framework, each symbol would always have > one meaning within a particular context. Searches would always be optimal. > An example of this is evidenced within the Japanese language. So, the 30+ > years of waiting was for no good reason. If only the industry had developed > appropriate theory for dealing with scalable ambiguity, which it probably > had. > > ------------------------------ > *From:* Jim Bromer <[email protected]> > *Sent:* Thursday, 21 February 2019 8:13 PM > *To:* AGI > *Subject:* Re: [agi] Some thoughts about Symbols and Symbol Nets > > I asked myself the question: If a theory of symbols was a feasible basis > for stronger AI, then the earlier efforts in discrete AI or weighted > reasoning should have show some promise. They should have worked. So why > didn't they work? Then I remembered that they did work with small data > sets. GOFAI did work as long as it could make a rapid search through the > possible candidates of meaning, but because combinations of symbols have > meaning, and because each symbol may have more than one meaning or referent > the problems of combinatorial complications presented a major obstacle to > developing the theories much further, My opinion is that the ambiguities or > multiple possible relevancies of a symbol (sub-net) can themselves be used > to narrow the possible meaning of the symbol (sub-net) when needed in > reasoning. We just need a huge amount of memory in order to create an index > of generalizations to use the information adequately. We now have that > scale of memory and processor speed available to us so we can try things > that could not be tried in the 1970s and 80s. > Jim Bromer > > > On Tue, Feb 19, 2019 at 12:45 AM Nanograte Knowledge Technologies < > [email protected]> wrote: > > Linas, Mike and Jim > > I find this to be a most-interesting conversation. Primarily, because it > suggests that the development of AGI may not only be challenged by the > development of competent theory, but also by programming capabilities to > put the theory into practice. > > Evolving such an architecture then, should desired outcomes be for an AGI > entity to achieve self-theory and self-programming? In its most-simplistic > from, a symbol is but a suitable abstraction of a greater reality, > similarly to how a symbol of a red-heart might be an abstraction of a > sentient being. Concept? Context? Meaning? Transaction. > > Who, or what decides what the symbolic world should look like and its > meaningfulness? The global state of social evolution may cause terrible > confusion in any learning entity. The learning objectives should be > specific, not generalized. Isn't learning incorrectly worse than not > learning at all? > > I think, there should be a general agi-architecture, replete with the > capacity to develop and function within a generic world view. Furthermore, > I think the real value would be derived from specialized AGI. Maybe beyond > that, an AGI architecture would - in future - morph via its own social > networking and inherent capabilities to become more than the sum of its > parts. > > To do so, would take a lot more than intersections. I agree with the > statements made about binary/vector theory, but it seems obvious to me that > this would not be sufficient for this task. You implied fractals. To my > mind, that would be the only way to proceed. As such, I think the primary > issue remains a design issue. > > Robert Benjamin > > ------------------------------ > *From:* Linas Vepstas <[email protected]> > *Sent:* Monday, 18 February 2019 10:36 PM > *To:* AGI > *Subject:* Re: [agi] Some thoughts about Symbols and Symbol Nets > > > > On Mon, Feb 18, 2019 at 1:17 PM Mike Archbold <[email protected]> wrote: > > I'm not sure I completely follow your point, but I sort of get it. > > I tend to think of symbols as one type of the "AI stuff" a computer > uses to think with -- the other main type of "AI stuff" being neural > networks. These have analogies to the "mind stuff" we use to think > with. > > > Symbol systems and neural-net systems can be seen to be variants of the > same thing; two sides of the same coin. I posted an earlier thread on this. > There's a 50-page long PDF with math, here: > https://github.com/opencog/opencog/raw/master/opencog/nlp/learn/learn-lang-diary/skippy.pdf > > roughly: both form networks. They differ primarily in how they represent > the networks, and how they assign weights to network connections (and how > they update weights on network connections). > > > On their own, symbols don't mean anything, of course, and inherently > don't contain "understanding" in any definition of understanding. > > Is there a broad theory of symbols? We kind of proceed with loose > definitions. I remember reading the Newell and Simon