Is a quantum basis fractal? To the extent you're suggesting some kind of quantum computation might be a good implementation for the structures I'm suggesting, though, yes. At least, Bob Coecke thinks quantum computation will be a good fit for his quantum style grammar formalisms, which kind of parallel what I'm suggesting in some ways. That's what they are working on with their Quantinuum, Honeywell and Cambridge Quantum spin-off (recently another 300 million from JP Morgan.) Here's a recent paper from their language formalism team (Stephen Clark a recent hire from DeepMind, I think, though I think Coecke did the original quantum and category theoretic combinatorial semantics papers with him when they were together at Oxford back from 2008 or so.)
>From Conceptual Spaces to Quantum Concepts: Formalising and Learning Structured Conceptual Models Sean Tull, Razin A. Shaikh, Sara Sabrina Zemljiˇc and Stephen Clark https://browse.arxiv.org/pdf/2401.08585 Personally I think they've gone off on the wrong tangent with that. I like the fact that Coecke has recognized a quantum indeterminacy to natural language grammar. But I think it is pointless to try to actually apply a quantum formalization to it. If it emerges, just let it emerge. You don't need to formalize it at all. It's pointless to bust a gut pushing the data into a formalism. And then bust a gut picking the formalism apart again to "collapse" it into something observable at run time. But these maths guys love their formalisms. That's the approach they are taking. And they think they need the power of quantum computation to pull it apart again once they do it. So there's quantum computation as a natural platform for that, yes. For the rest of what you've written, I don't well understand what you are saying. But if you're talking about the interpretability of the kind of self structuring sequence networks I'm talking about, paradoxically, allowing the symmetry groupings to emerge chaotically, should result in more "visible" and "manageable" structure, not less. It should give us nice, interpretable, cognitive hierarchies, objects, concepts, etc, that you can use to do logic and reasoning, much like the nodes of one of OpenCogs hypergraphs (it's just you need an on the fly structuring system like I'm talking about to get adequate representation for the nodes of an OpenCog hypergraph. They don't exist as "primitives". Though Ben's probably right they could emerge on top of whatever nodes he does have. But he's never had either the computing power, or, actually the LLM like relational parameters, to even start doing that.) So I see it as the answer for interpretability, logic, "truthiness", and all the problems we have now with LLMs (as well as the novelty, creativity, new "ideas" bit associated with the complex system side.) You only get the quantum like woo woo when you insist on squeezing the variability into a single global formalism. Practically, the whole system should resolve from moment to moment as clearly as the alternative perspectives of an Escher sketch appear to us. One or the other. Clear in and of themselves. Just that they would be able to flip to another state discontinuously depending on context (and essentially both be there at the same time until they are resolved.) On Wed, May 8, 2024 at 1:00 PM Quan Tesla <quantes...@gmail.com> wrote: > > If I understood your assertions correctly, then I'd think that a > quantum-based (fractal), evolutionary (chemistry-like) model would be > suitable for extending the cohesive cognition to x levels. > > If the boundaried result emerges as synonymous with an LLM, or NN, then it > would be useful. However, if it emerges as an as-of-yet unnamed, > recombinatory lattice, it would be groundbreaking. > > My primary thought here relates to inherent constraints in visualizing > quantum systems. Once the debate between simple and complex systems end > (e.g., when an absolutist perspective is advanced), then the observational > system stops learning. Volume does not equate real progress. > > With an evolutionary model, "brainsnaps in time" may be possible. This > suggests that scaling would be managable within relativistic and relevance > boundaries/targets. > > In the least, trackable evolutionary pathways and predictability of the > pattern of tendency of a system should become visible, and manageability > would be increased. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tb63883dd9d6b59cc-M210f900801eb7251599971d1 Delivery options: https://agi.topicbox.com/groups/agi/subscription