StarZero All shows Try the podcast clipper →

Apr 2, 2023 · 29m · another-podcast

AI, copyright and collective knowledge

Benedict Evans · 22m spoken Toni Cairn-Brown · 4m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Benedict Evans and Toni Cairn-Brown examine the complex challenges generative AI poses to copyright law, artistic intent, and creator monetization. Drawing on the history of photography, copyright precedents, and media industry disruptions, they explore how society can define and reward human originality in an era of automated synthesis.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 83% of the talking time here. How this is scored →

The hosts as informed peer 8.2 Guest teaching 0.8 Guest disagreement 1.3 The hosts pushing back 2.3
05100:0010:0020:003:06–5:49 · The hosts as informed peer 8/10 Tools Versus Artistic Vision in Photography and History Benedict dominates the discussion by drawing detailed analogies between AI generation and the history of photography, invoking Henri Cartier-Bresson's Leica and the evolution of Impressionism to argue that tool access does not equate to artistic vision. Toni readily agrees and reinforces the historical framing.5:50–10:50 · The hosts as informed peer 8/10 Prompt Engineering and Multi-Layered Style Synthesis Benedict demonstrates deep practical knowledge of generative models and style cues, detailing specific Midjourney and ChatGPT experiments using film stocks like Fujifilm Pro 400H and French 1960s sports cars. Toni enthusiastically validates this by reframing prompt engineering as its own distinct creative craft.10:51–15:46 · The hosts as informed peer 9/10 Legal Precedents and the Evolution of Copyright Concepts Benedict cites numerous legal and historical precedents, referencing the monkey selfie dispute, Ernest Bourguet's 1847 cafe performance lawsuit, and Voltaire's publishing maneuvers to contextualize how copyright frameworks struggle with technological shifts. Toni acts primarily as a supportive sounding board.15:46–21:04 · The hosts as informed peer 8/10 Attribution, Cultural Originality, and Collective Human Knowledge When Toni shares her experiment asking ChatGPT who owns its outputs, Benedict promptly intervenes with pushback, pointing out the secondary fallacy that the model cannot provide factual legal answers because it merely predicts text from training data. He expands on cultural shifts in attribution from Aztec sculptors to modern art.21:06–25:02 · The hosts as informed peer 8/10 Structural Disruption of Entertainment and Revenue Models Benedict explains entertainment industry contract shifts, contrasting traditional syndication windfalls like Jerry Seinfeld's with one-off Netflix streaming payouts and Scarlett Johansson's Disney lawsuit. Toni highlights the tension between treating AI training as collective human knowledge and the economic necessity of compensating individual creators.25:03–29:31 · The hosts as informed peer 8/10 Policy Objectives, Compensation Schemes, and Valuing Originality Benedict breaks down the mechanics and perverse incentives of pro-rata streaming royalty pools on platforms like Spotify, contrasting classical tracks with white noise spam and hip-hop sampling rules. Both conclude collaboratively that existing copyright concepts fail to map onto generative AI.3:06–5:49 · Guest teaching 0/10 Tools Versus Artistic Vision in Photography and History Benedict dominates the discussion by drawing detailed analogies between AI generation and the history of photography, invoking Henri Cartier-Bresson's Leica and the evolution of Impressionism to argue that tool access does not equate to artistic vision. Toni readily agrees and reinforces the historical framing.5:50–10:50 · Guest teaching 1/10 Prompt Engineering and Multi-Layered Style Synthesis Benedict demonstrates deep practical knowledge of generative models and style cues, detailing specific Midjourney and ChatGPT experiments using film stocks like Fujifilm Pro 400H and French 1960s sports cars. Toni enthusiastically validates this by reframing prompt engineering as its own distinct creative craft.10:51–15:46 · Guest teaching 1/10 Legal Precedents and the Evolution of Copyright Concepts Benedict cites numerous legal and historical precedents, referencing the monkey selfie dispute, Ernest Bourguet's 1847 cafe performance lawsuit, and Voltaire's publishing maneuvers to contextualize how copyright frameworks struggle with technological shifts. Toni acts primarily as a supportive sounding board.15:46–21:04 · Guest teaching 1/10 Attribution, Cultural Originality, and Collective Human Knowledge When Toni shares her experiment asking ChatGPT who owns its outputs, Benedict promptly intervenes with pushback, pointing out the secondary fallacy that the model cannot provide factual legal answers because it merely predicts text from training data. He expands on cultural shifts in attribution from Aztec sculptors to modern art.21:06–25:02 · Guest teaching 1/10 Structural Disruption of Entertainment and Revenue Models Benedict explains entertainment industry contract shifts, contrasting traditional syndication windfalls like Jerry Seinfeld's with one-off Netflix streaming payouts and Scarlett Johansson's Disney lawsuit. Toni highlights the tension between treating AI training as collective human knowledge and the economic necessity of compensating individual