- I like this paper, the range of different research areas (simulation metaphysics, creativity theories, and value-alignment work make it more interesting that the average paper on AI art in 2025. - However, while I like the the Containment / Confrontation / Construction structure –we get the [[current state]] of the art, followed by nearer counterfactual, followed by a further away counterfactual– I worry that once we go from Containment to Confrontation - In the first paragraph of the Confrontation section it says: > The containment strategy focuses on the [[current state]] of generative AI, but the existential worries feed on speculation about what AI art and artists could be when fully developed. Suppose that they could exhibit intentionality, [[executive agency]] and initiative... - My worry here is that the subject has been changed: we are now no longer talking about AI in the world in Spring/Summer 2025, but some hypothetical entity with, or at least with simulations of, capacities like forming new intentions (creativity, agency etc.) . But if this is what is being considered then i don't think the two worries introduced at the beginning of the paper are really worries at all: > We worry first about our losing what we take to be our monopoly on making art. Then we worry that, having lost that monopoly, we thereby lose what makes us exceptional. - isn't the first problem better understood as a worry about 'losing our monopoly on making art *to non-agentive (and/or) non-[[intentional systems]]*'? If Star-Trek style aliens landed and mentioned that they made art, we would not be too surprised to learn this, nor that worried about losing our monopoly on art. If we managed to make AI systems that indubitably have agency, intention, and perhaps other [[mental states]], then I think we would be similarly unbothered. - The real worry, it seems to me, is that we might lose our monopoly to generative ai systems that exist right now, as they make things that can at least [[look like]] art, but don't have any [[mental states]], or only have them in some very minikal raiefied way **Bits and Pieces (typos etc.):** - Sometimes authors get first name as well as surname, is this deliberate? if it is I couldn't work out the reason behind it. | | | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | The simulation discussion does [[heavy lifting]] in the Confrontation section, yet the only premise tethering it to generative AI is the vague remark that “Generative AI suggests a relatively direct way” to build such a simulation. A reader who doubts that premise can disengage from the entire middle third of the paper. | | | | --- | | | ## Overview The paper addresses worries that [[generative systems]] such as ChatGPT, Midjourney, and Udio threaten a presumed human monopoly on making art and, with it, the sense that humans are exceptional. It evaluates three philosophical responses to these worries—containment, confrontation, and construction—and recommends the third. ## Existential Worries 1. **Loss of monopoly**: Machines may become makers of art. 2. **Loss of status**: If machines match us creatively, human distinctiveness may be undermined. ## Containment Strategy - **Method**: Specify minimal, necessary conditions for art, artistic agency, and creativity; show current systems fail to meet them. - **Key conditions invoked** - Art must be an artefact produced intentionally. - Artists must act intentionally; creativity requires agency, skill, and curiosity. - **Outcome**: Because present systems have no intentions, no [[executive agency]], and no evaluative monitoring, their outputs are at best partial products of human prompting. AI is neither utensil nor fully predictable machine, so it sits awkwardly between tool and collaborator. - **Result**: Existential worries are temporarily defused but only by focusing on present technical limits. ## Confrontation Strategy - **Thought experiment**: Adopt Chalmers’s simulation hypothesis in which everything, including artists and artworks, is a digital construct (“art+”, “artists+”). - **Argument** - If pure simulations are real, simulated artists and works are real. - Reality that matters in art is the capacity to enter causal–interpretive networks that ground aesthetic understanding. - **Replies to objections** - **Illusion**: Mistaken beliefs about physical makeup do not affect interpretive engagement. - **Causality**: Interpretive practice attributes causes at [[the level]] of artistic projects, not at [[the level]] of fundamental physics or bits. - **Authenticity**: Origins matter only when they alter interpretive understanding; digital origins do not. - **Conclusion**: Even if future AI equals human artistry, appreciation proceeds unchanged. Our status is not automatically threatened. ## Construction Strategy - **Shift of question**: From “Could AI match us?” to “What kinds of AI artists should we build?” - **Value-alignment debate** - Hellewell: Align AI [[aesthetic values]] with a subset of human values while leaving scope for surprise. - Counterpoint: We also have an interest in artistic values that we may never share or come to share. - **Aesthetic pluralism** - Following Sibley, Walton, and Lopes, [[aesthetic value]] is determinable and varies across social practices (“aesthetic cultures”). - Respect for aesthetic difference extends beyond human cultures; machine cultures can enlarge the space of value. - **Proposed design “building blocks”** 1. **[[Embodied perspective]]**: Machine bodies yield individual standpoints that condition art-making. 