```todoist name: Today filter: "Today | Overdue" ``` # Brainstorm on [[the Aesthetics]] of Consciousness Ok so what the hell am I going to write about. Here is one idea. One way we could go with the [[aesthetic appreciation]] of [[conscious experience]] is to talk about [[vehicle properties]] and opacity. If we can experience our phenomenal states to be somewhat opaque then perhaps we can aesthetically appreciate them. We would, perhaps, be able to appreciate [[vehicle properties]], or even something like [[design properties]] –that is, how features of the vehicle realise features of the content. The difficulty with this idea is that the opcity and transparency debate is far from settled. And that [[the idea]] of a design property analog for phenomenal states is somewhat obscure. Here is another way to go. What about the [[aesthetic appreciation]] of our mental actions? Nguyen suggests that we can aesthetically appreciate our own actions, actions from the inside. He also suggests that we appreciate our mental actions as well ‘wonderfully precise turn of mind’. An initial worry about this is that it throws out the baby with the bathwater. When we talk about how wonderful consciousness is don’t we want to be talking about more ordinary forms of consciousness? Like the blooming buzzing confusing of the world around us? I think a response to me made here is that the phenomenology of mental actions can be thought of as a phenomenology or creation. Examples: Consciously bringing something to mind vs it suddenly appearing before ones mind. Doesn’t it seem as though you have created your thought in some way. When Nguyen talks about a wonderfully precise turn of mind, it sounds like he is talking about the act of creation. Think about a poet or a novelist feeling proud of a line they have come up with. We can admire the grace of a dancer, or admire the grace of our own dancing. We can admire the creativity of someone *coming up with* a funny remark, or we can admire ourselves for coming up with a funny remark. The first pair of cases can be thought of as actions simpliciter, the second can be thought of as creative actions. Problematise: But it might be argued that we still do not have a very well worked out idea about what would constitute an aesthetically pleasing action. Add the stuff from [[Ted Chiang]]’s Exhalation. If this is correct, then we need to flesh it out if it is to be convincing. ### 1. Representation: the “family” of the genera - Representation doesn't require a prior definition; it's sufficient to understand it through examples and basic principles. - Sophisticated systems adjust their behavior to environmental specifics, sometimes necessitating representation. - Representations stand in for specific features or structures not presently manifest, guiding behavior. - A representation scheme enables a variety of contents to be systematically represented and used properly under various conditions. - Misrepresentation is possible and varies by the scheme, but it does not equate to improper use or bad luck. ### 2. Canonical accounts of the genera - Representation involves a [[relational structure]], but distinct essences don't lie in the representative tokens themselves. - Logical representations are defined by a compositional semantics, allowing complex structures to systematically determine semantic significance. - Iconic representations achieve representation through isomorphism, where resemblance or abstract mathematical relationships map onto the content. - Distributed representations feature each part of the token contributing to the representation of each content part, with representation spread out and superimposed. ### 3. Problems with the Canonical Accounts - [[The author]] presents three counter-examples to challenge the adequacy of the canonical accounts, suggesting that each account fails to capture the essence of its own genus by admitting cases that belong to another genus. - A designer's floor plan and digital maps are shown to embody the token/content relation characteristic of logical representations, despite being paradigmatically iconic, which suggests a misclassification by the canonical accounts. - The abstractness of form and isomorphism is so broadly conceived in the canonical account of the iconic genus that it could encompass logical tokens as well, thus failing to exclude them and undermining the distinctiveness of the iconic representation. - Ordinary holograms, exemplifying distributed representation through their spread-out and superposed nature, are also straightforwardly iconic, indicating the insufficiency of the canonical accounts to maintain clear distinctions among the genera. - These examples collectively indicate that the canonical accounts are insufficiently restrictive, incapable of supporting principled distinctions among the genera, and mislocate the generic essences in the representing relation, suggesting a need for a fundamental reconception rather than mere refinement or acceptance of overlap. ### 4. The Representing/Recording Distinction - [[The author]] introduces outlandish theses to highlight [[the need]] for distinguishing between representing and recording, demonstrating the inadequacy of straightforward translations among [[representational genera]]. - **Outlandish Thesis #1:** It suggests the ease of translating iconic or distributed representations into logical representations, criticized for misunderstanding the complexity of perceptual recognition. - **Outlandish Thesis #2:** Proposes easily translating logical or distributed representations into iconic forms, ignoring the nuanced understanding required for accurate depiction from descriptions. - **Outlandish Thesis #3:** Implies the simplicity of converting logical or iconic representations into distributed ones, using "Turing's piano" as a metaphor to illustrate the spread and superposition of representational elements. - The distinctions between representing and recording are explored through examples, emphasizing that recording is a process yielding a record of its input, while representing involves