This final article in a four-part series lays out a vision for what the future of books with AI could look like, outlines a 6 part plan of what needs to change, and provides a CTA on how to get involved

In the previous article, we discussed how licensing of book data should work in AI / book publishers agreements, and where the primary concerns of consent, credit, and compensation break down in the traditional model.

We touch on how Inference represents a much more massive opportunity than licensing Training, and how books are currently better situated for use in Training AI, and not for Inference.

In this article, we continue that conversation by outlining a vision for the future of AI and books from both the reader and the publisher's perspective, discuss 6 things that need to change to unlock that future, and ultimately provide a call-to-action for you, the reader, to get involved.

A 360° Vision for the Future of Books

The future of books and AI can be viewed from two perspectives — that of the reader and that of the book industry.

The central assumption in this section is it requires (1) some changes in how we think about licensing book data and (2) some technological infrastructure that puts control of consent, compensation, and credit in the hands of the IP holders, which we’ll discuss further in the final section.

If we assume those to be true, let’s explore what that future could look like.

Also…if any of these ideas feel like science fiction, I assure you that in every case, they are science fact—completely doable with existing technology—much of which I have firsthand knowledge of because I’ve written the code for it myself.

Let’s start with the reader’s experience.

The Future of Books for Readers

Integrating AI into the realm of books opens up many possibilities that extend far beyond intelligent book recommendation systems.

There are a few ideas that I think represent the tip of the iceberg.

AI books for recreational readers

Imagine an AI reading companion that transforms your reading experience into a dynamic, interactive journey.

This companion could seamlessly blend listening and reading, allowing you to switch between or even combine both modalities simultaneously.

Need a summary of the last chapter because you set the book down for a month? Ask the AI companion and it can bring you up to speed.

Confused about a character’s motivation? Ask the AI companion, and it will provide contextual insights to enhance your understanding.

This AI could offer hands-free functionality, and allow you to talk with an AI narrator as if the human narrator was reading the book to you personally.

I often listen to books when I’m running or biking, and hear something and think “This is perfect for my colleague, they need to hear this” or “I want to look this up later”.

With audiobook apps today, there isn’t a simple way to handle these types of interactions. Have you ever tried to find your notes in Audible?

The way I solve this today is I stop and pull out my phone, and then do a quick Google search, and leave the tab open so I can come back and find it later.

But imagine if you could just ask your AI companion, hands-free, and it would take a contextual note for you, dig into specific parts of the book, or handle the request immediately.

Personalized Audio Narration

With AI-powered audiobooks, you could tailor the narrator’s voice to your liking.

Want Barack Obama to narrate your business non-fiction reads? Or perhaps you’d prefer Michael Kramer for all of your fantasy novels?

Or perhaps you want Harry to have the voice of Daniel Radcliffe and Alan Rickman as Snape.

Audiobook narrators could license their voices as digital clones, giving readers unprecedented control over their auditory experience.

AI books in Early Education

For early childhood education, an AI reading companion could revolutionize how children learn to read.

It would assist them in recognizing letters and words, providing guidance with unlimited patience. The AI could make learning interactive and fun, adapting to each child’s pace and style.

It could offer instant translation, teaching new languages on the fly without any additional effort required from the IP holder to manage translations.

Want your children to learn Spanish or Chinese? Your child’s AI reading companion could help teach them.

AI books in Higher Education

In higher education, textbooks could become interactive learning platforms powered by dynamic AI tutors.

Students could ask questions and receive not just factual answers but also guidance on where to find more information within the text.

The AI tutor could generate quizzes on the fly, assess a student’s understanding, and adapt the difficulty of questions in real time to address specific learning needs.

AI books in Business

There is already a proven market for book-based businesses for business.

Book Summaries

For instance, consider the market for book summaries, where companies like getAbstract, Blinkist, and Headway play.

The process of licensing books for content creation is a demanding and laborious process that requires dedicated resources.

For this reason, summary libraries are limited in scope, and some of the companies don’t even license book content directly and instead skirt the fringes of copyright law.