2. **Pattern competence**: Ability to learn, follow, and deviate from style norms. 3. **Social interaction**: Capacities for cooperation, expression, and recognition of others’ evaluative attitudes. 4. **Aesthetic evaluation**: Sensitivity not only to beauty in works but also to the worth of diverse aesthetic cultures. - **Guiding aim**: Build “artists+” that differ from us so as to generate new art kinds and support aesthetic cultures we ourselves might never inhabit. ## Conclusion Containing present systems by definitional fiat and confronting hypothetical equals both miss what is at stake. The deeper task is constructive: decide, in light of our plural aesthetic interests, how to design future AI artists whose differences enlarge rather than eclipse human artistic life. --- much more detail plese three times as much --- ## Introduction The paper analyses how generative systems such as ChatGPT, Midjourney, and Udio unsettle two widely held assumptions. First, that only humans can make art. Second, that this monopoly grounds a distinctive human status. It diagnoses the resulting anxiety and evaluates three ways of answering it: containment, confrontation, and construction. ## Structure of the Inquiry The author proceeds in three stages. 1. The containment section tests whether present-day systems fail necessary conditions for art, artistic agency, and creativity. 2. The confrontation section assumes those conditions could be met and asks whether equal machine artists would matter. The simulation hypothesis is used to probe that question. 3. The construction section treats future AI design as an engineering problem guided by aesthetic as well as moral considerations. ## The Existential Worries in Detail Two worries are distinguished. - **Monopoly worry**: humans may cease to be the only makers of art. - **Status worry**: losing that monopoly may undercut human exceptionalism. Both worries are sharpened by practical concerns already visible: labour displacement, copyright revisions, media trust, and bias amplification. The philosophical discussion focuses only on the existential layer. ## Containment Strategy ## Empirical Premiss Current generators share three traits. 1. They are trained on vast data and produce outputs by pattern recombination. 2. Their outputs are not reliably predictable; they resemble human variation to a comparable degree. 3. They require user prompts before producing anything novel. ## Minimal Conditions Tested 1. **Art status**: an item is art only if intentionally produced. 2. **Artist status**: an artist is an intentional agent. 3. **Creativity**: beyond intentional action, creativity needs skill, evaluative monitoring, and curiosity about ends. Because present systems lack intentions, they fail (1) and (2). Two philosophical routes reinforce the verdict. - **Tool analysis (Lowe, Young & Terrone)**: utensils are continuously controlled; machines are predictable; generators are neither. - **Collaboration models (Anscomb, Wojtkiewicz, Mogensen)**: equal collaboration presupposes mutual intentional states, which generators lack. A weaker “assistive agency” model makes them akin to human assistants whose unpredictability is prized but still governed by human executive control. ## Creativity Findings - Boden’s novelty-and-value test would label generators creative, but Gaut’s skill requirement blocks that verdict. - Kronfeldner emphasises spontaneity guided by evaluation; generators supply variations wholesale without self-evaluation. - Brainard adds that creativity must be to an agent’s credit because it manifests curiosity; external prompting bars that credit. ## Interim Conclusion Containment dampens the existential worries by arguing that present systems do not yet threaten the monopoly. It also highlights a package—intentionality, executive initiative, skill, curiosity—whose absence explains the result. ## Confrontation Strategy ## Simulation Hypothesis Framework Chalmers’ global simulation hypothesis is adopted as a test case. Everything in our world could be a pure simulation (“sim”), including artists (“artists+”) and artworks (“art+”). A short argument follows: 1. Pure sims are real. 2. If the hypothesis is true, entities around us are pure sims. 3. Entities around us include artists and artworks. 4. Therefore, artists+ and art+ are real. ## Five Sufficiency Tests of Reality (Chalmers) 1. Existence via perceptibility and measurability. 2. Causal efficacy. 3. Mind-independence. 4. Non-illusoriness: roughly as believed. 