What a brilliantly written and argued article, Cameron. I feel smarter for having read it, and you make some very valid arguments. The only bit I don't fully agree with is to start this America only. I think the EU could be well positioned to do this simultaneously.
I’d like to think a US-EU partnership is possible from the start, but unfortunately a pretty fractious time in our relationship (bc of this administration mostly)
I don't think that should stand the design of proper regulation in the way. This could be a chance for the EU to pivot from reactive and retroactive legislation, to future-oriented legislation, that also leaves room for investing in further research.
The vast bulk of what these labs did was drudgery: improving the cost and reliability of switches, not chasing genius. It was a repair process, not a research process. The patents were what fell out of it, not what it was for.
Which is worth sitting with, because it means "capture" wasn't extraction from a genius-producing machine. It was extraction from an optimization process that occasionally produced something patentable. Whoever owned the optimization process owned the surplus, genius or not.
Amazing read, am highly piqued by the proposed policy measure. I hope it somehow hapoens in the near future.
I also think that Proudhon’s notion of the “collective force” also deserves a place in the discussion. What we humans have aggregated on the internet as a result of our interactions is much much greater than the mere sum of the parts of the interaction. The fertility of a paddy field is not only in the sum of the paddy planted in the flooded waters, but also their interactions and interconnections (not only amongst themselves but with other species: fish and aquatic creatures help develop fertility more!)
Similarly, the real value of the internet is hence not only what happens in each of the sites (web pages etc) but what happens *between* them too (in terms of discussion, interaction, callback, reference, retweet, restack….list goes on). Hypertext for a reason. I cannot think right now of a suitable way to account for this monetarily, but this must be addressed in some way.
I definitely agree. It’d be an interesting thread to explore the dual formal + informal “collective force” dynamics with the internet. It’s very much a “parts become much more valuable collectively and iteratively rather than individually” situation.
1. It might be useful to think about it in terms of a public surplus, not just a commons. Most of what we think of as a commons is something claimed (field, stream and sky) not something created like a model or a network. But how to deal with the surplus of a social unit has a huge literature. Political science largely starts at economics which begins with surplus.
2. Surplus, or the distribution of it, is one of the primary reasons we have government, so we as a society have to determine a form of government that can make the best reasonable disposition of the surplus. This is the distribution function of the royalty, and again, not novel.
3. It is unlikely that a flat distribution would be smart in a capital-driven economy, simply because it would just serve to raise prices, inevitably returning the distribution right back to the same class of wealthy people who will benefit from the models under their current capital structure.
4. The interesting plan would be to use the proceeds to lever open the commons in a way that disproportionately benefitted creators, which would be to lower the cost of living, largely by reducing the rent burden on both home and livelihood. We do not contribute equally to the intellectual patrimony that is being exploited - and those who write and do math and philosophy and those who know how to rebuild a transmission are precisely the people who need a lower cost of living.
No offense, but well-paid executives are likely contributing just about zero to our intellectual asset base.
5. This logic can be applied to every extractive intermediary, including food delivery, ride share, health providers, Facebook, Shopify and eBay. Earnings are capped at a percent of revenue, fees are capped at a percent of transaction, and taxes are used to reduce the cost of existence for the people actually creating the value (drivers, nurses).
—-
The trillions of dollars being spent by Russia to defeat Ukraine or Anthropic and the data center investors to own access to hyperscaled intellect on tap are the same - a massive bid for dominance over resources. Now is the time to put the structures in place to ensure that we are not the immiserated subjects.
Thanks for taking the time to read and def agree that public surplus is a useful framing to add to the discussion! On the allocation of value, I generally agree with you there are may be more socially productive ways to split if we can figure out good distribution models. However, that is a big if and so I ultimately went with the universal royalty mechanism for two reasons. 1 - I could not come up with a strict definition of "creator" that I felt confident in both spiritually and legally as its v hard to split the value of creation vs interaction vs participation and 2 - I wanted to anchor the essay in something simple and tangible enough that I think actually has a chance to pass into law within the next decade.
I do think using some of the proceeds to replenish the internet knowledge commons is worth developing alongside what I'm proposing. Anyway, I genuinely appreciate the thought with which you approached your reply
Yea distillation is a really fertile ground from arguing about what is valuable and who owns it. I think we’ll see a lot of footgun arguments made over the next few years on this topic haha
Brilliant essay, up to the point of suggesting a "solution" based on money:
Imagine you hadn't fallen into that trap: I, the reader, could then have imagined that a "solution" did not just belong to USA and US-national legalities, but was in fact a necessity belonging to the world at large - possibly including sanctions against countries doing nothing to equalize data as belonging to humanity/every creator on the internet. Which is just a preposterous thought of my inner 5-yr old.
These guys, the tech moguls and their conscious constituents know, that sucking up to the presidential environment in USA is equal to USA being the only country in the world that matters.
Solution: Go in the opposite direction. Keep reminding everyone that the world is round and that ALL thoughts matter. Cough.