creators.25:03–29:31 · Guest teaching 1/10 Policy Objectives, Compensation Schemes, and Valuing Originality Benedict breaks down the mechanics and perverse incentives of pro-rata streaming royalty pools on platforms like Spotify, contrasting classical tracks with white noise spam and hip-hop sampling rules. Both conclude collaboratively that existing copyright concepts fail to map onto generative AI.3:06–5:49 · Guest disagreement 1/10 Tools Versus Artistic Vision in Photography and History Benedict dominates the discussion by drawing detailed analogies between AI generation and the history of photography, invoking Henri Cartier-Bresson's Leica and the evolution of Impressionism to argue that tool access does not equate to artistic vision. Toni readily agrees and reinforces the historical framing.5:50–10:50 · Guest disagreement 1/10 Prompt Engineering and Multi-Layered Style Synthesis Benedict demonstrates deep practical knowledge of generative models and style cues, detailing specific Midjourney and ChatGPT experiments using film stocks like Fujifilm Pro 400H and French 1960s sports cars. Toni enthusiastically validates this by reframing prompt engineering as its own distinct creative craft.10:51–15:46 · Guest disagreement 1/10 Legal Precedents and the Evolution of Copyright Concepts Benedict cites numerous legal and historical precedents, referencing the monkey selfie dispute, Ernest Bourguet's 1847 cafe performance lawsuit, and Voltaire's publishing maneuvers to contextualize how copyright frameworks struggle with technological shifts. Toni acts primarily as a supportive sounding board.15:46–21:04 · Guest disagreement 2/10 Attribution, Cultural Originality, and Collective Human Knowledge When Toni shares her experiment asking ChatGPT who owns its outputs, Benedict promptly intervenes with pushback, pointing out the secondary fallacy that the model cannot provide factual legal answers because it merely predicts text from training data. He expands on cultural shifts in attribution from Aztec sculptors to modern art.21:06–25:02 · Guest disagreement 2/10 Structural Disruption of Entertainment and Revenue Models Benedict explains entertainment industry contract shifts, contrasting traditional syndication windfalls like Jerry Seinfeld's with one-off Netflix streaming payouts and Scarlett Johansson's Disney lawsuit. Toni highlights the tension between treating AI training as collective human knowledge and the economic necessity of compensating individual creators.25:03–29:31 · Guest disagreement 1/10 Policy Objectives, Compensation Schemes, and Valuing Originality Benedict breaks down the mechanics and perverse incentives of pro-rata streaming royalty pools on platforms like Spotify, contrasting classical tracks with white noise spam and hip-hop sampling rules. Both conclude collaboratively that existing copyright concepts fail to map onto generative AI.3:06–5:49 · The hosts pushing back 1/10 Tools Versus Artistic Vision in Photography and History Benedict dominates the discussion by drawing detailed analogies between AI generation and the history of photography, invoking Henri Cartier-Bresson's Leica and the evolution of Impressionism to argue that tool access does not equate to artistic vision. Toni readily agrees and reinforces the historical framing.5:50–10:50 · The hosts pushing back 2/10 Prompt Engineering and Multi-Layered Style Synthesis Benedict demonstrates deep practical knowledge of generative models and style cues, detailing specific Midjourney and ChatGPT experiments using film stocks like Fujifilm Pro 400H and French 1960s sports cars. Toni enthusiastically validates this by reframing prompt engineering as its own distinct creative craft.10:51–15:46 · The hosts pushing back 2/10 Legal Precedents and the Evolution of Copyright Concepts Benedict cites numerous legal and historical precedents, referencing the monkey selfie dispute, Ernest Bourguet's 1847 cafe performance lawsuit, and Voltaire's publishing maneuvers to contextualize how copyright frameworks struggle with technological shifts. Toni acts primarily as a supportive sounding board.15:46–21:04 · The hosts pushing back 4/10 Attribution, Cultural Originality, and Collective Human Knowledge When Toni shares her experiment asking ChatGPT who owns its outputs, Benedict promptly intervenes with pushback, pointing out the secondary fallacy that the model cannot provide factual legal answers because it merely predicts text from training data. He expands on cultural shifts in attribution from Aztec sculptors to modern art.21:06–25:02 · The hosts pushing back 3/10 Structural Disruption of Entertainment and Revenue Models Benedict explains entertainment industry contract shifts, contrasting traditional syndication windfalls like Jerry Seinfeld's with one-off Netflix streaming payouts and Scarlett Johansson's Disney lawsuit. Toni highlights the tension between treating AI training as collective human knowledge and the economic necessity of compensating individual creators.25:03–29:31 · The hosts pushing back 2/10 Policy Objectives, Compensation Schemes, and Valuing Originality Benedict breaks down the mechanics and perverse incentives of pro-rata streaming royalty pools on platforms like Spotify, contrasting classical tracks with white noise spam and hip-hop sampling rules. Both conclude collaboratively that existing copyright concepts fail to map onto generative AI.

