Exploring OpenAI’s Game Theory Model for Fair Author Compensation: The Self-Publishing News Podcast with Dan Holloway

In this episode of the Alliance of Independent Authors Self-Publishing News Podcast, Dan Holloway discusses a new proposal from OpenAI that introduces cooperative game theory to determine compensation for authors whose works help train AI. This approach aims to calculate the specific contributions of individual creators to AI outputs and allocate earnings accordingly, similar to determining the mix of colors in a paint blend. Dan examines the potential of this system to fairly reward creators and its computational feasibility in revolutionizing how authors are compensated for their contributions to AI.

Thoughts or further questions on this post or any self-publishing issue?

If you’re an ALLi member, head over to the SelfPubConnect forum for support from our experienced community of indie authors, advisors, and team. Simply create an account (if you haven’t already) to request to join the forum and get going.

Non-members looking for more information can search our extensive archive of blog posts and podcast episodes packed with tips and advice at ALLi's Self-Publishing Advice Center.

Listen to Self-Publishing News: Game Theory

Don't Miss an #AskALLi Broadcast

Subscribe to our Ask ALLi podcast on iTunes, Stitcher, Player.FM, Overcast, Pocket Casts, or Spotify.

Show Notes

An Economic Solution to Copyright Challenges of AI

About the Host

Dan Holloway is a novelist, poet, and spoken word artist. He is the MC of the performance arts show The New Libertines, He competed at the National Poetry Slam final at the Royal Albert Hall. His latest collection, The Transparency of Sutures, is available on Kindle.

Read the Transcripts to Self-Publishing News: Game Theory

Dan Holloway: Hello and welcome to another Self-Publishing News podcast from an Oxford that is pretty much apocalyptically wet at the moment, which of course it is, because today is a big day of athletics here in Oxford. We have been celebrating on this May bank holiday, which is when I am recording, Roger Bannister's 70th anniversary of the first four-minute mile down at our very own Iffley Road track, and the whole day here in Oxford has been dedicated to that celebration. As we speak, we are gearing up for elite milers to come from all over the world to see if they can repeat his feat.

They've been running at the track all afternoon and this morning I was one of 1,500 people who took part in the mass mile running down Oxford High Street from my old College of Christchurch and ending at the track on Iffley Road.

I didn't, needless to say, break four minutes, I came in at a very modest seven and a half, but one that I'm nonetheless very happy with and delighted to be part of such a historic event.

First Publication of Indie Book with No-AI Warranty

Talking of historic events, this week has seen, not particularly historic event in the world of books, but we have nonetheless had a first, and that first is the first publication of a book that has what it proudly professes to be, a no AI logo, or a no AI warranty. So, this is R. R. Haywood's, Rain, the latest book in the Undead series.

R.R. Hayward, fabulous to see this indie author, always good to report on something that indie authors are doing, a very high-selling indie author, and they have jumped on the bandwagon of the backlash that is coming against the use of AI and they have put this logo on the front of the book.

Naturally original. Authentically invented. The acronym, as you can gather, works out at saying “NO AI”, and there is even a warranty that goes along with it. The warranty reads, this book was authored by Richard Haywood and not generated by any machine or artificial intelligence. This statement serves as a legally binding guarantee of authenticity and can be used to establish the true creator of this work should any doubt arise.

I'm not, I have to say, sure how saying something is true provides a way of establishing in a legally binding way that it is true, something that lawyers may want to pick over, but it is a good marketing gimmick, shall we say, and best of luck to Richard Haywood for, for doing that, which brings us on to the main news of the week.

OpenAI Publish Paper Hints at How They Will Compensate Creators

The main news of the week is also AI related, sorry, but it is really quite big news, and it relates to how authors get paid for the use of their work by AI platforms.

It is the publication of a paper from OpenAI. So, obviously OpenAI who own ChatGPT, they have published a paper, proper academic paper that hints at how they are planning to compensate creators.

So, it looks at different ways you could go about paying people for their work, and the model that it is suggesting might be a viable way of paying those people whose work has been used to train an AI is what they're calling cooperative game theory.

So, the long and the short of this seems to be they are going to, essentially for any output, this is a way of figuring out which of the many inputs into the AI's training has had the most significant contribution to that output.