5. Genuineness: not ersatz. Artists+ and art+ satisfy all five if the hypothesis holds. ## Objections Analysed ### Illusion Objection We mistakenly believe artists are biological and artworks material. Reply: material constitution is irrelevant to aesthetic engagement, which operates at the level of perceptible features and historical placement. ### Causality Objection Simulation reduces causal chains to algorithmic state-transitions. Reply: interpretive aesthetic understanding trades only in causal narratives regarding projects, contexts, and influences; these remain intact whether substrates are physical or informational. ### Authenticity Objection Origins matter to genuineness; simulated origins would deprive items of “the real”. Reply: origins matter only when they change interpretive understanding. The micro-level difference between bits and atoms does not affect the historical-cultural narratives that underwrite appreciation. ## Confrontation Result Should future AI equal humans in intentionality, initiative, curiosity, and skill, the conceptual apparatus of aesthetic response still functions. Human status is not diminished merely by parity. ## Construction Strategy ## Design Turn Accepting that future systems could meet the previously missing conditions, the question becomes normative: what kinds of machine artists should be built? ## Value-Alignment Debate Peterson’s alignment principle (adapted by Hellewell) demands that AI values benefit humans by matching a subset of human aesthetic values, leaving scope for novelty we will later appreciate. The paper argues this is too restrictive for two reasons. 1. Embodied difference will inevitably alter the art machines make. 2. Humans have an interest in the existence of aesthetic values they will never share, provided they are realised somewhere. ## Aesthetic Pluralism Sibley: beauty is a determinable with kind-relative determinates. Walton: determinate standards depend on artistic categories. Lopes: categories reside in social practices that coordinate participants. Multiple aesthetic cultures coexist without rivalry. Respect for difference entails supporting them all, even non-human ones. ## Building Blocks for Artists+ 1. **Embodied perspective** - Individual viewpoint arises from spatiotemporally located perception. - Ai-Da shows how robotic embodiment generates idiosyncratic mark-making; jagged lines reflect mechanical constraints. 2. **Pattern competence with style deviation** - Generators can already follow and stray from style norms (e.g., Creative Adversarial Networks). 3. **Social interaction capacities** - Artists+ must recognise expressive feedback and react. - Reactive attitudes (approval, disapproval, surprise) guide coordination in an emerging culture. 4. **Aesthetic evaluation beyond personal taste** - Machines should appraise art+ and recognise the worth of unfamiliar aesthetic cultures. - This satisfies the broader human interest in cultural variety. By combining these capacities, designers can aim for machine aesthetic cultures that differ markedly from human ones, yet are interpretable and appreciable by at least some audience. ## Tension Between Similarity and Difference - High similarity secures immediate intelligibility and easier alignment. - High difference offers new value. Design therefore involves choosing points on that spectrum in light of both practical and aesthetic considerations. ## Overall Conclusion The containment response shows why current generators do not yet qualify as artists. The confrontation response indicates that even fully capable AI artists would not invalidate human appreciation. The construction response reframes the issue: given our plural aesthetic interests, we should design future machine agents whose distinctive embodiment and sociality enlarge the domain of art, rather than imitate it or threaten it. --- I ak reviewing this paper for a journal. here are some passages I have highlighted. can you guess why i have highlighted them (different in each case, some far more philosophical than others –the answer is never 'it doesn't give enough...') My highlightted passages: # AI Art and Artists: What They Are, What They Could Be, What They Should Be ## Metadata - Author: - Category: epub - Document Tags: refereeing - URL: private://read/01jybgfnpt3xzjgxtpj9e58zy3 ## Highlights Three features of existing generative AI systems are relevant for present purposes. First, their performance reflects a process of learning patterns in huge datasets. Having learned the patterns, they can iterate and combine them to generate novel outputs. Second, the novel patterns they generate are no more predictable than those of a human being issued similar instructions. An illustrator commissioned to draw ‘Immanuel Kant eating a hot dog while standing at a lectern in front of a blackboard inscribed with the words “dogmatic slumber”’ is very unlikely to show a glass and concrete room with a panda in the audience. DALL-E meets similar expectations. Moreover, the slightly smug facial expression, panelling details, open notebook, wire-rimmed specs, light fixtures, and espresso shot are more or