For what it’s worth I am writing this essay series as a group of nested policy proposals designed to directly persuade regulators with practical recommendations implementable within the next decade or so.
I have at least 5 more essays in the works that will go into this and other issues more deeply
The short version is that morally i'm right there with you, but practically I think it will end up being flagship US regulatory action benefitting its citizenry first and global remedies following over time.
I suspect some countries will claim similar chains of logic as mine in the near future, but I can't come up with a practical way for them to tax/enforce against an American frontier lab without just driving them out of the country or it becoming a pass through tax onto model users (which probably ends up just a tax on corporations). Maybe there's an elegant solution? Open to ideas.
The river delta analogy is dangerous and should be avoided. I've read countless arguments where the same subtle mistake is made over and over again. It's so tempting to conflate copying with moving, but it's so fundamentally different.
With moving, you take my thing, which I then no longer have. With copying, we now both have a copy. Just like moving, this can be incredibly undesirable.
The comparison with libraries is solid though. They lend you a book with implied and possibly formalized usage expectations. You borrow the book, read it, then bring it back. The same goes for buying a book. You buy it, then you read it, then you could give it to others. But you don't copy it. This is an almost unspoken rule. The cost of copying is close to the cost of buying the book, so people don't do it.
I agree with your assessment when considering the text data layer by itself. It’s certainly not taking the tweets/essays/comments/etc in the sense of they disappear after ingestion by LLMs, but when we consider the knowledge generation ecosystem as a whole system then I do believe we are creating the conditions to effectively “move” the text data layer into a worse state than it is today via raw volume of low quality AI output and destroying the incentives to contribute.
I see. Maybe theft is an accurate description then after all. Those whose content is stolen will eventually stop contributing this way. The question is, what will they be doing the future. Creating content exclusively for AI companies? Some kind of paywalling/DRM? No doubt a lot will change.
What a brilliantly written and argued article, Cameron. I feel smarter for having read it, and you make some very valid arguments. The only bit I don't fully agree with is to start this America only. I think the EU could be well positioned to do this simultaneously.
Thank you for reading, Arjan!
I’d like to think a US-EU partnership is possible from the start, but unfortunately a pretty fractious time in our relationship (bc of this administration mostly)
I don't think that should stand the design of proper regulation in the way. This could be a chance for the EU to pivot from reactive and retroactive legislation, to future-oriented legislation, that also leaves room for investing in further research.
I agree! I just hope the frontier labs don’t retaliate by exiting the market if so.
The vast bulk of what these labs did was drudgery: improving the cost and reliability of switches, not chasing genius. It was a repair process, not a research process. The patents were what fell out of it, not what it was for.
Which is worth sitting with, because it means "capture" wasn't extraction from a genius-producing machine. It was extraction from an optimization process that occasionally produced something patentable. Whoever owned the optimization process owned the surplus, genius or not.
Amazing read, am highly piqued by the proposed policy measure. I hope it somehow hapoens in the near future.
I also think that Proudhon’s notion of the “collective force” also deserves a place in the discussion. What we humans have aggregated on the internet as a result of our interactions is much much greater than the mere sum of the parts of the interaction. The fertility of a paddy field is not only in the sum of the paddy planted in the flooded waters, but also their interactions and interconnections (not only amongst themselves but with other species: fish and aquatic creatures help develop fertility more!)
Similarly, the real value of the internet is hence not only what happens in each of the sites (web pages etc) but what happens *between* them too (in terms of discussion, interaction, callback, reference, retweet, restack….list goes on). Hypertext for a reason. I cannot think right now of a suitable way to account for this monetarily, but this must be addressed in some way.
Thanks for taking the time to read and reply!
I definitely agree. It’d be an interesting thread to explore the dual formal + informal “collective force” dynamics with the internet. It’s very much a “parts become much more valuable collectively and iteratively rather than individually” situation.
Props for mentioning a club good and the internet in the same article!
https://www.linkedin.com/pulse/club-good-protocol-s-mike-dierken-hqpbc
Appreciate it and thanks for reading 🫡
This is a fantastic article! A few observations:
1. It might be useful to think about it in terms of a public surplus, not just a commons. Most of what we think of as a commons is something claimed (field, stream and sky) not something created like a model or a network. But how to deal with the surplus of a social unit has a huge literature. Political science largely starts at economics which begins with surplus.
2. Surplus, or the distribution of it, is one of the primary reasons we have government, so we as a society have to determine a form of government that can make the best reasonable disposition of the surplus. This is the distribution function of the royalty, and again, not novel.
3. It is unlikely that a flat distribution would be smart in a capital-driven economy, simply because it would just serve to raise prices, inevitably returning the distribution right back to the same class of wealthy people who will benefit from the models under their current capital structure.
