Apr 12, 2024 · 51m · more-or-less
#42: OpenAI, IP Theft and a New Incentive Structure for the Internet · More or Less Podcast
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the 'More or Less' podcast, co-hosts Jessica Lessin, Dave Morin, Brit Morin, and Sam Lessin critically analyze the legal and ethical fallout of generative AI data scraping, examine the collapse of digital publishing economics, and evaluate emerging tech trends from Google Vids to the Apple Vision Pro.
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 99.8% of the talking time here. How this is scored →
The hosts as informed peer 6.6 Guest teaching 2.6 Guest disagreement 2.7 The hosts pushing back 3.1
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 25:43 Sam dismisses media executive talent
Sam aggressively rejects standard sympathetic narratives about publishers, stating flatly that media business talent consists largely of B-players who are fundamentally not very good.
Hardest push from the hosts ▶ 25:03 Jessica refutes guaranteed link compromisesWhen Brit suggests that AI platforms could placate publishers by guaranteeing referral links, Jessica immediately shuts down the premise by pointing out that Facebook guaranteed traffic to BuzzFeed News right before BuzzFeed News went bankrupt.
Biggest teaching moment ▶ 18:44 Sam's structural analysis of model training valueSam educates the room on the mechanics of LLM training economics by cleanly dividing ingested data into commodity scale, protectable intellectual property, and real-time news.
The host holds their own ▶ 41:20 Jessica relays market-moving IPO thresholdsJessica leverages exclusive reporting from her interviews with top asset managers at the NYSE to explain how ETF structures have pushed the viable public debut baseline up to five to ten billion dollars.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Pre-Show Check and Appearance Banter | 3 | 1 | 1 | 2 | The episode opens with casual pre-show banter covering Jessica's makeup, Sam traveling with kettlebells, and Brit citing speed-dating trends on Eventbrite. The exchange is lighthearted and conversational with no real substantive debate. | |
| OpenAI YouTube Scraping and Fair Use Scrutiny | 7 | 3 | 2 | 3 | Dave and Jessica unpack the New York Times investigation into OpenAI scraping YouTube transcripts via Whisper to train models under fair use claims. The group shows high domain literacy regarding copyright doctrine and model training practices. | |
| Regulatory Front-Running and the Tech Ethics Choice | 7 | 3 | 3 | 4 | Sam frames tech practices around front-running regulatory frameworks like Uber and Netscape did, while Dave argues AI founders actively faced an explicit moral choice between licensing and theft. Jessica adds nuance with historical corporate parallels like Zenefits. | |
| Publisher Concessions and the Three Data Categories | 8 | 4 | 3 | 3 | Jessica details publisher desperation in New York settling for minor licensing sums, while Sam introduces a precise taxonomy of LLM training data into pure scale, IP, and topical news. Dave sharpens Sam's scale premise using Meta's internal findings on dialogue. | |
| The Synthetic Data Fallacy and Human Essentiality | 7 | 2 | 3 | 3 | Dave highlights the lack of scientific proof that synthetic AI data can sustainably improve frontier models, and Sam backs him up by describing synthetic degradation as photocopies of photocopies losing resolution to reality. | |
| Search Traffic Myths and Digital Publishing Pitfalls | 8 | 3 | 4 | 5 | Brit inquires whether guaranteed referral links could save publishers, but Jessica forcefully pushes back by citing the demise of BuzzFeed News after relying on platform traffic promises. Sam harshly critiques media executive talent and defeatist attitudes. | |
| Digital Feudalism and the Crisis of Democracy | 7 | 3 | 4 | 4 | Sam articulates an economic thesis where common knowledge clears at zero price while proprietary reality commands extreme fees, predicting a return to information feudalism. Jessica pushes on the downstream societal risks to democratic systems. | |
| Sci-Fi Parallels: Three-Body Problem and Wallfacers | 5 | 2 | 2 | 2 | The panel pivots to discussing The Three-Body Problem, drawing an analogy between Wallfacers keeping strategic thoughts unexpressed and humans needing to protect un-indexed cognitive value from AI scraping. | |
| Private Capital Conference Insights and IPO Realities | 8 | 2 | 2 | 2 | Jessica reports back from The Information's Private Capital Conference, relaying firsthand insights from Coatue's Philippe Laffont, General Atlantic's Anton Levy, and Mubadala's Ibrahim Ajami regarding sovereign wealth AI capex and rising IPO baseline thresholds. | |
| West Coast Check-in: Spring Break and Ski Week | 6 | 3 | 3 | 3 | The hosts discuss Google's new enterprise video tool Vids, drawing parallels to Google Wave which Brit worked near while at the company. Sam calculates an exaggerated cost-per-minute metric for the Apple Vision Pro. |
Statements from this episode (20)
Morin: Scraping internet content for AI under fair use is egregious
“Going out and saying that you're going to steal the entire internet, steal all the content and then generate new content and claim fair use just seems, it seems beyond the pale to me and a little bit egregious.”