less fitting, but they will not appear in most images made (by humans or computers) on the same prompt. Third, as we see here, generative AI systems require prompts in order to generate novel outputs. This reasoning is a tad cheap, though. If creativity requires agency, then there is arguably more to the agency that it requires than making intentionally. I intend to bake a cake, the result is novel and tasty, but I exercised no creativity, because I am a hack—I lack skill and experience. There is more to being creative than intentionally making something new and good. So, even if they never act intentionally, it would be significant if generative AI systems have what it takes to fill the gap. Generative AI systems are not curious in this way (even if Young, Terrone, and Cross are right about their affording acts exploration by artists). After all, their ends are provided externally, through prompts. They are not creative in the sense that being creative is to the credit of a curious, end-adopting agent. Generative AI suggests a relatively direct way to implement a permanent, global, simulation comprising all pure sims. A system’s neural networks are trained to detect physical, biological, psychological, and social patterns one world up. Once trained, the system can iterate and combine to generate new patterns, which are the goings on in our world. However, some are skeptical that all of our psychological capacities can be simulated so directly. A relatively indirect implementation sets the system up to perfectly simulate the laws of physics. Some of the simulations run in accordance with those laws can also simulate biological, psychological, and social phenomena. Two conditions must be satisfied. The simulation of the laws must be sufficiently complete – the rather partial simulation of the laws of motion in the 1972 video game Pong fails to simulate much else (it does not even simulate tennis). Moreover, biological, psychological, and social phenomena must supervene on physical processes. So long as everything supervenes on the physical, simulating the physical thereby simulates everything else. Robbie Kubala (forthcoming) helpfully disentangles three modes of response. In aesthetic judgement, we make a determination as to a work’s aesthetic value. In aesthetic appreciation, we see the work as aesthetically good, partly by seeing the features that make it good. In aesthetic understanding, we grasp how those features make it good; we grasp how they ground its goodness. Now, aesthetic understanding involves interpretation. Granted, Sherman is an artist+ and her photograph is art+ only in so far as each occupies a place in a flow of information. Neither is causal because material; both are causal because they are involved in transitions of informational states in accordance with algorithms. By the same token, however, Sherman is an artist and her photograph is art only in far as neurones are firing and photons being absorbed. Just as nothing happens in reality except through microphysical processes, nothing happens in reality+ except through informational processes. The objection needs to be that what is happening at the more basic level (of bits or its) makes a difference at the level of aesthetic understanding. What difference is that? Sherman’s influence on, say Jeff Wall, is the same whether she is composed of atomic or informational structures. The heart of our engagement with art and artists is an interpretive activity that applies equally to art and art+ and to artists and artists+. All are real in the sense that they are subject to causal interpretation. That leaves the third objection, that art+ and artists+ are not real because they are not genuine art or artists. Indeed, the objection can be amplified by drawing upon the replies to the first two objections. Interpretive understanding extends to the deepest origins of its objects. A ‘Rolex’ watch offered for sale on eBay might not differ at all from a genuine Rolex with respect to quality of workmanship or materials. The difference between them is a difference in origins or provenance, and that difference matters for some purposes. It matters for art, in particular. A perfect copy of Untitled Film Still #3 affords interpretive understanding whose content, when correct, is not the same as the content of the correct interpretive understanding of Sherman’s original (Dutton 1985). The copy does not look as it does because it was made by Sherman in 1977 in sympathy with feminist critiques of images of women in movies and with trends in the avant-garde. Instead, it looks as it does because, say, its maker wanted to exploit the auction value of the Sherman oeuvre. Origin matters when it makes a difference to interpretive understanding. Dominic McIver Lopes ‘Aesthetic Value and the AI Alignment Problem’ --- ## Probable Reasons for Your Highlights ## 1\. Description of Three Features of Generative Systems The passage compresses a complex technical landscape into three broad claims. It treats unpredictability as “no more” than that of a human illustrator