The improved infrastructure required for AI-powered books would expedite the licensing process, allowing more of these companies to not only legally use book content, but also expand the number of books that are included in their libraries.

This could expand their audiences, reduce costs, and allow them to channel more revenue back to IP holders.

Professional Development & Consultants

Companies like BookClub and Campfire build professional development programs using AI and books and often work with creators to provide them with digital tools to serve their customers.

Better infrastructure could open up a world of resources for authors and consultants to magnify their abilities and expand the impact of their IP both in scope and duration.

Book-based Coaching

Today, Jim Collins, author of Good to Great, has a consulting agency that runs workshops, seminars, and coaching.

Imagine Jim Collins creating and then licensing a virtual leadership coach powered by AI to all of his clients.

This virtual coach could be trained on the knowledge from all his books, enabling leaders at every level to interact with a “virtual Jim” for guidance on leadership challenges and personal development.

Subscribers to his service could become part of a broader network, sharing experiences and learning collaboratively, and Jim Collins could learn how his clients are using his coach and what the most common challenges they’re facing are, giving him insight as to what to write his next book about.

Corporate L&D Teams

Businesses that leverage books as part of their company culture could access entire libraries of content to build customized training programs.

AI-powered text, audio, and video materials could be contextualized to meet specific organizational needs, making training more relevant and engaging for both readers and non-readers alike.

Relevant case studies could be tailored to align with the company’s industry or challenges, enhancing the applicability of the learning material.

The Future of Books for the Book Industry

The potential for AI to completely transform the book experience for readers is massive. But potentially even more massive is what AI could unlock for the book industry itself.

License content to book-based AI applications

Today, readers can purchase copies of “digital books”, in the form of ebooks, but they don’t own these books the way they own physical books.

Their purchases end up living in specific ecosystems, and other than reading it, there is very little the user can do with it.

Imagine a scenario where readers purchase copies of digital books through transactional payments, much like they do with physical books today.

However, these digital books would be fully owned by the reader, and the reader could choose to “plug” their book into different applications.

For example, each of the reader experiences described in the previous section could be a separate application, with their own business models layered on top.

With this model, the IP holder — the author or publisher — could have full control over the terms of how consent, credit, and compensation work within each application.

If they didn’t want an application to use their data, or there was a use case they were not comfortable with — then they could specify that in the digital book, and it would prevent it from happening.

New Revenue Models

Books have historically been confined to very limited subscription-based business models, such as Audible or a “book-of-the-month” type subscription.

However, with ownable digital books, we could create applications that allow readers to subscribe to their favorite publisher or author, gaining access not just to content but to a community.

Similar to how you can subscribe to video streaming services like Peacock or HBO Max and access all the content from those media companies, the same could happen for specific authors or publisher imprints.

Take Brandon Sanderson, for example.

He has run record-breaking Kickstarter campaigns to fund his books and ideas.

But what if readers could subscribe directly to Brandon Sanderson, with different tiers providing varying levels of access to his IP, including the entire library of books he has written, and will write in the future?

Much like his Kickstarter campaigns, but on a recurring basis, this would eliminate the need to launch new campaigns and build a community of supporters who are continuously engaged.

Fine-grained licensing controls

Publishers could maintain ownership of their IP while tracking where it is used, who is using it, and what they are doing with it.

If they are uncomfortable with a particular use case, they would have fine-grained control to disable it. By establishing common licensing agreements, the ability to connect use cases with approved revenue-generating activities could be dramatically accelerated.

Currently, publishers and agents spend countless hours and dollars negotiating contracts and ironing out the details of partnership arrangements. I know, because we had an entire team dedicated to this in the early days of BookClub.

With better systemic tooling, this process could become nearly instantaneous for digital applications, saving time and money, and bringing exciting experiences to readers faster.

Publishers that adopt these solutions would also have a competitive advantage in winning authors to their firms, as they could move more swiftly and be more responsive to market needs.

Unprecedented data access

Publishers and authors could gain access to data in ways they’ve never experienced before.

