#meeting #work Conversation with Enrico about the [[The Environmental Aesthetics of Generative AI]] paper ## 1. Overall Context and Purpose 1. The participants are revising a draft paper on **[[environmental aesthetics of generative AI]]**—a framework adapted from Carlson’s [[environmental aesthetics]], which they apply to AI systems. 2. The discussion is primarily about refining the paper’s structure and clarifying [[key concepts]]. They frequently refer to approaching AI [[not just]] as an artifact but as an _environment_ shaped by various “forces” (technical, conceptual, cultural, etc.). **Primary Goal**: Finalize the paper, which is due the [[next day]], by removing redundancies, clarifying ambiguous language, and improving the flow of ideas (particularly around how they articulate “three ways of appreciating AI”). --- ## 2. Conceptual/Mathematical/Technical Forces vs. Cultural/Technical Forces - Early in the conversation, the speakers address the mention of **“conceptual, mathematical, and technical forces”** in the draft. - One [[speaker notes]] that “technical forces” can be understood as causal aspects of the system—e.g., how code or engineering shapes outputs—but “conceptual” or “mathematical” forces are less obviously “forces.” - They hypothesize these terms might be **an artifact of large [[language models]]** (where LLMs often like to list two or three parallel terms). - **Proposed Revision**: They consider replacing “conceptual or [[mathematical forces]]” with terms like “cultural forces,” since the data and social context shaping AI might be described more aptly as “cultural.” For instance, [[training data]] reflect cultural influences, user practices, and so on. - The upshot: They want to ensure it is clear what these “forces” actually do in shaping AI outputs. --- ## 3. Structure and Organization: Two Introductions, Abstract, and Redundancies ### 3.1 Two Introductions (Redundant Paragraphs) - The transcript reveals there are **two introduction paragraphs** in the draft that cover almost the same ground—both frame how Carlson’s [[environmental aesthetics]] relates to generative AI. - They see them as “two alternatives,” acknowledging they do not want to overload the opening. - **Plan**: Merge or replace one of them so the introduction is concise and clearly sets up the argument. ### 3.2 Carlson References: “Arousal Model” vs. “[[Environmental Model]]” - They spot a mismatch in the draft referencing **Carlson’s “arousal model”** (sometimes misheard or mistranscribed as “arouse or model,” “arise or model,” etc.). - Carlson contrasts the “arousal model” with his “[[natural environmental model]]” in his own writings, but the direct quote in the draft might confuse readers (it references “arousal” in a way that does not fully align with how [[the authors]] are discussing environment-based appreciation). - **Proposed Fix**: They consider either using a smaller part of the quote They do not want to derail the argument with too much detail about Carlson’s “arousal model,” which might not map neatly to generative AI. ## 4. Three Ways (or Modes) of [[Appreciating Generative]] AI A core portion of the transcript is spent on ordering or clarifying three distinct but overlapping modes of appreciating AI systems—derived from Carlson’s approach to environment. These modes are: 1. **Scientific / Theoretical Engagement** - In Carlson’s environmental aesthetics, knowledge of geology, biology, etc. can deepen one’s aesthetic appreciation of nature. Analogously, having some grasp of how AI works (training data, architecture, etc.) enriches appreciation of the system. - This parallels “looking at a natural environment through a scientific lens,” except now one is looking at a generative AI environment. 2. **Appreciating Outputs (like a Garden)** - They compare it to appreciating a _garden_—the final “cultivated” output. But the real focus is on how that output illuminates the **system** itself. - The conversation clarifies that they do **not** want to treat the output alone (like a standalone piece of art) but to highlight that _enjoying or examining the outputs_ can be a route to appreciating the system that produced them. - In an analogy: the speaker says if you see a beautiful garden, you might appreciate it, but you are also appreciating the “environment” and perhaps the gardener’s skill. 3. **Practical / Hands-On Engagement** - They liken this to being a _gardener_ inside the system, planting seeds and seeing how they grow. In AI terms, one is actively tweaking prompts, refining parameters, or coding new features. - This active involvement fosters a deeper “practical knowledge,” akin to “knowing by doing,” which further enhances the overall aesthetic appreciation. ### 4.1 Potential Reordering - They debate the **best sequence** of presenting these three modes: 1. **Scientific** → **Outputs** → **Hands-On**, or 2. **Scientific** → **Hands-On** → **Outputs**. - One speaker argues the “hands-on gardening” approach is the most distinctive, so it might be a good place to finish the discussion. Another speaker sees a narrative logic in going from theoretical to practical to final results. - They also consider a “dialectic or chronological progression” (from learning, to doing, to appreciating final forms). - **Tentative Decision**: They lean toward a structure that will be easiest to follow in the paper. They do not finalize the exact order in the transcript, but they emphasize that no matter how it is structured, they must make it _explicit_ that they are still discussing the **same system** from three perspectives. ### 4.2 Emphasis: “One Object of Appreciation, Three Methods” - They clarify that all three modes are about the **same “object”**—the AI system as environment—not three different objects. - The conversation points out that the “outputs” mode can be misunderstood as appreciating _just_ the outputs; the speaker wants to be sure the paper explains that outputs are part of how we see the system’s aesthetic value, _not_ the entire story. --- ## 5. Deadline Pressures and Additional Projects - The group notes the paper is due **“tomorrow by 23:59.”