So, if you imagine that every output generates a pot of, shall we say, a hundred units that is going to be paid out to authors whose work has been used, they would establish how many hundredths of that output was contributed to by each author whose work has been used to train the work, and the winners in that battle, the ones whose work has contributed the most, they will be proportionately rewarded.

So, the way I referred to in the column this week is, if you imagine going into a hardware store, a DIY store, they have those things where you take in a bit of, I don't know, your favourite bit of fabric, or the cover of your favourite journal. That's something I can imagine myself doing. If I want to match that to the paint I use to paint my walls, I say, please make me a paint that's this colour. So, they scan that and then they use their basic ingredients. If it was digital? It would be RGB or CMYK. It would be red, yellow, blue, primary colors, bit of white, bit of black, and they would mix those to certain proportions in order to get the color I wanted. That might be, say, I don't know, 42 percent red, 17 percent blue, 41 percent yellow. That works, I think, that adds up to 100.

It's the literary equivalent of that. If that were the mix that had led to the output, it would be that author red would get 49 units of currency, 17 to author blue, and 41 to author yellow. You can comment if I've got those colors wrong, but you get the principle behind that. Anyone wants to look up what that color looks like, send me a copy, that would be fabulous. I would be intrigued. Sort of magentar-y pink-y purple, I would imagine.

But that's the principle. You work out which inputs have contributed most to the outputs, and you divvy up a pot and reward them accordingly, and that is how it looks like OpenAI are imagining a payment system might work.

How this works with consent, of course, that's an interesting one. That assumes that people have already consented. That's very much a, once the horse has bolted, sort of a situation for how to go about paying people.

We know, of course, that they do also favour this, drawing up agreements with publishers, but of course, as indies, on an ongoing basis, it's not necessarily going to be feasible to draw up a big agreement between one author and OpenAI.

Registering in some way for this kind of thing, if your work is going to be used, might be a way to do it. How people opt in, how people opt out, who knows? I'm sure that is something they will be thinking about. I still think it's most likely that we are going to see a one-off ‘register for your payments now and then forever hold your peace' model.

But this is an alternative, and it is an alternative that they are looking at that would be sustainable. Whether or not it works out in practice, that authors who don't opt into this will then have their works not used to train AI models, who knows? It sounds a little bit as though it's idealistic thinking, whether it works in practice or not. It could be incredibly complicated.

One of the things that it actually does point out is that it would take a lot of computing power just to run this system. But if you don't want to go and read the whole paper, I understand. I've copied out the abstract in the column this week. Do go and check that out, it's a really interesting abstract. It sums up their philosophy. It's a very much a tech-based, mathematical philosophy. There's not much about creation and there's not much about the value of the creative process in there, and that's pretty much as you might expect. But it's an interesting story.

Google Anti-Trust Lawsuit Ongoing

Other interesting stories this week. The lawsuit around Google is ongoing. So, this is the big antitrust lawsuit against Google that is accusing them of basically throwing their weight around, paying people lots of money to operate as the default search engine on their operating systems.

So, it's said that Google had paid Apple $18 billion to be the search engine of choice on iOS. Very interesting. So, that's anti-competitive behavior is what is being alleged, and whether or not it is going to be adjudicated upon soon. One of just many antitrust cases that is coming up that will be of interest to us.

“No One Buys Books” Article Debunked

The other thing that keeps rumbling on is Elle Griffin's, No One Buys Books article, which claims that no one is buying books. That books don't make any money, and so yet more people have leapt on that bandwagon and said actually people are buying a lot of books. So, Slate have done a piece on this, and they've crunched the numbers and they have figured out Americans buy over a billion books a year, and these books, they do make a lot of money and Publishers aren't going out of business any time soon. In addition to that, indies are making lots of money too. Yeah, everyone's buying books. It's a really odd thing to say that no one's buying books.

It feels like the latest version of the death of the novel. Talking of which, that's something I haven't seen for a long time.

It seems like several years since we've had some big-name author who is a best-selling novelist telling us that the death of the novel has happened, usually when they stop selling novels.

There will be a crossover at some point, you know, death of the novel, no one buys books. What they actually mean is no one buys my books anymore.

So yeah, maybe go out and write a different book that people actually want to buy and read, because lots of people still love reading and still love books.

And what a lovely way to finish this week's self-publishing news. I hope to be back next week with some non-AI news, but nonetheless some good news again.

So, I will speak to you again then, and I will get back to the rainy athletics of Oxford. Thank you all.