Jim, I've been trying to wrap my mind around a significant semantic gap that seemingly exists between the philosophy, concepts, and methods, which different researchers have been employing for understanding AGI. I'm going to try and breach this gap.
When I read what you're saying, I comprehend it. Yet, when I speak in a systems language, which is well documented by Science, and not a developer' language per se, then it's rather foreign sounding to some. For example, you said: "And I wasn't thinking of hypergraphs as being completely connected all of the time, since relationships in AI are conditional." This should be true for narrow AI in reductionism only, which is the more physical (in the sense of programmable) aspect of any, functional system. Furthermore: "And I was also thinking that relations (or edges of hypergraphs) could be used as references to subsystems (or subgraphs)." In theory, this might be so, but in practice, the dimensional shifts implied within a hypergraph already scales way beyond many invisible boundaries. Unless you set the boundary down firmly, in a hypothetical sense, as you would for placing a scope limit on a project, then the workspace may become too large to be dealt with at the programmable level. Even further: "Can you use stochastic methods to compare theories to models (or to constraints or goals of the theory)?" If you meant by "stochastic methods", such methods which resulted in any weighted state, then they should help to elucidate the understanding of those aspects of a system under consideration, but it's unlikely that they would be sufficient for a holistic system all on their own. Second to last you stated: "I don't see why not. It is like using what-if situations. I guess a stochastic method - in the traditional sense - would be part of what-if modelling. That makes sense right?" I'm open to correction, but "traditionally" - in the West - , stochastic methods for "what-if modelling" were employed to help support biased decision making, very pertinent to large-scaled strategic contexts. Last: By "testing theory for stochastic, CAS at the highest level of abstraction." I meant that I develop theory for predicting systems of ultra complexity, (as CAS), but dealing with the dynamical environment of the system as well as it tends to evolve and shifts independently along a stochastic scale (more at a logical level than measuring every little thing at the physical/operational level). I include these perspectives assuming any system, or parts thereof, or wholes thereof, would tend towards optimal efficiency, as effective complexity, zero-point energy, AKA states of equilibrium. In other words, for holistic, natural systems, one always has to include the perspectives of the linear and alinear for any subsystem. Even so, there other perspectives to include. For example, considering a rainbow in the sky. This is being done in order to develop sufficient understanding for "weaving" the integrals of a "single" system from it. I think programmers far too often try to "weave" systems from their code, trying to manifest the idea of the system they have via the hard constraints of a simple tool. I just cannot see how typing a million lines of code would result in a dynamical CAS. Maybe, others realized this as well. For this reason alone, most of AI would eventually fall flat on its ass, and perhaps cause devastating damage to the economies and control systems they are being deployed in. For example, consider the Boeing disasters. That was a direct failure in conditional narrow AI, a design and programming oversight. The whole stinky fish was supposed to have been blamed for those material losses and deaths, but only the head was. Frightening, really. Further, consider the covid-19 experience, as zero accountability attached to experimental biotechnology being released among - forced upon - living systems and life-supporting ecologies. Perhaps even in the sense of AI integration into living systems, trying a shortcut to AGI (in my opinion). Many recent patent approvals support this notion. How can this possibly not end up being devastating to the living ecology? To continue with answering your question on abstraction. Quantum systems could be in states of chaos and order near instantly, many times over, as a life cycle of naturally-scripted chain reactions, yet what would emerge may be tremendous bursts of energy. An active volcano and everything seen or unseen it affects over time, might be one such an example. As such, in this context of my limited answer; abstraction simply means different levels/aspects of "visible/tracable" knowledge pertinent to any system being considered. By the same token, deabstraction would mean including methods for instantiating "unseen" knowledge and expressing that in a language suitable to the rest of the system (preparing it to be coded in an application), which was already made visible to the public (integrating it with the modules of code that was already written as an application). Without running ahead of myself, such deabstraction would then be likened to an act of integrating a previously-unseen perspective, into the wholeness of the knowledge model being assembled, manifesting the functionality thereof. For example, it would be similar to integrating quantum entanglement into an existing AGI systems model. None of this would become possible without having the support of scientifically sound methods. In my view, the correct toolkit of philosophy, language, methodology, convention, information processing technology, and testability has to be employed as a suitable ontology for doing AGI-related development. Even so, it has to be kept in a ready-state of flexibility. Simply by doing that, massive uncertainty would be introduced into the process. We should be asking then: "How to reduce uncertainty within AGI developments." >From what I'm observing, the theory of order from chaos is being applied in a >fatalistic sense. First, there's no real control at all. Where there's >control, little real progress is being made. And so on. There are deep cracks appearing in this "rocking-boat scenario", but the fatalist Nihilists are seemingly saying: "Be damned with this world and its inhabitants. It was shit anyway to begin with. Let's start over." The resultant destruction is being welcomed as an "expected outcome". I think this kind of thinking is insane. Maybe it's a case of; humankind should not be messing around with biosynthetic AGI at all? Who would be willing to consider that? If not, how can such persons with restrictive mindsets have the mind to develop trustworthy AGI in the first instance? How can they be trusted by the unknowing society at all? Too many really complex questions there, right? ________________________________ From: Jim Bromer <[email protected]> Sent: Sunday, 24 October 2021 03:02 To: AGI <[email protected]> Subject: Re: [agi] Re: Ideas Have to Act on Other Kinds of Ideas Nongrate: I did not understand what it was that you getting at. I still don't completely get it. And I wasn't thinking of hypergraphs as being completely connected all of the time, since relationships in AI are conditional. And I was also thinking that relations (or edges of hypergraphs) could be used as references to subsystems (or subgraphs). However, I started wondering: Can you use stochastic methods to compare theories to models (or to constraints or goals of the theory)? I don't see why not. It is like using what-if situations. I guess a stochastic method - in the traditional sense - would be part of what-if modelling. That makes sense right? And, what did you mean by, "testing theory for stochastic, CAS at the highest level of abstraction." ? 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/T1675e049c274c867-M252989fe0130e882846e488a> ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1675e049c274c867-M5259f045ebafcc14bab4c857 Delivery options: https://agi.topicbox.com/groups/agi/subscription