** One speaker says they will refine the draft in the morning, then forward it for final checks. - They mention they have multiple other writing projects: 1. A **Venice paper** on generative AI or aesthetics, though not identical to this one. 2. A “no shutter” paper (the context is not fully elaborated in the transcript, but it presumably also relates to AI or aesthetics). 3. A **psychedelics paper** (6,000 words, due in about two months). 4. Some form of **engineering paper**, with no clear mention of a deadline. - They briefly reference a **French Society of Analytic Philosophy event** (organized by “Alexander de claw”), where one speaker will present on AI aesthetics, and a **poetry conference** they have to attend or prepare for. --- ## 6. Minor Edits and Corrections Throughout, they identify small fixes needed: 1. **Consistent Title Usage**: They mention at some point that a title or heading references “appreciating natural answer” instead of “appreciating nature” or “natural environment,” calling it a likely glitch. 2. **Quotations from Carlson**: Some quotations or page references need to be updated. The portion that references Carlson’s “arousal model” (and everyday knowledge vs. scientific knowledge) may be shortened or replaced. 3. **Balancing Word Count**: They observe that some sections (e.g., the introduction) exceed the recommended length, so they plan to cut unnecessary redundancies. --- ## 7. Illustration Generation with AI (Midjourney) Towards the end, the speakers shift briefly to discussing **Midjourney**: - They are trying to design a **poster or visuals** for a presentation, apparently wanting an image of “playing cards” or “philosophical playing cards.” - They note that generating consistent suits or symbols on playing cards is challenging with AI. They also look at a few AI-generated images that produce “weird” or “abstract” outputs. - The conversation reveals they are trying to find an aesthetically pleasing but thematically relevant image—perhaps referencing four suits or four philosophers, but generative models often produce “strange” variations that do not fully align with standard playing-card visuals. - They scroll through example outputs (the speaker: “Go down, down, that’s cool,” or “This one is interesting… not quite playing cards, but it has the right vibe.”). - **Conclusion**: They will keep experimenting with Midjourney to arrive at a design that evokes the idea of “philosophical playing cards” or otherwise suits the event. --- ## 8. Addressing Ambiguous or Mistranscribed Elements Using the Draft Below are key ambiguities and how the draft clarifies them: 1. **“Arise or model” / “arouse or model” → “arousal model”** - The draft references Carlson’s “arousal model,” which stands in contrast to the “natural environmental model.” This clarifies the words heard in the transcript. 2. **“Conceptual, mathematical forces”** - The conversation in the transcript clarifies that the speakers want to rename these “cultural” and “technical” forces (or something similar) to avoid confusion. 3. **Outputs vs. The System** - From the draft, it is evident that the “outputs” dimension is a _sub-part_ of appreciating the entire system; the transcript’s conversation underscores this to ensure the final version of the paper does not appear to treat outputs as the sole focus. --- ## 9. Key Takeaways 1. **Integration of Carlson’s Environmental Aesthetics** - The group reaffirms that Carlson’s framework of knowledge-based appreciation of environments can be mapped onto AI systems, as if they were natural habitats. 2. **Three Interconnected Modes of Appreciation** - They find it essential to show how scientific knowledge, engagement with outputs, and hands-on creation each provide different but _complementary_ avenues for aesthetic appreciation of AI. 3. **Streamlining the Paper** - The final text must remove duplicated introductions, refine references, and present the “three modes” clearly and systematically. 4. **Deadline and Future Work** - The next steps are finalizing the draft by the next day’s deadline and then revisiting the other pending papers and conferences. --- ## 10. Conclusion This transcript shows a detailed conversation among collaborators refining a paper that seeks to extend _environmental aesthetics_ to _generative AI._ They examine the best way to describe the “forces” shaping AI, how to reorder or combine separate introductions, how to articulate three distinct yet unified methods of appreciating generative AI, and how to ensure the text is consistent, concise, and clear. They also address side topics such as designing conference posters with Midjourney and preparing for other upcoming papers and conferences. Overall, the conversation affirms that: - **Knowledge** (scientific or common-sense) is crucial for aesthetic appreciation (in line with Carlson), - **Outputs** are a means to appreciate the system rather than the only object of appreciation, and - **Hands-on or “gardening-like” engagement** reveals another, more immersive layer of aesthetic understanding. They plan to finalize these details, meet their submission deadline, and carry these insights forward into future research and presentations.