They could see exactly which books are sold, at the moment they’re sold, who it was sold to, and through which channels (instead of waiting for the high-level weekly NPD report).

They could understand how readers engage with their content — what apps they’re using, which parts of the books are most popular, and which ideas are trending in the market.

They could analyze how consumption patterns change over time, providing valuable insights to predict which types of books will most appeal to readers, what the market needs or wants, and perhaps most intriguingly, where there are gaps in the content landscape.

I think one of the big roles book publishers play today, and will play an even bigger part in the future, is the ability to help bring big ideas to market, and then help structure those ideas in a way that resonates with people.

Having more direct, accurate data around what data is used and how will allow publishers to identify and bring big ideas to market faster than ever.

Participation in secondary book markets

Opening up secondary markets is another exciting possibility.

When readers truly own their digital books, they gain the power to do anything they want with them, including reselling.

With physical books, authors and publishers cannot dictate the rules of reselling, but with digital books, they would have a say.

They could participate in the resale market of digital books, with a percentage of the resale transaction returning to the IP holder.

This would have huge implications in higher education and collectors markets.

Readers could rent out their books.

Today, you can borrow books from services like Libby, but for popular titles, you might wait months to get access.

Meanwhile, there are purchased copies of the same book sitting unused in people’s Audible accounts eternally after they’ve listened to them.

We live in a sharing economy — people rent out their homes (AirBnB), their cars (Turo), their time (TaskRabbit), and their talents (Fiverr).

What if you could do the same thing with your digital books, with a portion of the rental fee going back to the IP holder? Let your audience do your marketing for you.

Limited-run digital prints

People think just because something is digital means there could be infinite copies of it.

But what if there weren’t?

What if you had digital books as limited-run series? Or what if authors provided personalized notes in specific copies, or annotated their own book with commentary and included it with limited editions?

It could help drive demand and capture latent interest from the author’s biggest supporters. It could help incentivize usage in secondary markets and prevent brand dilution.

These limited-run editions could entitle the owner to special access and events.

Maybe JK Rowlings holds a private screening of the next Fantastic Beasts movie for the owners of the pre-release edition of her next book, or Jim Collins provides unlimited personalized access to his services to the 10 owners of his $100k limited edition run of Good to Great.

The opportunities are endless.

Digital narration

We discussed digital narration from the reader's perspective, but there’s a huge opportunity from a business perspective as well.

For instance, audiobook narrators could license their voices.

They could set limits on the types of use cases where their voice could be used, perhaps avoiding genres or authors they’re uncomfortable with.

This opens up new revenue streams for narrators and adds personalization options for listeners.

Or services could be created that charge parents to generate AI models of their voices, which would allow them to “read” every audiobook for their children.

Direct access to the market

One of the most significant opportunities, I believe, is that with the right tooling, publishers could have direct access to their audience.

They wouldn’t have to sell exclusively through other channels or remain disconnected from their readers as they are today.

Through these systems, publishers could build direct relationships with readers, shifting the balance of power away from distributors like Amazon and back into the hands of the IP owners.

The opportunities are limitless

Truly. The ideas I have listed here are the ones my team and I have come up with over the years as we’ve pondered the possibilities of AI and books.

But the incredible thing about building systems that open up ecosystems is the imagination of the human mind is endless, and provided the proper tooling, the things people can create are truly endless.

My hope is that presenting a few of the big ideas here, it can help to paint a picture of what could be possible.

What needs to change to enable this vision of the future of books with AI?

Before I dive into the specifics, I’d like to reiterate my stance.

I believe books deserve to have a vibrant future, and book publishers play a critical role in making that future happen. I am an entrepreneurial technologist who is pro-books, pro-publishers and pro-authors.

Everything I have to say comes from a desire to help.

I understand there are reasons why things are the way they are, but as the saying goes — no pain, no gain.

Some of these changes might be no-brainers, while others could be challenging — and I know there are countless reasons people might say it won’t work.

Think of me as a trainer at the gym. I see the potential you have, and I’m going to push you in ways that might be uncomfortable but are necessary to achieve your potential.

