What’s up folks. Just wanted to say that I appreciate the love that everyone is giving, whether that is via Twitter or email or text, it is motivating to keep me posting on time every week. We’re hitting a nice stride here, and as I keep telling you, we are still “pre-viral” :). I think it’s important to have this sort of blinding optimism about the future, that things will always work out, that humanity will always find a way to prevail, and that as long as you are doing something, pushing the ball forward, good things will continue to happen.
To put my own spin on an FDR classic: There is no stress but stress itself.
If you are constantly looking forward, while staying in the moment, you realize that all change is incremental. There is no need to ever stress about progress. The massive leaps in technology and evolution are not massive leaps, but the accumulation of microscopic goals being achieved over time. Case in point - my newsletter. I will share my subscriber numbers when we reach something to be proud of, but just know that you are part of a community that is growing exponentially over time:
I hope the corpus of my content circles around a worldview that people can utilize agnostically amongst sectors. Stay objective and optimistic. DYOR. BYOB. Lean into the trends but don’t get sucked into the hype. You can be pro-Crypto and anti-speculation. You can be high on AI and skeptical of AGI. You can be registered to one political party and vote for the other party. Nothing is binary, so community should not be bifurcated. Human society is an infinitely-expanding 4D Venn Diagram of our interconnected alignments.
Anyways, enough preamble. Here is this week’s topic.
Infringement Binging
OK so by now everyone is fully in AI mode, right? If you aren’t experimenting with it, you’re falling behind. Whether you believe AI is going to take over the world, take over your job, or just make your life slightly easier, there is no denying that there is something here right now. I am using Midjourney for all of my images. There is a new goldrush of AI companies being built off of the OpenAI API key. ChatGPT surpassed 100 million users, making it one of the fastest growing consumer applications of all time. Even the most ardent contrarian has to have their ears perked up right now.
There is a lot of hype around the space. Often when this happens, people get out over their skis in predicting all of the future ramifications that are 2, 3, 4 levels away from the current status. Things don’t work so deterministically in reality, however. There are always unforeseen obstacles that change the course of history. Progress is not a straight line, but helical. If AI is ever going to get anywhere, the immediate known obstacles need to be confronted, so I think it is important to talk about what is going on with AI right now in the context of the law.
Corpus of Content
It has to be noted that everything generative AI (that is, things that generate text, images, etc) is coming from the set of content that the AI is trained on. ChatGPT is trained on the open internet, while other products might be trained on more localized or concentrated data sets. The point is, the content these AI models are being trained on are definitively someone else’s work. If I want to spit out an image of Van Gogh’s Starry Night in the style of Monet, I (the user) am quite literally stealing a single piece of art, and remixing it with the AI’s interpretation off of the corpus of another artist’s content. Artistic integrity aside (which I have discussed in prior articles), the question now is about legality and attribution.
Does Van Gogh get credit for this piece of artwork? If so, how much? More than Monet? More than me, the prompter? More than the AI model itself? How do we distribute that value?
More pointedly, am I allowed to use this piece of art to create new art?
AI is all about the content it is trained on, and the most immediate debate is going to be around what is or is not allowed into a training set. Like Napster though, it will be impossible for creators to stop the massive onslaught of change that is coming. If it exists on the internet, it is available to anyone who can Save As. Challenges may be tossed up in court as people who have spent massive amounts of money collecting content are suddenly sucked dry as the business physics shift around this. What is the definition of ownership? What is proof of ownership? What does ownership mean when open source ethos are applied to creative licenses? NFTs start to make a little more sense..
Fair Use
We have copyright laws in America to protect creators from theft. Going back to 1790 when it was first established, the Founding Fathers gave 14 years of protection to copyright. It wasn’t until the mid 1970s that extended it to "Either 75 years or the life of the author plus 50 years", and the late 1990s that it became “120 years, or the life of the author plus 70 years, whichever ends earlier.”
That is pretty powerful protection, and it allowed brands like Disney to become the behemoth they are today. If anyone could create their own Mickey Mouse cartoon and market it as such, Disney could just slap an infringement suit on them and shoo them away like a fly. Because how easy would it be to tap into the brand equity of Mickey Mouse and monetize my own version of his character? You're stealing the brand equity, not the artwork.
Derivative works are allowed of course, and they receive the same protection as the original artwork. I can imitate Van Gogh and have my new piece of art protected as it’s own entity. Does Van Gogh get credit? Only through cultural reinforcement of his existing value, as the value of my new art only exists because of his existing/relevant/acknowledged value. Also, no money is exchanging hands so it’s all good (for now).