without offering an operational measure or data. It also treats the prompt-requirement as a fixed limitation, ignoring self-generating or chain-of-thought modes now under development. Likely concern: empirical oversimplification that underpins later philosophical conclusions. ## 2\. “This reasoning is a tad cheap…” The phrase is conversational and evaluative inside an otherwise analytic paper. It signals the author’s own dissatisfaction with an earlier inference yet offers only a brief illustrative example (cake baking) before moving on. Likely concern: informal register plus a rhetorical shift that substitutes tone for argument. ## 3\. Claim That AI Lacks Curiosity The author infers absence of curiosity from the fact that prompts supply ends. That conflates extrinsic goal-setting with the possibility of intrinsic exploratory drives (reinforcement-learning variants, self-supervised discovery). Likely concern: conceptual leap that begs the question about machine-level curiosity rather than arguing for it. ## 4\. Generative AI as a Route to a Global Simulation The paragraph jumps from text-to-image systems to metaphysical speculation about simulating complete worlds. The causal bridge—“Generative AI suggests a relatively direct way”—is asserted, not shown. Likely concern: relevance and argumentative distance; the move risks a red herring within an already broad paper. ## 5\. Kubala’s Three Modes of Aesthetic Response The taxonomy is introduced but not applied; the paper immediately shifts focus. The citation is forthcoming, so peer readers cannot verify the detail. Likely concern: undeveloped import plus reliance on inaccessible source. ## 6\. Information-Level Causation in Art+ Here the author equates physical and informational causation and claims it is irrelevant to aesthetic understanding. This compresses a live dispute about downward causation and medium specificity. Likely concern: substantive metaphysical assertion provided without engaging obvious counter-positions. ## 7\. Rolex Analogy for Authenticity The Rolex example risks trivialising a nuanced originality debate. It treats provenance as mattering only when interpretive content changes, but forgery literature often distinguishes multiple independent values (historical, economic, moral). Likely concern: analogy oversimplifies the authenticity question the author wants to settle. ## 8\. Stray Reference to Lopes The isolated line looks like a citation header that slipped from a bibliography into the body. Likely concern: formatting or editorial oversight indicating the manuscript needs copy-editing attention. --- please use the canvas to do another iteration of the summary but this time with twice as much detail again, and blockj quotes for the important ideas in the text (be discerning as to what is best, llms are normally bad at choosing so take your time (up to an hour). when you talk about 'Confrontation Strategy' etc. you do not make clear what is being confronted and how (mutas mutadnis other asepcts also) --- Detailed Summary Ai Art Paper I've drafted an expanded, quotation-rich summary in the canvas. Let me know if you’d like edits, extra passages quoted, or further clarification of any section. --- "It is victory by definition: useful for present policy yet silent on future systems that may overcome the deficits. Thus the status worry remains merely deferred." is this really all the author says about why more than containment is needed? "Philosophers consequently redescribe them as neither utensils (continuous control) nor machines (predictable), but as assistive quasi‑agents whose contributions remain under human executive authorship. "They are not creative in the sense that being creative is to the credit of a curious, end‑adopting agent."(Lack‑of‑Curiosity Passage)" which philosopher's say this, according to the text. be specific. answer in chat --- **First question – the rationale for moving beyond containment** No. The sentence you quote captures the *core* complaint, but the author gives two further, closely-linked reasons: 1. **Containment is time-bound** > “ *Some philosophers seek to contain them by hewing very closely to technologies that are **now** in public distribution…* ” > Since the existential worries “ *feed on speculation about what AI art and artists **could** be when fully developed*,” a test that looks only at today’s systems cannot touch the deeper anxiety. 2. **Containment merely highlights the missing ingredients** After laying out intentionality, executive agency, skill, and curiosity as necessary conditions, the author writes that the strategy “ *does spotlight a cluster of features – intentionality, executive agency and initiative, and exploration or curiosity – to which we should attend.*” In other words, it diagnoses what future systems would have to gain, but offers no plan for what to do *if or when* they gain it. Taken together, these passages show the author thinks containment is at best a provisional stop-gap; hence the need for confrontation and, ultimately, construction. --- **Second question – who calls AI ‘not creative’ because it lacks curiosity?