speaking balance: gold is the hosts, purple is the guest (3 minute bins)

0:00 · the hosts 84% · guest 16%0:00 · the hosts 84% · guest 16%3:00 · the hosts 84.1% · guest 15.9%3:00 · the hosts 84.1% · guest 15.9%6:00 · the hosts 100% · guest 0%6:00 · the hosts 100% · guest 0%9:00 · the hosts 65.9% · guest 34.1%9:00 · the hosts 65.9% · guest 34.1%12:00 · the hosts 96.7% · guest 3.3%12:00 · the hosts 96.7% · guest 3.3%15:00 · the hosts 73.1% · guest 26.9%15:00 · the hosts 73.1% · guest 26.9%18:00 · the hosts 77.7% · guest 22.3%18:00 · the hosts 77.7% · guest 22.3%21:00 · the hosts 80% · guest 20%21:00 · the hosts 80% · guest 20%24:00 · the hosts 90.3% · guest 9.7%24:00 · the hosts 90.3% · guest 9.7%27:00 · the hosts 76.5% · guest 23.5%27:00 · the hosts 76.5% · guest 23.5%
Sharpest disagreement ▶ 22:52 Toni highlights the ethical tension in collective knowledge framing

Toni directly challenges the friction between Benedict's romantic framing of collective knowledge and the harsh economic reality that individual creators are going uncompensated while their work is ingested.

 Hardest push from the hosts ▶ 19:21 Benedict rejects relying on ChatGPT output for legal determination

Benedict immediately halts Toni's point about asking ChatGPT about copyright ownership by identifying the logical fallacy of expecting an autoregressive language model to understand legal reality.

 Biggest teaching moment ▶ 9:46 Toni reframes prompt engineering as standalone creative art form

Toni sharpens Benedict's extended technical anecdote into a clear artistic argument, establishing that the skill and intentionality behind prompt construction constitutes the actual modern creative medium.

 The host holds their own ▶ 12:05 Benedict contextualizes copyright history with specific 18th/19th century cases

Benedict demonstrates high domain fluency by detailing the 1847 Ernest Bourguet cafe concert dispute alongside Voltaire's anti-piracy syndication strategy to show how copyright norms have continually evolved.
the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Tools Versus Artistic Vision in Photography and History 8011 Benedict dominates the discussion by drawing detailed analogies between AI generation and the history of photography, invoking Henri Cartier-Bresson's Leica and the evolution of Impressionism to argue that tool access does not equate to artistic vision. Toni readily agrees and reinforces the historical framing.
Prompt Engineering and Multi-Layered Style Synthesis 8112 Benedict demonstrates deep practical knowledge of generative models and style cues, detailing specific Midjourney and ChatGPT experiments using film stocks like Fujifilm Pro 400H and French 1960s sports cars. Toni enthusiastically validates this by reframing prompt engineering as its own distinct creative craft.
Legal Precedents and the Evolution of Copyright Concepts 9112 Benedict cites numerous legal and historical precedents, referencing the monkey selfie dispute, Ernest Bourguet's 1847 cafe performance lawsuit, and Voltaire's publishing maneuvers to contextualize how copyright frameworks struggle with technological shifts. Toni acts primarily as a supportive sounding board.
Attribution, Cultural Originality, and Collective Human Knowledge 8124 When Toni shares her experiment asking ChatGPT who owns its outputs, Benedict promptly intervenes with pushback, pointing out the secondary fallacy that the model cannot provide factual legal answers because it merely predicts text from training data. He expands on cultural shifts in attribution from Aztec sculptors to modern art.
Structural Disruption of Entertainment and Revenue Models 8123 Benedict explains entertainment industry contract shifts, contrasting traditional syndication windfalls like Jerry Seinfeld's with one-off Netflix streaming payouts and Scarlett Johansson's Disney lawsuit. Toni highlights the tension between treating AI training as collective human knowledge and the economic necessity of compensating individual creators.
Policy Objectives, Compensation Schemes, and Valuing Originality 8112 Benedict breaks down the mechanics and perverse incentives of pro-rata streaming royalty pools on platforms like Spotify, contrasting classical tracks with white noise spam and hip-hop sampling rules. Both conclude collaboratively that existing copyright concepts fail to map onto generative AI.