There are six fundamental changes I believe need to take place:

1. Establish Industry-Wide Licensing Standards for AI Applications

Licensing books today is like navigating a labyrinth blindfolded — it’s overly complicated, time-consuming, stifles innovation, and I believe is the primary driving force behind Alex’s concern.

It needs to be easier to license book content for usage with AI.

Consider how other industries have tackled similar challenges.

In the banking sector, consortia of banks have come together to build financial infrastructure, making it seamless to transfer money across institutions.

Airlines have formed alliances, aligning on standards to simplify booking tickets and coordinating logistics across networks.

In the tech world, organizations like the World Wide Web Consortium (W3C) develop web standards so that the Internet functions smoothly for everyone.

Book publishers could do something similar.

By forming a consortium or alliance, publishers can create standardized licensing agreements and protocols that make it easier for AI companies to license book content for AI-based applications.

This not only streamlines the process but also ensures that publishers and authors maintain control of their IP.

2. Secure Book IP with Public Advanced Digital Rights Management

We don’t want books leaking into the public domain unintentionally or being used in ways they shouldn’t be — securing book IP should be of top concern.

Therefore, one way to securely and publicly store books would be to encrypt and write them on the blockchain.

This would provide a secure ledger that ensures only those with permission can access and decrypt the content, with full traceability.

With blockchain-based smart contracts, you could dictate, on a per-book basis, who has access to a book, what they can do with it, and for how long.

For instance, maybe JK Rowling would be comfortable with her books being used with an AI reading companion app, but not with an app that helps fans write “authorized” fan fiction — she would have control to do that using a system like this.

We could also implement digital watermarks for traceability, which can help monitor where and how book content is used, and provide culpability in security breaches.

By managing access carefully, we can prevent unauthorized use while still encouraging legitimate, innovative applications.

3. Create a Standard Digital Book Format for AI

After working with digital book files for the last few years, I can authoritatively say that they are a mess.

Not only are there a plethora of incompatible, competing formats, but there is no consistency in how their content is structured.

This makes it extremely difficult for developers to build applications with them, which ultimately slows innovation.

Standards unlock ecosystems.

Consider the example of Milwaukee’s battery system for power tools.

Milwaukee created a standardized battery that works across all their tools, which not only simplifies things for consumers but also encourages them to invest more deeply in the ecosystem.

Similarly, a standard digital book format would encourage AI developers to create new and exciting applications and continue to invest in the ecosystem knowing they can reliably work with the content.

4. Simplify Access to Book Content

Building on the previous points, we need to make it easier to access books digitally, both for individuals and applications.

Tech has already solved for how to securely, easily, and safely transfer sensitive data between applications.

Have you ever signed up for a website using your ID from another website, like your Facebook, Apple, or Google ID?

Spotify allows users to login with their credentials from other platforms, for convenience and security

Spotify does this.

Spotify had to obtain permission from Facebook, Apple, and Google just to request access to your data, and then once they have permission from those applications, they still need your permission (as the user) to actually access your personal information.

It’s a two-step process that prioritizes security and consent, yet it’s seamless for the user and broadly adopted.

A similar model could work with books.

Book applications could apply with publishers for access to their IP, specifying how they’re going to use the data.

Once the application is approved, individual access can be checked to ensure individuals have the right to access specific pieces of data. Once those permissions are in place, accessing the content should be easy.

This approach ensures that consent is obtained at both the application and individual levels, without putting unnecessary barriers in the way of legitimate uses.

5. Develop New Payment Models

The way we sell and license books needs to evolve.

Currently, it’s binary — you purchase the entire book or not at all.

But AI applications like LLMs might not need access to the entire book; they might only need a chapter or even a few paragraphs.

Allowing books to be purchased or licensed in parts opens up new possibilities.

Facilitate micro-payments by allowing applications to pay to access only the portions it needs, for the specific use case it has, both on a temporary and permanent basis.

This can facilitate both transactional payments and subscription-based models.