Lessin: AI fair use lawsuits will take a decade to resolve
“There are court cases that will take a decade, unfortunately, and they're saying it's fair use, which is a part of copyright law when you're taking copyright content and then just paraphrasing or taking out a sentence or two.”
Morin: Multiple AI teams knowingly chose data theft over paying publishers
“Here's a bunch of teams, not just one, but many AI teams standing on the precipice of making a moral decision. And they knew they either needed to pay for it or steal it. And some of the teams chose to steal it instead of paying for it.”
Lessin: The New York Times stands alone in litigating against OpenAI
“The only one really left standing up is the New York Times, which has a big lawsuit against OpenAI, and others that, like News Corp, my former employer, was talking a really tough game, and then did a one-eighty in their tone, and is now like, we're gonna nego…”
Lessin: Publisher lawsuits will not put OpenAI out of business
“Will it put OpenAI out of business? No, I don't think so, because like, They can fight these lawsuits. You know, there's not clear, like, other kind of regulatory stuff breathing down their neck”
Lessin: LLMs break internet economics by stealing without returning ecosystem value
“The problem with these kind of LLM things is they break economics because they steal, but they give nothing back. And so even like a cents on the dollar licensing deal or whatever, that's not investing in the ecosystem of knowledge of information globally, rig…”
Morin: Science has not proven LLMs can improve on synthetic data
“Science has not yet proven that you can create higher quality LLMs using AI generated data.”
Lessin: AI is a deeply communist technology against the free market
“AI is like a deeply communist technology, right? It is against the free market. It's again, like in, in every way, shape, or form.”
Lessin: Micro-payments for human data input to AI won't work
“If you're like, oh, I'm going to text 10,000 people to get the right answer. Like, you're not going to get paid on that. Like, it's not actually an economic model that's going to work in practice, although it's a nice idea in theory.”
Lessin: Search referral traffic was an economic illusion for journalism
“The traffic back from these links wasn't the economic engine to support the high-quality journalism they all wanted to do, and in fact, in many cases, it distracted them from doing that.”
Lessin: Most media business talent are 'B-players' and not talented
“You have the media business talent, which with a few exceptions, I have to say that because there are some people I like, they're bees, right? They're not very talented.”
Morin: AI will stall without economic incentives for content creators
“We have to figure out how to create an economic incentive structure that incentivizes both sides. Otherwise, we end up at this kind of, I think, mediocre middle, and the technology doesn't reach its potential.”
Sam Lessin: Widely Known Information Has an Economic Clearing Price of Zero
“There's information that lots and lots of people know. The clearing price for that information is precisely zero. Right? Like if lots of people know something, it doesn't matter how it gets on the internet or if open AI bids one cent for someone to tell them a…”
Lessin: Media will bifurcate into AI slop and expensive elite newsletters
“The future is going to be almost like, this is what happened after the printing press is the news got super scummy and shitty, and all the elites retreated to super high price Italian private newsletters to know what was actually going on in the world. That's …”
Dave Morin: ByteDance investors don't understand influence because they never built platforms
“That's also because these guys are capital allocators. And not builders of social networks. They've never built social networks, and then they don't know what is possible in terms of influence.”
Lessin: ByteDance investors ignore national security risks for massive financial incentives
“Oh, they're pretty smart guys. They know exactly what's up. They, they're trading between what is obviously true to any American versus their enormous financial incentive, and empathy is a great word for that.”