4. The interesting plan would be to use the proceeds to lever open the commons in a way that disproportionately benefitted creators, which would be to lower the cost of living, largely by reducing the rent burden on both home and livelihood. We do not contribute equally to the intellectual patrimony that is being exploited - and those who write and do math and philosophy and those who know how to rebuild a transmission are precisely the people who need a lower cost of living.
No offense, but well-paid executives are likely contributing just about zero to our intellectual asset base.
5. This logic can be applied to every extractive intermediary, including food delivery, ride share, health providers, Facebook, Shopify and eBay. Earnings are capped at a percent of revenue, fees are capped at a percent of transaction, and taxes are used to reduce the cost of existence for the people actually creating the value (drivers, nurses).
—-
The trillions of dollars being spent by Russia to defeat Ukraine or Anthropic and the data center investors to own access to hyperscaled intellect on tap are the same - a massive bid for dominance over resources. Now is the time to put the structures in place to ensure that we are not the immiserated subjects.
Thanks for taking the time to read and def agree that public surplus is a useful framing to add to the discussion! On the allocation of value, I generally agree with you there are may be more socially productive ways to split if we can figure out good distribution models. However, that is a big if and so I ultimately went with the universal royalty mechanism for two reasons. 1 - I could not come up with a strict definition of "creator" that I felt confident in both spiritually and legally as its v hard to split the value of creation vs interaction vs participation and 2 - I wanted to anchor the essay in something simple and tangible enough that I think actually has a chance to pass into law within the next decade.
I do think using some of the proceeds to replenish the internet knowledge commons is worth developing alongside what I'm proposing. Anyway, I genuinely appreciate the thought with which you approached your reply
It’s an interesting project. And creativity is inherently communal ;)
I wrote something quick that was looking at "distillation attacks" a charge leveled against open weights models that taking your argument and looking at it from there: https://nathangrouns.substack.com/p/the-transformative-shield?utm_source=share&utm_medium=android&r=1stfm
Yea distillation is a really fertile ground from arguing about what is valuable and who owns it. I think we’ll see a lot of footgun arguments made over the next few years on this topic haha
Points! for referencing Elinor Ostrom.
🫡
What an informative, thought-provoking, and smooth read. Thank you!
Thank you kindly and appreciate your thoughts 🫡
Great job.
Thank you kindly
This is excellent! Belongs in the LRB.
That’s very kind. Thank you for reading!
This slaps. Really glad I read it
Thanks Spencer! I appreciate you taking the time to check it out.
Brilliant essay, up to the point of suggesting a "solution" based on money:
Imagine you hadn't fallen into that trap: I, the reader, could then have imagined that a "solution" did not just belong to USA and US-national legalities, but was in fact a necessity belonging to the world at large - possibly including sanctions against countries doing nothing to equalize data as belonging to humanity/every creator on the internet. Which is just a preposterous thought of my inner 5-yr old.
These guys, the tech moguls and their conscious constituents know, that sucking up to the presidential environment in USA is equal to USA being the only country in the world that matters.
Solution: Go in the opposite direction. Keep reminding everyone that the world is round and that ALL thoughts matter. Cough.
best, kk
Thanks for reading and appreciate your feedback
For what it’s worth I am writing this essay series as a group of nested policy proposals designed to directly persuade regulators with practical recommendations implementable within the next decade or so.
I have at least 5 more essays in the works that will go into this and other issues more deeply
The short version is that morally i'm right there with you, but practically I think it will end up being flagship US regulatory action benefitting its citizenry first and global remedies following over time.
I suspect some countries will claim similar chains of logic as mine in the near future, but I can't come up with a practical way for them to tax/enforce against an American frontier lab without just driving them out of the country or it becoming a pass through tax onto model users (which probably ends up just a tax on corporations). Maybe there's an elegant solution? Open to ideas.
The river delta analogy is dangerous and should be avoided. I've read countless arguments where the same subtle mistake is made over and over again. It's so tempting to conflate copying with moving, but it's so fundamentally different.
With moving, you take my thing, which I then no longer have. With copying, we now both have a copy. Just like moving, this can be incredibly undesirable.
The comparison with libraries is solid though. They lend you a book with implied and possibly formalized usage expectations. You borrow the book, read it, then bring it back. The same goes for buying a book. You buy it, then you read it, then you could give it to others. But you don't copy it. This is an almost unspoken rule. The cost of copying is close to the cost of buying the book, so people don't do it.
I agree with your assessment when considering the text data layer by itself. It’s certainly not taking the tweets/essays/comments/etc in the sense of they disappear after ingestion by LLMs, but when we consider the knowledge generation ecosystem as a whole system then I do believe we are creating the conditions to effectively “move” the text data layer into a worse state than it is today via raw volume of low quality AI output and destroying the incentives to contribute.
I see. Maybe theft is an accurate description then after all. Those whose content is stolen will eventually stop contributing this way. The question is, what will they be doing the future. Creating content exclusively for AI companies? Some kind of paywalling/DRM? No doubt a lot will change.
It’s a tough issue to wrangle for sure. Thanks for reading and participating in the discussion!