We also have Fair Use laws which help facilitate such derivative works. All art influences other art, so it would be really hard to create if no allusions could be drawn to historical greats, or even contemporary peers. “No Ideas Original” as Nas said. Fair Use basically says that “hey, we recognize that this copyrighted material is cool and you should be able to use it for reasonable purposes as long as you don’t try to make money off of it or hurt the nature of the original work in any way”. In a country full of weird and wonky laws, that one is pretty tight.
What happens when money does get involved though? Copyrighted material is more valuable, which is why it is protected in the first place, so there is more to be gained by stealing it over some other random, unprotected piece of art. But also, stealing some unprotected piece of art could rob that artist of the vertical mobility associated with increased attention. Countless examples of this have happened throughout history. X artist was the first to do Y, but Z artist took the ball and ran with it for massive success. Sorry X, all is fair in art and money..
Attribution
The digital world is mostly about posting content and selling products. Not all content is monetizable individually, but all content is monetizable in the aggregate. We know this from the Facebook data. No one cares about the post you made showcasing your limp homemade waffles, but it is notable to the network that Cousin Suzy liked your limp waffle photo, so that when Eggo starts marketing to a waffle audience, they can include Suzy and make reasonable assumptions about Suzy’s expanded network as well. Should you get paid for your limp waffle post because it contributed to the overall value of the network? If social networks become decentralized, the advertising revenue that gets inevitably piped through can be pushed to the fringe user, instead of being scraped off the top.
Which leads back (once again) to the Original Sin of the Internet, which was not allowing a native payment system, which forced advertising to be the native currency, and attention its non-renewable resource.
We have to recognize that advertising revenue is like gravity, it’s as close to a physical law of the internet that we have. As long as there is content, there will be monetization through attention.
The reason digital marketing works so well is because of attribution. Eggo runs an ad to a waffle lookalike audience, and your cousin Suzy’s cousin Lucy gets targeted and buys a big ‘ol box of crispy Eggos and now Eggo knows that the lookalike audience based on cousin Suzy (the limp waffle photo lover) is performing. All of this is repeated over and over, amassing billions of dollars for Big Tech, as well as advertisers. If there were no deterministic attribution models, Big Tech would not exist, full stop.
So how does this tie back to AI? Well, the generative-ness of the models are taking the entire corpus of content (copyrighted or not) and putting it in a great big blender and whipping up entirely new things. Just like Eggo can attribute a conversion to a lookalike audience, we can all attribute my Van Gogh image to Van Gogh. But what about your part in all of this? It was your limp waffle post that identified cousin Suzy as a limp waffle lover, and led to the construction of the waffle audience from which the lookalike audience was a derivative of. Surely your limp waffle is of some value in the chain here. What if instead of your limp waffle post, it was an article on the top 20 restaurants in LA? And you asked an AI search engine where should I eat dinner tonight, and it aggregated the top 10 articles about top restaurants in LA and it spat out a unique top 3 list for my review. This AI search engine is taking some monetization away from the creators of these top 20 lists and repurposing it.
Creators are either going to be compensated based on an attribution model that only the AI can provide, or they are going to be washed away in a tide of unbridled infringement binging, where copyright is uncopyrightable and content procurement is the new Wild West.
Captured attention is always going to be monetizable, and Content Creators are the magicians that conjure that attention straight out of the ether, and they will always have value. But, generative AI can dilute some of that value. There is sooo much money at stake here that creators/publishers/advertisers/algorithms/platforms/AI are all going to have to find a way to get along.
Section 230
Let’s take a quick divergence and talk about the protection that is granted to the Big Tech companies as this ties in tangentially to this topic. Section 230 recognizes the platforms as platforms, and not publishers. Therefore, Facebook is not liable for your limp waffle post should it be particularly scandalous. Facebook does however host all of these images, and by nature is making it free game for AI models to download that content into their database for generative purposes. A law could be passsd that says AI models are not allowed to touch certain things (such as social media networks), the problem is enforcing that law when these models are open source protocols, not people.
As I have predicted in previous articles, free speech is probably going to be litigated further, and with that, Section 230 has the possibility to face additional scrutiny. If the government starts to feel uncomfortable as to how “free” the speech is getting on Twitter, for example, they could start to reel in some of that protection that s230 provides. The massive exploitation of content that AI exposes, mixed with the inevitable outcry of creators/publishers, could facilitate such re-litigation from another angle. If your limp waffles go viral and you create an Eggo competitor, and then some Facebook-sanctioned AI model scrapes Facebook and repurposes your content, you’re going to be like “wtf Facebook?!”. That scenario is of course happening every single day, and users acknowledge this as part of their Terms and Conditions, but sentiments can change. Especially when things start to happen at scale.
