5 Future works
In future, following such conclusion, it is my purpose to extend those structures in which I laid out, verify their validity and sufficiency of such, and either replace them by something else, or proceed to construct them further on. That includes the theoretical system in which we can define, though temporarily, the research program in which start from ground zero of the constructs (i.e. the autopoietic system construction), the definition of intelligence revisited and to reinterpret benchmarks and verification methods as of current, model definition and more works on the evolutional, percolation patterns of models itself, clarifying and putting in it the technical sufficiency for the neural network formalism, and many more works to come.
5.1 Limitations, Errors, and Author Contact
Because it is inconclusive, mostly theoretical, and the author recognize inconsistency in certain parts of the paper, there are limitations and errors that would emerge in detail future revisions. Furthermore, because the paper also introduces conceptual frameworks for understanding artificial intelligence, and so on for future theoretical understanding, there exist errors and caveats of which should be addressed, or fixed if suitable reasons are to be provided. These are intended to guide readers in interpreting the materials, and to clarify the assumptions underlying the argument.
- Certain conjectures rely on vague conceptual interception, while some hints at idealized assumptions. For example, in conjecture 1.1 and conjecture 1.2, these are intentionally vague as to clarify in an operational sense, the perception of AI in modern standards. Those can be confusing, and there might exist fallacy within the conjecture itself. The Universal Approximation Theorem (UAT) and its simplified notion assume smooth, continuous functions over compact subsets, which may not hold for all real-world representation of data-encoding.
- Many concepts are theoretical, and there are no sufficient guarantees in which they would be accepted or not superseded by better replacements. This accounts for the dual formalism listed in section 3, and so on of the learning theoretic. In the appendix, the same goes for the layered Chinese Room formalism and its associated recursive threshold $Q$.
- Architectural limitations, specifically the examples on GNN and others, rely on current theoretical and practical results, and may not generalize to future architectures, alternative encoding schemes, or any relevant architectures outside the author’s knowledge.
- The definition of artificial and intelligence are inherently subjective. While this paper attempts to formalize even the notion of subjectiveness in such definition, interpretations might differ across contexts, disciplines, observers, and so on.
- Analogies such as ”man-in-the-box”, the argument about Searle, should be taken as heuristic pictures, rather than literal cognitive model. This also hints at a potential structuralization of such concept in a detailed form.
- Discussion on AGI and emergent intelligence also reflect current understanding, trends, hypes, directions, and also philosophical reasoning, which may evolve overtime as new theoretical or empirical findings emerge.
- Some claims, particularly regarding scaling laws, bottleneck phenomena, or functional expressivity, are derived from simplified or idealized settings, and thus is debated in the paper. For the neural scaling law, we directly argued it as a weak form of correlation via morphed reference space, though it does not inherently rule out the sufficiency in which such observation might help in practical settings. Thereby, it is also to be noted of those claims as to have controversial positions.
- The results in certain cases would be regarded mainly as conceptual guidance rather than prescriptive implementations, as such, production of certain analyses, and verification, very much relies on practical and experimental designs, of which the paper does not cover.
All interpretations, derivations, and conjectures henceforth in this paper are the responsibility of the author. While cares have been attributed to provide as rigorous as possible reasoning and many fail-safes to be added, errors in logic, assumptions, or presentations (spelling or grammatical errors) may exist. It is then encouraged to critically assess, challenge, or refine the presented ideas, and also contribute correcting inconsistencies or presentation issues. For questions, discussion, suggestions, and fixes, the author can be reached at this e-mail.