Compensating the I in AI

Human creativity is a black box. We are inspired by multiple sources and create things by merging ideas that bounce inside our heads. Meanwhile, our current copyright system compensates creators only when their ideas are used directly — when someone samples their song, copies their text, or uses an idea that is clearly theirs. A creator doesn't earn anything when his work inspires someone else because such inspiration is impossible to track.

Machines are different. Artificial Intelligent seems like a black box, but at least some of its inspiration can be traced and described. In fact, many machine creations are explicitly aimed at copying specific creators. Consider this clip from DJ David Guetta (If you can't see it in your email, open the article in your browser).

What happened here? A famous DJ used two AI tools to create a piece of a song in the style of Eminem. It is not an Emined song, only a sample that was generated to sound like him.

Will Eminem get paid for this? Maybe. Assuming he might, we are seeing the emergence of a whole new revenue stream for creators: Get paid when people name your style when they generate new content. And, as a reminder, content is not just music — it is also code, games, video, and news reporting. When I ask ChatGPT to "generate an article in the style of Scott Galloway," — should Professor Galloway get a piece of the revenue the article generates?

Strictly speaking, the answer is "yes!".

But the story doesn't end there. So far, we have discussed cases in which AI is explicitly asked to copy/resemble a specific creator. In such cases, it's obviously easy to figure out the source of inspiration (since it is named).

But AI can help us unpack inspiration in two other important ways: By keeping track of its training base and by tracing the inspiration of content that it process.

Let me explain.

Even when it is not asked to draw specific inspiration, AI is inspired by something. If I ask Stable Diffusion to create a "surreal drawing of a house," the image it generates will likely draw on paintings by Salvador Dalí, even if I don't explicitly mention Dalí in my prompt. As the user, I may not know what inspired the algorithm, but the algorithm itself does know. The algorithm knows exactly which sources it processed. It may not know which specific images inspired it, but it knows something about its source material.

Even if the algorithm doesn't know which references it used to generate an image (or text), it can discern these references. For example, looking at the picture below, an algorithm can surmise that René Magritte inspired it given the color choice, objects, and brush strokes. In this example, even a trained human can surmise Magirtte's style. But notably, machines can discern patterns in many more situations than humans.

All this is to say that AI creates cracks in the black box of creativity: It can trace the sources from which a piece was derived much better than a human can. This means AI can do so in many more situations as well as faster and cheaper.  

Why is this meaningful?

Last week, Microsoft's CEO Satya Nadella described how AI would permeate human work:

"All computer interaction is going to be mediated with an agent helping," Chief Executive Satya Nadella said at a launch event at the company's headquarters in Redmond, Washington. "We're going to have this notion of a co-pilot that's going to be there across every application."

This is where we are headed. I already see it in my own work and reliance on AI to generate and refine images, music, and text for my newsletter and podcast. Machines will constantly refine and remix our work, drawing on other people's ideas and styles. Sometimes, this will happen explicitly (we will ask the algo to do something in the manner of someone). In most cases, it will happen implicitly. But even in the latter cases, AI itself will be able to figure out what inspired our own work and which bits of other people's work we are using. Above a certain benchmark, those other people deserve to be compensated.

Do you see where this is going?

I'm envisioning a dramatic expansion of the copyrightable universe. Today, we only trace and enforce copyright for the works of a minority of notable people — people with publishers, lawyers, and trademarks. We do so because it is not practical to trace and enforce everyone else's rights. AI makes it cheap to trace the reuse and repurposing of all content all the time. This means that instead of only enforcing copyright laws over a small number of songs, books, or articles, we can now enforce usage across smaller bits of content that less prominent people generated.

Funnily enough, this is exactly the future I contemplated when I wrote about... crypto! In NFTs and The Future of Work, I noted people often associate NFTs with stupid cartoons and financial speculation. But NFTs can apply to all content, and content is not just silly cartoons:

"Content" is not just Beyonce's new song or the highlight real of yesterday's playoff game. Content is the two lines of code that an engineer somewhere just added to an open-source project. That Slack message your colleague sent you this morning is also content. And of course, the newsletter you sent to your 72 subscribers last week is also content, even if you're not a celebrity.

Can these two lines of code, Slack message, or an email that nobody read be sold just like a work of art? No, because they are not a work of art. But they are work. And so, someone deserves to get paid for them if they end up being useful.

I pointed out that machine learning can play a critical role in helping us trace and enforce copyright across a broader universe of "content":

One of the most useful things about machine learning is its ability to turn unstructured data into structured data. It can convert a photo into a list of the items that appear in it. It can summarize books. It can scan a video and tell you who's in it and create a transcript of what each participant said. It can keep track of who said what, and what ended up being repeated and used.

All this is happening now. Unsustainable as it was, the 2021 Crypto market offered us a glimpse of the future. And as I wrote in Don't Show Me the Money, crypto might not play a direct role in that future, but it does provide a breeding ground for many of the ideas that will shape it:

It's quite possible that web3, the internet of the future, will end up running on completely different technologies and protocols. Or that blockchain protocols will be abstracted away by powerful middlemen that will save users the trouble of knowing anything about the underlying technologies and protocols (see Shopify's latest NFT solution for an example). But if you want to understand what people will be doing on the internet in 5-10 years, and if you want to identify new business models before anyone else — you must pay attention to what's happening in crypto.

What's the bottom line of this story?

• It's becoming easier than ever to create content.
• It's becoming easier than ever to trace the origins of content
• Software makes it cheaper to enforce and monetize intellectual property rights across an expanding universe of "content" — from images and articles to single lines of code and specific painting styles.

This will unlock new ways for people to earn money from work. It will also enable those with the most recognizable style to scale themselves further and make more than ever before — and much more than their peers.  

In other news...

I am experimenting with a new podcast format. Instead of recording an article, I recorded a conversation about an article. My friend Zach Valenta joined me to explore how we can increase innovation and improve the way we distribute the fruits of innovation. The discussion is based on this piece from last month.

You can listen to the whole thing on Spotify, Apple Podcasts, and wherever you get your podcasts.

In other news... #2

I spoke to Jim Bianco on RealVision about the future of offices and cities. The full interview is paywalled (they're offering a $1 trial — excellent financial content for those who are into it). Below are a couple of highlights from the interview. Again, if you can't see the videos in your email, open this article in your browser.

And...

Thank you for reading. Have a great weekend.