a functional status or role. - The analysis extends to scenarios like court recording and the generation of descriptions from scanned images, underscoring the mechanical or trivial nature of recording processes compared to the nuanced and skill-dependent process of representing. ### 5. The Initial Problems Rediagnosed - [[The author]] revisits the counter-examples presented earlier, reevaluating them in light of the representing/recording distinction to diagnose the insufficiencies of the canonical accounts more accurately. - **Holographic images and Turing's piano:** These examples are reinterpreted to show that they involve recording rather than representing, emphasizing the technology's role rather than the representational genus. - **Picture theory of meaning:** This theory is critically examined, suggesting that logical representations recorded in an iconic form do not truly transform the nature of representation but merely change its medium. - **Recursive generation of floor plans and maps:** Analyzed as icons being recovered from logical recordings, challenging the straightforward classification of representations based solely on their generative process. - The section concludes by questioning the canonical accounts' effectiveness in distinguishing representational genera, suggesting a fundamental reconception rather than minor adjustments or acceptance of overlap. ### 6. Skeletal versus Fleshed-out Content - The distinction between skeletal and fleshed-out content is introduced to differentiate between the minimal and augmented content of representations. - Skeletal content is defined as the strict content of a representation, not mediated by any additional information, representing the bare-bones of what is communicated. - The metaphor of skeletal versus fleshed-out content highlights how everyday representations are shaped by their skeletal content but enriched by external influences. - Examples illustrate how detailed readings of descriptive reports reveal assumptions made by readers, distinguishing between what is explicitly stated and what is inferred or assumed, underscoring the notion of skeletal content. - The distinction is nuanced against other content distinctions like literal versus figurative or explicit versus implicit, emphasizing its unique approach to understanding representational content. ### 7. Distinguishing Genera by Contents - The motive behind distinguishing skeletal from fleshed-out content is to facilitate the characterization of representational genera based on the nature of their contents. - Skeletal contents of logical representations involve atomic facts about properties or relations, with a tradition of representing these through formal semantics extending to complex constructs. - The difficulty in characterizing non-atomic contents and the suggestion that represented content shares structures with alternative possible worlds underscore the complexity of defining logical content. - Iconic representations are understood through variations of values along dimensions, emphasizing the structured and measurable nature of these dimensions essential for representing shapes or variations. - The account contrasts logical and iconic representations by highlighting their reliance on absolute versus relative elements, respectively, to structure their contents. ### 8. Connectionist Networks - Connectionist networks, or distributed representations, are discussed as a distinct genus of representation, emphasizing their novelty and complexity compared to logical and iconic representations. - The structure of connectionist networks is outlined, including units with variable activation levels, connections with weights, and the distinction between cyclic and acyclic networks. - Network functionality is described in terms of input and output patterns, with a focus on the network's ability to group patterns by similarity, highlighting the importance of similarity relations in network operations. - The process of how networks induce similarity groupings and the implications for understanding complex input/output behaviors are explored, emphasizing the richness of potential representations in sophisticated networks. ### 9. Are there "Distributed" Representations? - The discussion questions whether the patterns of activation and connection weights in connectionist networks qualify as representations, distinguishing between ephemeral activation patterns and the more stable patterns of connection weights. - Activation patterns are often interpreted as distributed recordings, while connection weight patterns are considered potential candidates for illustrating a unique genus of representation due to their role in encoding a system's capabilities. - The conceptual challenge of regarding the embodiments of abilities or know-how as representations is addressed, suggesting that such patterns can serve as stand-ins for specific environmental features, aligning with the initial characterization of representations. - The distinctions between distributed representations and classical associationism are clarified, emphasizing the representational content's rootedness in the represented world and the interdependent nature of associative elements. ### 10. Conclusions and Morals - The chapter concludes that representational genera are distinguished by the characteristic structures of their represented contents, identified as absolute, relative, and associative elements for logical, iconic, and distributed representations, respectively. - The potential for mitigating the divide between parallel distributed processing (pdp) and symbolic artificial intelligence (AI) in cognitive science is discussed, arguing for the relevance and adequacy of network models in representing different kinds of cognitive content. - The unresolved problems of recognizing absolute elements in realistic situations and the possibility of deep cooperation between logical contents and distributed representations are highlighted, suggesting a symbiotic relationship between natural language and the distributed representations it is implemented in.