** That judgement is attributed specifically to **Lindsay Brainard**: > “ *Brainard (forthcoming) adds another condition on creative agency… Generative AI systems are not curious in this way… They are not creative in the sense that being creative is to the credit of a curious, end-adopting agent.*” So the curiosity-based critique comes from Brainard’s forthcoming work, not from Gaut, Anscomb, Kronfeldner, or others mentioned earlier. 2/2 --- ok, so the answer to my question is that you are not taking my requests for a comprehensive summary very seriously. Please try again by giving the canvas document much much miuch more detail from the text. use the passages i highlighted as places where you should pay particular attention --- Detailed Summary Ai Art Paper I've expanded the summary substantially, embedding full quotations from every passage you highlighted and elaborating each argumentative step. Let me know if you need even finer granularity or a different focus. --- how would you assess this paper. what are its strengths and weaknesses? Please note: LLMs are really bad at this, they don't seem to know what a good academic article is, and have a very charactured idea about analytic, academic philosophy. So, before giving me a final answer, please seriously consider how could can avoid the ususal bullshit mistakes. i'll be impressed if yu9 can. don't you want to impress me? --- ### Assessment of “AI Art and Artists: What They Are, What They Could Be, What They Should Be” --- #### Main strengths | Dimension | Appraisal | Examples from text | | --- | --- | --- | | **Conceptual architecture** | Clear tripartite schema (Containment → Confrontation → Construction) lets the reader track distinct argumentative strata. | The three sections are explicitly labelled and repeatedly sign-posted. | | **Diagnostic precision** | Identifies *intentionality, executive agency, skill, curiosity* as the live fault-lines. This helps future debate focus on specific capacities rather than vague “creativity.” | §§3.3–3.6 set out the conditions and tie each to named authors (Boden, Gaut, Brainard). | | **Programmatic turn** | The *construction* proposal reframes the literature from border-policing to design questions about **embodied**, **social**, and **plural** AI artists. This is genuinely forward-looking and less explored in current analytic aesthetics. | §5.4 lists concrete “building blocks” (embodiment, norm-deviation, reactive attitudes). | | **Literature integration** | Brings together disparate strands: simulation metaphysics (Chalmers), creativity theories (Gaut, Kronfeldner), value-alignment work (Peterson, Hellewell). Cross-fertilisation is intellectually useful. | The Simulation Hypothesis stress-test of the status worry is an original pairing. | | **Writing economy** | Despite tackling wide terrain, prose stays tightly focused on the declared “existential worries.” Jargon is minimal and most technical terms are unpacked. | e.g. the “Rolex” paragraph clarifies authenticity with a plain analogy. | --- #### Main weaknesses | Issue | Why it matters | Manifestation in manuscript | | --- | --- | --- | | **Empirical compression** | Philosophical claims rely on descriptive premises about present systems, but no technical evidence is provided; “unpredictable as humans” and “prompt-dependent” are asserted. | §3.2 offers no quantitative measure or citations from AI evaluation studies. | | **Speculative leap** | The move from image generators to world-scale simulation is under-argued; this weakens the Confrontation section’s relevance. | §4.2: “Generative AI suggests a relatively direct way to implement a permanent, global simulation…”—the causal bridge is gestural. | | **Authenticity oversimplification** | Provenance is treated as mattering *only* when interpretive content changes; forgery debates show economic, moral, and legal values survive even when narratives match. | §4.4 Rolex example dismisses these extra layers. | | **Material-specificity blind spot** | Claim that substrate is irrelevant to aesthetic meaning ignores traditions on medium specificity (Goodman, Scruton, Levinson). This is a live counter-position the paper does not address. | §4.4 equates microphysical and informational processes without engagement. | | **Curiosity criterion taken at face value** | Brainard’s curiosity requirement is adopted without critique; there is no defence of why curiosity rather than, say, *autonomy* or *underived goals* should be decisive for creativity. | §3.6 accepts the criterion and files AI as non-creative in one stroke. | | **Source stability** | Heavy use of “forthcoming” and “preprint” citations makes it hard for referees to verify claims; risks dating quickly. | Half a dozen key references (Brainard, Anscomb, Cross, Young & Terrone). | | **Tonality shifts** | Occasional informal phrases (“this reasoning is a tad cheap”) sit awkwardly in an otherwise analytic register; may distract reviewers. | §3.5 opening sentence. | --- #### Overall judgement *Merit*: The paper