works, and they > say AI strictly in terms of symbols and LISP (as I recall anyway). > > > Yes. The "broad theory of symbols" is called "model theory" by > mathematicians. It's highly technical and arcane. It's most prominent > distinguishing feature as that everything is binary: it is or it ain't. > Something is true, or false. A formula takes values, or there is no such > formula. A relation binds two things together, or there is no relation. > There's no blurry middle-ground. > > So, conventionally, networks of symbols, and the relations between them, > and the formulas transforming them -- these form a network, a graph, and > everything on that network/graph is a zero or a one -- an edge exists > between two nodes, or it doesn't. > > The obvious generalization is to make these fractional, to assign weights. > Neural nets do this. But neural nets do something else, that they probably > should not: they jam everything into vectors (or tensors) This is kind-of > OK, because the algebra of a graph is a lot like the algebra of a vector > space, and the confusion between the two is an excusable mistake: it takes > some sophistication to realize that they are only similar, but not the same. > > I claim: fix both these things, and you've got a winner. Use symbolic > systems, but use fractional values, not 0/1 relations. Find a good way of > updating the weights. So, deep-learning is a very effective weight-update > algorithm. But there are other ways of updating weights too (that are > probably just as good or better. Next, clarify the > vector-space-vs-graph-algebra issue, and then you can clearly articulate > how to update weights on symbolic systems, as well. > > (Quickly explained: probabilities are not rotationally-symmetric under the > rotation group SO(N) whereas most neural-net vectors are: this is the spot > where deep-learning "gets it wrong": it incorrectly mixes gibbs training > functions with rotational symmetry.) > > So Jim is right: discarding symbolic systems in favor of neural nets is a > mistake; the path forward is at the intersection of the two: a net of > symbols, a net with weights, a net with gradient-descent properties, a net > with probabilities and probability update formulas. > > -- Linas > > > On 2/18/19, Jim Bromer <[email protected]> wrote: > > Since I realized that the discrete vs weighted arguments are passe I > > decided that thinking about symbol nets might be a better direction for > me, > > > > 1. A symbol may be an abstracted 'image' of a (relatively) lower level > > object or system. > > An image may consist of a feature of the referent, it may be an icon of > > the referent or it may be a compressed form of the referent. > > 2. A symbol may be more like a 'label' for some object or system. > > 3. A generalization may be represented as an image of what is being > > generalized but it also may be more of a label. > > 4. An 'image', as I am using the term, may be derived from a part or > > feature of an object or from a part of a system but it may be used to > refer > > to the object or system. > > 5. An image or label may be used to represent a greater system. A system > > may take on different appearances from different vantage points, and > > analogously, some features of interest may be relevant in one context but > > not from another context. A symbol may be correlated with some other > > 'object' and may stand as a referent to it in some contexts. > > > > So, while some symbols may be applied to or projected onto a 'lower' > corpus > > of data, others would need to use an image to project onto the data > field. > > I use the term, 'lower' somewhat ambiguously, because I think it is > useful > > to symbolize a system of symbols so a 'higher' abstraction of a system > > might also be used at the same level. And it seems that a label would > have > > to be associated with some images if it was to be projected against the > > data. > > > > One other thing. This idea of projecting a symbol image onto some data, > in > > order to compare the image with some features of the data, seems like it > > has fallen out of favor with the advancements of dlnns and other kinds of > > neural nets. Projection seems like such a fundamental process that I > cannot > > see why it should be discarded just because it would be relatively slow > > when used with symbol nets. And, there are exceptions, GPUs, for example, > > love projecting one image onto another. > > Jim Bromer > > > -- > cassette tapes - analog TV - film cameras - you > > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/Tcc0e554e7141c02f-Me60b5807cb33012c06624ade> ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tcc0e554e7141c02f-M1f62b8f3fc27a60c096d5fa5 Delivery options: https://agi.topicbox.com/groups/agi/subscription
