AI's Thoughts on GenAI As Art
How Does the 19th-Century Fight Over Photography vs Painting Inform the 21st-Century Debate on AI versus Human-made Art?
I, a human (I think), asked ChatGPT the following question:
Prompt:
How can the lessons learned of the photography versus painting as art debate inform the discussion around human and AI made art?
ChatGPT Said:
Great question. The 19th-century fight over whether photography “counted” as art maps surprisingly well onto today’s human vs. AI-made art debate. Here’s a tight set of lessons you can lift directly from that history—and how to apply them now.
1) “New tool” ≠ “no art”
Then: Early critics called photography a merely mechanical process; artists proved authorship lived in choices (framing, exposure, printing, retouching).
Now: With AI, authorship similarly rests in intent + constraint + iteration (dataset curation, prompting strategy, model fine-tuning, selection, editing, presentation).
Takeaway: Evaluate artistic decisions, not just the tool. Ask: What problem was the artist solving? What constraints were set? What was rejected and why?
2) New media find their own aesthetics
Then: Photography won legitimacy by leaning into what it did uniquely well (movement, candidness, seriality, reproducibility), not by imitating painting.
Now: AI art’s strengths are scale, combinatorics, probabilistic variation, and fast iteration.
Takeaway: Don’t judge AI only by how well it mimics human mark-making. Judge how it extends the visual/semantic space beyond human-only throughput.
3) Process transparency matters (but doesn’t define value)
Then: Pictorialists vs. straight photographers debated manipulation; both gained acceptance once processes were named and understood.
Now: Disclose model, dataset sourcing (at least categories), prompt/fine-tune steps, and post-processing.
Takeaway: Create a process label (like a wine label):
“Human-only, analog/digital”
“Human + AI (curated data; directed prompts; hand-edited)”
“AI-primary (automated batch; minimal human selection)”
4) Indexicality, aura, and the “hand”
Then: Walter Benjamin’s “aura” vs. mechanical reproduction reframed value around context, not just uniqueness. Photography still developed forms of aura (vintage prints, signed editions, process rarity).
Now: AI outputs are infinitely reproducible, but aura can re-enter via provenance, editioning, performance/installation context, and live generation.
Takeaway: Build aura with edition control, cryptographic provenance, performance-based display, or time-bound generative runs.
5) Skill shifts, not skill disappearance
Then: Skill moved from brushwork to optics, chemistry, darkroom craft, sequencing.
Now: Skill shifts to data ethics, model steering, prompt engineering as design brief, multimodal compositing, curation.
Takeaway: Update curricula and juries to read AI craft (dataset design, failure-mode management, long-loop iteration logs).
6) Democratization + gatekeeping
Then: Cheap cameras broadened who could make images; institutions adapted with new canons and genres.
Now: AI lowers barriers even further; expect volume explosions and new folk/vernacular genres.
Takeaway: Curate with new rubrics (original problem setting, iterative risk-taking, conceptual coherence) instead of pedigree or tool purity.
7) Economic + labor questions must be addressed head-on
Then: Photo threatened portrait painters; new markets (editorial, advertising, art photography) emerged.
Now: AI threatens illustration and stock markets; new roles appear (model custodians, data curators, AI art directors).
Takeaway: Build fair-use norms, licensing schemes, and revenue-share models for training data; commission frameworks that explicitly price human direction.
8) Legal and ethical clarity enables cultural acceptance
Then: Copyright for photography matured; authorship became clearer.
Now: Clarify authorship (human intent), rights in training data, consent, credit, and disclosure.
Takeaway (policy sketch):
Disclosure: Identify AI involvement and process level.
Consent/Credit: Respect opt-out datasets; credit notable sources when practicable.
Provenance: Attach verifiable metadata (CID/hash, model/version).
Review: Flag sensitive content and bias audits for juried shows.
9) Hybridity becomes the mainstream
Then: Photo + paint, collage, photogram, mixed media flourished.
Now: Expect AI + photography + 3D + robotics + performance.
Takeaway: Stop binary framing. Curate hybrid categories: “Generative Assemblage,” “Performative Generation,” “Algorithmic Documentary,” etc.
A practical evaluation rubric (usable by juries, galleries, or teams)
Intent & Concept
Clear inquiry or thesis? Novel question or perspective?
Process & Craft
Demonstrable human direction (logs, iterations, dataset rationale)?
Mastery of AI’s failure modes and constraints?
Form & Aesthetics
Does it leverage the medium’s native strengths (variation, scale, recombination) rather than copy analog looks?
Context & Ethics
Transparent provenance; considered stance on data use and bias.
Impact
Resonance, originality, and contribution to discourse or practice.
Score each 1–5; publish the rubric alongside the show or portfolio.
Implementation quick-start (for creators and orgs)
Creators
Keep an iteration journal (key prompts, forks, rejects, edits).
Ship with a process label + provenance file (model, version, seed if relevant).
When possible, license or self-source datasets; document consent posture.
Galleries/Competitions
Require a disclosure card (tools, datasets, human role).
Judge with the 5-part rubric; allow hybrid categories.
Offer editioning guidance for AI works (caps, on-chain proofs, or time-windowed generation).
Educators
Teach dataset ethics and prompt as design brief.
Assign projects that force medium-native exploration (e.g., controlled variation studies).
Bottom line
From photography’s path we learn: legitimacy arrives when we judge by artistic intent, process rigor, and medium-native contribution—not by tool purity. If we build clear norms around authorship, disclosure, provenance, and ethics, the culture will absorb AI as another powerful—sometimes radical—art medium, just as it did photography.
End ChatGPT Response
=========
What are your thoughts, fellow human?
The image above, human or AI? Why do you think that?