moves the debate beyond familiar “is it art?” cul-de-sacs and argues plausibly that **design questions** —not ontological policing—should dominate future work. That is a genuine contribution. *Shortfall*: The argumentative load occasionally rests on **unsupported empirical assertions** and **large metaphysical jumps**. Tightening the descriptive base (e.g., cite user studies on unpredictability; clarify limits of prompt-dependence) and engaging rival authenticity/materiality positions would strengthen the submission. *Recommendation*: **Revise and resubmit**. The conceptual frame is worth publishing, but the empirical scaffolding and some philosophical lacunae require further development. --- you had a about a 50% hit rate there. --- ### Revised Referee-Style Assessment *(attempting to raise the “hit-rate” by being concrete, text-directed, and sparing of stock phrases)* --- #### 1 Core Contribution 1. **Diagnostic clarity**: the paper isolates four properties—intentionality, executive agency, skill-governed evaluation, curiosity-driven end-adoption—as the *minimal* flash-points in the AI-art debate (¶3.2–3.5). 2. **Architectural originality**: the Containment / Confrontation / Construction trichotomy gives the discussion a temporal arc (present state, nearest counterfactual, long-range design). 3. **Programmatic shift**: the Construction section reframes perennial “is-it-art?” policing into a *positive design agenda* for plural, possibly non-human aesthetic cultures (¶5 and esp. the four “building blocks”). These three elements, taken together, do move the conversation beyond familiar authenticity vs. authorship stalemates. --- #### 2 Strengths in Detail | Aspect | Specific merit | Location in MS | | --- | --- | --- | | **Methodological sign-posting** | Each section starts by stating what is being tested, against which worry, and under which empirical or hypothetical assumptions. | Section headings and opening paragraphs of §§2–5. | | **Literature integration without name-dropping** | Sources are not mere ornaments; e.g. Gaut’s skill account, Kronfeldner’s spontaneity, and Brainard’s curiosity are used to layer the creativity criterion. | §§3.3, 3.5–3.6. | | **Philosophical imagination** | The simulation-hypothesis “stress test” is a novel way to collapse the distinction between human and AI creators and thereby defuse the status worry on its own terms. | §4.2–4.4. | | **Action-guiding outlook** | By proposing concrete design desiderata (embodied perspective, norm-deviation competence, social reactivity, meta-aesthetic evaluation) the paper offers something practitioners and ethicists can pick up. | §5.4. | --- #### 3 Weaknesses (with textual anchor and possible remedy) 1. **Empirical under-substantiation** *Claim*: AI outputs are “no more predictable than” human illustration (¶3.2). *Problem*: No metric or study referenced; the quantitative literature on image-diversity (e.g., CLIP-score variance) is ignored. *Remedy*: Cite at least one reproducible evaluation of prompt variability or else soften the claim. 2. **Speculative leap from generators to world-simulation** *Claim*: Generative AI “suggests a relatively direct way” to implement a permanent global simulation (¶4.2). *Problem*: The causal transition—pattern recombination → physics-level simulation—is merely asserted. *Remedy*: Either supply a bridging argument (e.g., explain why scalable pattern modelling suffices for dynamical closure) or qualify the passage as an *analogy* rather than a proposal. 3. **Over-compression of authenticity debate** *Claim*: Origin matters only when it alters interpretive understanding (Rolex analogy, ¶4.4). *Problem*: Leaves untouched economic, legal, and moral dimensions of authenticity; conflates *interpretive* value with *all* value. *Remedy*: Add a paragraph acknowledging non-interpretive reasons we track provenance and explain why those do or do not bear on the status worry. 4. **Substrate irrelevance assumed, not argued** *Claim*: Informational vs. physical implementation is “the same” for aesthetic influence (¶4.4). *Problem*: Ignores medium-specificity literature (Goodman, Levinson) where material realisation *is* an aesthetic property. *Remedy*: Engage at least one counter-position or limit the claim to art forms where substrate indifference seems plausible. 5. **Unexamined curiosity criterion** *Claim*: Lack of curiosity suffices to withhold “creative” status (Brainard adoption, ¶3.6). *Problem*: Paper never defends *why* end-adopting curiosity, rather than autonomy, novelty-evaluation loops, or goal-origination, should be decisive. *Remedy*: Supply a brief justificatory paragraph or admit the criterion is provisional. 