The music industry made this same adjustment with the introduction of the iPod in the early 2000’s, by allowing users to purchase a single song instead of the whole album, and building streaming services instead of transactional marketplaces.

This flexibility can create new revenue streams and make it more attractive for developers to work with book content, and for readers to consume it.

6. Empower Digital Book Ownership

Currently, when you buy a digital book, you don’t truly own it — not in the way you own a physical book.

With a physical book, you can highlight it, lend it to a friend, annotate it, or even resell it if you want to.

Early editions of books have value, and copies owned by notable individuals can become collectors’ items.

In the early days of technology and the internet, when it was impossible to prove ownership of a digital asset—it made sense why this was not a thing.

But with the ability of the blockchain to immutably cement IP and prove individual ownership—why should digital books be any different?

People should be allowed to truly own their digital books and control what they do with them.

This could involve allowing for personal annotations, lending their digital books to a friend, or even reselling them.

By embracing this, we honor the traditions of book ownership while adapting to the digital age.

Additional considerations

I recognize I’m pretty pragmatic about my recommendations, and various other aspects of this conversation demand attention, such as the broader cultural, ethical, and educational issues that I didn’t even touch.

My hope is this starts a conversation that drives toward the changes needed to forestall Alex’s grim prediction and enable the hopeful vision I’ve outlined in this ebook.

Now what?

The transformative potential of AI in the world of books is immense, but realizing it requires the active engagement of forward-thinking individuals like you. If you’re interested in shaping this future and discussing ways to get involved, I’d love to connect.

Let’s talk directly about how we can make this vision a reality together. You can connect with me on LinkedIn or, if you prefer, schedule some time with me directly for a 1:1 conversation on how to get involved. Let’s build the future of books with AI, side by side.

This was the last article in a four-part series discussing the Future of AI and Books. If this was the first article you found, I’d encourage you to read the previous 3 articles to help provide some critical context.

1. AI and the Future of Book Publishing
2. The Current State of AI and Publishing
3. Why the Established AI Content Licensing Model Breaks with Books
4. Vision and Call to Action for the Future of Books and AI

If you’re a bit fuzzy on some of the AI terminology used in this article, I’d recommend reading my “Beginner’s Guide to Understanding Generative AI”.

This third article in a four-part series discusses where the established content licensing models break for books and makes the case for why this is a problem worth solving properly

In the previous article, we discussed what’s currently happening in the relationship between AI companies and publishers, specifically focusing on digital publishers.

We evaluated how the 3C’s of consent, credit and compensation are likely working their way into those partnerships, both through the lens of training and inference.

In this article, we will continue that conversation by discussing why that model doesn’t work well with books, and why it’s a problem worth solving.

Book publisher content licensing for usage with AI companies

We’ve talked a lot about training and inference in this ebook, and by now, I hope you’re beginning to understand the implications of each.

Here’s another huge one.

Experts estimate that the AI inference market will be 100x the size of the AI training market.

This is a massive issue because in their current format, books are best positioned to train AI, and not for AI inference; whereas digital publishers, due to their inherent structural differences, are well-positioned to participate in both training and inference opportunities.

This issue is compounded by the fact that there’s reason to believe the days of AI companies paying to license content to train their models will be short-lived.

Consider Meta’s (Facebook) approach to AI.

Where companies like Google, OpenAI, and Microsoft have built closed-source LLMs, Meta has made their line of LLMs (called Llama) open-source.

If you’re unfamiliar with the two terms, it’s fundamentally the difference between making software private (closed-source) versus public (open-source).

As you might imagine, if Meta makes their AI open source and free to use, Meta has to rely on different sources of data to train their AI, as they don’t have the same budget to purchase content as OpenAI has done.

Meta CEO Mark Zuckerberg recently stated that creators overestimate the value of their work in training AI, and they won’t pay to license content to train their AI.

He claims that there are enough other public resources available for training AI, and if creators don’t want them to use their content, they just won’t use it.

Plain and simple.

On the one hand, it may be comforting to know that book IP may not be as critical as some may say to the training of AI.