Re-litigating old laws doesn’t always make sense in the context of actual society, and is more political in nature (see: Roe v Wade). But the current sentiment landscape formula looks something like this:
Anti-Tech + Anti-Billionaire + Anti-1A + Loss of Self-Monetization + Increased Demand for Data Privacy = Re-litigation of Section 230.
This creates a disastrous precedent in which Big Tech is liable for your content, and will thusly moderate the shit out of it (per whatever current-Government regulations are popular at that time).
This ties AI back to Crypto, as the need for decentralized social networks evolve from a nice-to-have into a necessity. Once again, I am not drawing a straight line from here to there, and any number of other precedents could be set along the way, creating an alternate path. AI is a exponential driver of change though, and should it cause an uproar, it is much easier to regulate the centralized Big Tech than an onslaught of open source data models.
CC0, Open Source, and The Future of Copyright
As I have mentioned in the past, 80-90% of code written today is open source. There seems to be an ethos among programmers that the value is not in of the code itself, but what you are able to execute with it. Additionally, by sharing this information with the entire world, things like GitHub and their AI copilot can proliferate, which expands the total economic pie. This is actually a quite ironic perspective, as the production code is utilized by the most staunchly competitive and capitalistic markets.
“Shopify Merchants build applications for each other. Imagine if a startup found a growth hack and just gave it away, that would never happen.” - Kaz, COO of Shopify said recently on Balaji’s new podcast.
Open source and Shopify Merchant apps are just different forms of decentralization, the ethos of the blockchain. Distributed power in many instances can increase the pie for all, rather than just a few. It all starts with letting go. One must let go of the code you write and allow it to create to the greater good. You must let go of your limp waffle photo and allow others to build businesses off of its greatness.
CC0, or “Creative Commons” is an ideology that pre-emptively surrenders all copyright ownership and dedicates those rights directly to the public domain. This is where AI and the blockchain really converge. There is all this talk about NFTs and the ownership privileges they provide in the digital realm. One could conceivable make an NFT out of your limp waffle picture, which grants that person the rights to that intellectual property, with which they could start say, a limp waffle restaurant. (this has already happened with BAYC) This is cool, but what if you don’t want such constricted use of your limp waffles? What if more value was derived from you personally if you said “no rights reserved” and you let your limp waffles fly?
These are the two competing ideas:
Protecting your IP means that only sanctioned use of your material is allowed. This narrows the scope of use, but directs it in ways that you approve of. All uses of limp waffles must adhere to strict limp waffle brand guidelines. You do this because this adherence to brand ethos is what you believe will drive the most economic value.
Not protecting your IP means that anyone can use your material for anything. Criminal organizations can wave the limp waffle flag and say that you inspired them to conduct their nefarious acts. You allow this because you believe that the broadest use of your IP will inevitably bring the recognition back to the original, and drive the most economic value.
Idea #1 is the standard practice today because there is no direct way to attribute the use of Idea #2. Marketing and Economics are built completely around Idea #1. Only through the use of an immutable ledger could you directly attribute all iterations back to your IP in Idea#2, making it ever more valuable.
AI is pushing CC0 to the forefront of the discussion without the protections of an immutable ledger. At some point, the uphill battle of infringement cases will result in the collective sigh of the creators and they will lean into whatever outcome benefits them the most economically.
This is what happened when Napster cracked the music industry open. Record companies stopped fighting with the streaming services and leaned into it, changing the monetization of the industry forever.
There are a lot of things in play here, so the calculus is hard to map out. The cost of content creation is approaching zero, but the need for authenticity is increasing. At the same time, content creators are fighting over the share of ad revenue that the platforms provide them. You could see an alternate path where platforms themselves start iterating with generative AI off of their own creator’s IP, maximizing their own profits and diluting that of their users. There are also societal (with the potential to be legislative) trends bubbling around what can and can’t be said online, which creates a cap on content. Additionally, as wealth inequality grows, new money (Tech) comes into conflict with old money (Media). There could be a flippening of new money over old, bringing about a new era of technocracy, or the old money could put their thumb on the scale (through politics/legislation) to squash the new money back into the pre-internet age. Alongside this, we are about to enter the largest demographic power shift in history as Boomers transfer $70T of wealth to younger generations. AI plays a thematic role in all of this as it democratizes code and media, influencing both old money and new.
Where does all of this end up? It’s hard to bet against the trend of technological progress. Any pushback is likely to be just a blip, especially when tech is largely deflationary and democratizing, and traditional institutions are largely inflationary and centralizing. Of course, all of this is just in America, there are many other cultures and governments that have a say in global sentiment as well! All we can do is ride the wave and push our individual goals forward.
This is an amazing big picture of the future of AI. Brian Doran has an amazing mind
Fantastic article! New big big fan