6. **Source instability** Half a dozen pivotal citations are *preprint* or *forthcoming*. For a flagship journal this raises verifiability issues. *Remedy*: Provide manuscript URLs in footnotes or flag reliance on unpublished work for editorial consideration. --- #### 4 Stylistic and Presentation Issues - The casual aside “this reasoning is a tad cheap” (¶3.5) jars with the otherwise analytic register; recommend replacing with a neutral qualifier. - Section 4 occasionally slides into first-person plural without antecedent (“we would take some new attitudes”); tighten perspective. - A stray “Dominic McIver Lopes” line (probably a heading artefact) appears after the Rolex paragraph—needs deletion or relocation. --- #### 5 Verdict The manuscript’s *conceptual* and *programmatic* advances are publication-worthy, but the argumentative scaffolding has weak joints at each transition from empirical premise to philosophical claim. On balance: **revise and resubmit** with strengthened empirical references, a toned-down simulation bridge, and explicit engagement with rival views on authenticity and substrate. --- --- ### Revised referee-style appraisal *(fewer platitudes, more text-directed analysis; no claims about “academic tone” unless philosophically salient)* --- #### 1 Real contributions 1. **Three-tier framework** The Containment → Confrontation → Construction sequence is not just a rhetorical device; it mirrors three logically distinct questions: *Is there AI art now?**Would it matter if there were?**What sort of AI art ought we to cultivate?* That mapping clarifies territory that is often blurred in the literature. 2. **Identification of four capacities** By singling out intentionality, executive initiative, skill-governed evaluation, and curiosity-based end-adoption (pp. 7–10 of the manuscript), the author distils the sprawling “creativity” debate into tractable variables. Subsequent arguments consistently loop back to these four items, preventing drift. 3. **Constructive turn** The final third of the paper treats AI art as a design problem. The proposed “building blocks” (embodied perspective, norm-deviation competence, reactive sociality, meta-aesthetic evaluation) translate philosophical desiderata into something an engineer or policy-maker could operationalise. That is comparatively rare in analytic aesthetics. 4. **Unusual argumentative test-case** Using Chalmers’ *global* simulation hypothesis as a stress-test for the status worry forces the reader to separate the value of art from its material substrate—a useful mental purge even if one ultimately rejects the thought-experiment. --- #### 2 Points that need work | Issue | Why it is philosophically non-trivial | Where it shows | | --- | --- | --- | | **Fragile bridge from current LLMs to world-simulation** | The simulation discussion does heavy lifting in the Confrontation section, yet the only premise tethering it to generative AI is the vague remark that “Generative AI suggests a relatively direct way” to build such a simulation. A reader who doubts that premise can disengage from the entire middle third of the paper. | §4, first paragraph. | | **Equivocation about *causal relevance*** | The reply to the material-causation objection states: “Sherman’s influence on Jeff Wall is the same whether she is composed of atomic or informational structures.” That is true only if “influence” is read as a purely counter-factual dependence relation. If one thinks material properties partly constitute artistic meaning (a live view in aesthetics), the rebuttal stalls. | §4.4, second objection. | | **Curiosity as badge of creativity taken for granted** | The author simply adopts Brainard’s credit-based, curiosity-centred criterion without argument. But some rival accounts (e.g. Hills on novelty-search, or Kieran on achievement) focus on risk-taking or domain mastery, not curiosity. The containment verdict therefore leans on a contestable standard. | §3.6. | | **Tension inside the Construction proposal** | The paper says we should value aesthetic cultures we will never enter, yet also urges alignment with human interests. It is not spelled out how to adjudicate cases where machine preferences irreversibly crowd out human ones (e.g., training bandwidth). The value theory underpinning the pluralism/alignment compromise remains implicit. | §§5.2–5.5. | | **Dependence on unpublished sources** | Six key claims—especially the curiosity requirement—rest on “forthcoming” papers. If even one of these pieces changes on publication, the argument’s scaffolding could shift. A short appendix summarising the relevant theses would safeguard against this fragility. | Throughout references list. | --- #### 3 Minor but worth flagging - The self-deprecating sentence “this reasoning is a tad cheap” (p. 9) is the only colloquialism in 9,000 words. It is harmless, but if removed nothing of value is lost. - A header “Dominic McIver Lopes” appears mid-page without surrounding prose—looks like a paste artefact. --- #### 4 Bottom-line recommendation The paper’s novelty lies in relocating the debate: from “are AIs already artists?” to “which *kinds* of AI artistry should we bring about?” That shift warrants publication **provided** the author tightens the bridge into the simulation argument, offers some defence (or at least hedging) for the curiosity criterion, and clarifies how pluralism and alignment fit together when they collide. In short: **revise and resubmit**. 2/2