On the other hand, since books are best positioned to participate in training AI, it means unless something changes, they could completely miss out on the much larger inference opportunity AI represents.

It’s early to make the estimate, but if it turns out the market for training data dries up like Mark Zuckerberg estimates, Inference will turn out to be an infinitely larger opportunity, not just 100x larger than the AI training market.

However, for the sake of discussion, let’s assume AI companies will continue to pay to license data for Training, and discuss why these partnerships need to look different from digital publishers.

We’ve had to cover a lot of ground to get here, but we’re finally arriving at the core issues we came here to talk about.

Why are books best positioned for training AI and not inference?

Similar to how we analyzed the digital publisher perspective, let’s answer this question by analyzing how the three C’s—credit, consent, and compensation—should be addressed in AI partnerships, first through the lens of Training and then Inference.

Book Content Licensing: Training

Not having access to the details, I’d bet this is the primary focus of the Wiley AI partnership; partially because their public announcement specifies that they licensed their data to AI companies to “train their Large Language Models (LLMs)”, but mainly due to my experience and understanding how the AI works, I have little reason to believe their partnership would include inference-based applications.

I’ll lay out my reasoning for this belief over the next few sections.

Consent and Compensation

Similar to the licensing of digital publisher data, licensing book IP for training would also likely fall under a similar perpetual licensing model.

The license for the IP should take into consideration each successive new model the AI company trains, which would make sense as the data requirements for each model shift as the researchers learn more about what type of data they actually need.

In this instance, both the consent and terms of compensation would be worked into the same agreement.

Books today are well-positioned for use in training LLMs because you need massive amounts of unstructured text with diverse language patterns, vocabulary, and context.

An LLM is designed to learn from raw natural language, which is inherently unstructured, to capture the extraordinarily complex relationships between words, sentences, and domains.

As long as the book exists in a digital format, the content can easily be extracted and used to train an LLM.

This is good news for books, because these days, nearly all books exist in some digital format.

So all a book publisher needs to do is sort out consent and compensation through a perpetual license agreement, and then dump their books into an FTP and share it with an AI company as a training “data set”.

While this “works”, the biggest challenge is once the AI company acquires the data, the book publisher loses complete control over how that data is used.

Credit

All of the same points from our analysis of Digital Publisher content apply here, which mainly point to the inherent difficulties of providing credit to source materials based on training alone; it’s the two piano challenges we discussed of (1) self-referencing and (2) how to source the training of every note — there’s no pragmatic way to do this.

However, AI companies can use formal partnerships with book publishers, like Wiley, as a way to publicly provide credit to them for their IP and convey the copasetic use of Book IP in training their models.

Book Content Licensing: Inference

Recall, experts estimate Inference will be 100x the market opportunity that Training represents.

And this is where we encounter the biggest problem: in today’s system,

There is no easy way to license and use books for AI inference.

This means there are no easy ways for book IP holders to leverage their content to participate in the largest opportunity around AI.

Let’s dissect this through the 3 concern lenses of credit, consent, and compensation — in that order.

Credit

This should be the easiest problem to solve for books.

One of the best ways to use books with an LLM is with RAG — to inject the book text into the context window and ask the LLM to reference the text and answer questions about it.

While there are legal considerations, technically performing RAG with book content is quite straightforward.

The challenge comes in how to provide credit for it.

If you are the one copying the text from the book and putting it into the LLM context window for your prompt, then you can be confident that the data the LLM references came from the book.

But what happens when you are not providing the data for RAG?

With Google and Reddit, this is the case. Users do not provide content directly; Google handles that behind the scenes.

Google has spent the last 25 years solving the biggest problems in connecting the immense depths of the data on the internet and making it easy to surface the information you’re looking for.

The reason you believe search results with Google is not because you take Google’s word for it, but because Google can instantly connect you with the source of their results, and then you can make your own determination around the validity of those sources.

But how would you do this with a book (specifically those under copyright)?