Statements from this episode (12)

Opinion
Evans: Generative AI is moving faster than the web or smartphones did
“This is all happening so much more quickly than previous changes like the web or iPhones did.”
Benedict Evans Apr 2, 2023 ▶ 0:50
Insight
Evans: Generative ML works by running pattern recognition models backwards
“What generative ML does is run that model backwards and make more of that pattern.”
Benedict Evans Apr 2, 2023 ▶ 1:21
Assertion Supported
Evans: Stable Diffusion occasionally reproduces Getty Images watermarks
“There was a thing in stable diffusion where every now and then you'd see like something that like a Getty image is watermark appear on the image.”
Benedict Evans Apr 2, 2023 ▶ 2:08
Insight
Cairn-Brown: Prompt engineering creativity determines generative AI output quality
“We're all being given the same tools, but you're, that creativity behind the prompt engineering is what's going to make the absolute difference here, which is fascinating.”
Toni Cairn-Brown Apr 2, 2023 ▶ 10:16
Assertion Supported
Cairn-Brown: Works lacking human creative input are not copyrightable
“Without the human creative input, that work isn't entitled to copyright protection.”
Toni Cairn-Brown Apr 2, 2023 ▶ 10:59
Insight
Evans: GPT aims for AGI by ingesting all collective human text
“The theory that all of this kind of is on a track to AGI in a sense is that you can create AGI out of the collective Knowledge of humanity, which is sort of what GPT is trying to do. It's just to give it all the texts that there is on the internet and like int…”
Benedict Evans Apr 2, 2023 ▶ 20:09
Insight
Evans: Synthesizing styles from millions of images isn't copyright infringement; direct copying is
“If I've now looked at millions of pictures of wizards and I draw something, do I have to pay them? Like, no. If I copy your thing, do I have to pay you? Well, yes.”
Benedict Evans Apr 2, 2023 ▶ 20:53
Insight
Evans: Streaming eliminates syndication lottery tickets for creators
“It used to be that when you did a TV show, if it became a hit, then it would go into syndication, and so you had kind of a lottery ticket, and this is why Jerry Seinfeld is worth half a billion dollars, because it's all the syndication money, but a Netflix sho…”
Benedict Evans Apr 2, 2023 ▶ 22:17
Insight
Evans: Existing copyright frameworks fail to map onto generative AI
“You can't just say, well, this is what copyright law says. And this is a derivative work, and that's fair use. That's not really what's, that doesn't map very well to what's happening here. It's kind of like taking live performance rights from the early 19th c…”
Benedict Evans Apr 2, 2023 ▶ 23:38
Assertion Supported
Evans: Music revenue rebounded to two-thirds of inflation-adjusted 2000 peak
“And we are now back at the point that revenue from the recorded music industry is sort of two thirds of what it was in 2020 adjusted for inflation, sorry, in 2000 adjusted for inflation and is heading back upwards.”
Benedict Evans Apr 2, 2023 ▶ 25:22
Prediction Not checkable as stated
Evans: Generative AI is going to demolish the porn industry
“It's also going to happen to porn. It's just going to demolish the porn industry. Because again, there's vast amounts of training data.”
Benedict Evans Apr 2, 2023 ▶ 27:01
Prediction Held up
Evans: Running image generation models locally on phones is months or years away
“You know, we're going, quickly going to go to a model in which you can run stable diffusion on the journey yourself on your computer or on your phone. That's like months or years away. It's not decades away. That's close.”
Benedict Evans Apr 2, 2023 ▶ 27:22
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 100 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more