The LLM could provide you with all the information you want to know about the book; it could even give you the exact location by word position in the book — but then as the user, unless you have the book sitting in your library and you’re willing to look it up yourself, how do you verify the results?

It’s not like you can go to http://www.INSERT_BOOK_TITLE.com/chapter1 and find a reference to the source material.

The inability to generate referenceable attributes with books is a massive impediment to leveraging books for inference and providing credit.

Consent

Previously we’ve connected to the idea of a pianist not being able to cite any specific location where they learned to play, but rather they built their skills through many sources — and thus, no single source stands alone or can be highlighted.

Part of the power of inference, however, is the ability to provide a source and intentionally highlight a specific piece of content. Like Google, this is how you build credibility with the user.

The headline of the article about the Wiley announcement confirmed that authors could not choose to “opt out” of that partnership.

For Wiley, this likely won’t ultimately be an issue. In addition to their strong market reputation, they’re providing training data “soup” for AI companies, and thus individual author’s content is likely to be immediately highlighted—which minimizes the potential impact to impact any single author.

But what would happen when an LLM through inference responds to a prompt and cites a specific author who doesn’t want their data in the LLM, or a book publisher didn’t have permission contractually to integrate?

The training of the LLM is complete — the data is already incorporated into the model and like music in the pianist’s brain — the author’s data cannot be easily extracted.

Reddit does not face this same challenge with Google, because as part of the Privacy Policy you agree to when you check that box to use their site, you own the copyright of your data, but you implicitly grant Reddit an unrestricted license to use your data.

The same is not true in most contracts with authors and book publishers.

Extensive resources go into every contract, defining who has permission for every single thing, and ultimately declaring the bounds of what they can do with it.

Inference moves from handling rights in aggregate, and looks at them from an individual perspective — and these are the scenarios that the current infrastructure is very poorly equipped to handle.

Compensation

As previously mentioned, the inference model most readily relates to a subscription model.

Reddit made the change of charging users every time they accessed data over their API.

Most AI companies operate the same way — they charge users for every time they access their models via their API.

Ideally, there would be some way to track how often (and how much of) a book is referenced in inference, and the consumer would be charged based on their usage of the source material.

However, with books, it’s not clear how this would work.

As we mentioned previously, there is no way to canonically reference the content of a book, much less track the usage of that data by an LLM in a way that provides the consumer access to it.

This is one of the biggest challenges that will prevent book publishers from capturing this potentially enormous revenue stream.

The crux of the issue

I’ve summarized the key points for this section in the below matrix:

Perhaps your natural inclination after reading thus far is to dismiss the challenges. Perhaps some of these thoughts crossed your mind

“Books aren’t going anywhere, people are always going to read. This issue is not relevant.”

“Books have survived every technological advancement of the last thousand years, this one is no different.”

“So all that’s needed is a book API, and everything is fixed. Easy peasy.”

“These issues seem trivial, I’m sure someone is working on them.”

It may seem simple and obvious, but there are three reasons why I think these issues are critical and imminently relevant.

First, the biggest impact often comes from the simplest changes.

Consider Stripe, an online payment processor.

Stripe was not the first to handle online payments; there were many companies before them.

They were the first to realize that their competition’s documentation was terrible, which made it difficult for software developers those solutions.

So Stripe focused on generating excellent documentation for their product, and as a result, many developers readily adopted their product. Stripe now handles nearly 25% of all online financial transactions

Sometimes you have to acknowledge the problems hiding in plain sight before something can change. I believe this is the case with books.

With a few simple changes and a little technology, I think there could be a massive opportunity waiting for book publishers.

Second is the question of who will own the problem and bring the solution to market.

If a little technology could go a long way, then the question is who is going to build it?

The natural response is, “Well this is the publisher’s problem, they should build it”.

But if this were true, why haven’t book publishers been able to realistically compete with Amazon? Isn’t selling and distributing books to people via the Internet also their problem?

I would argue it’s because book publishers focus on what they’re excellent at — discovering, curating, and bringing big ideas to market in the form of books — and not on building technology products.

My understanding is that book publishers have a tenuous relationship with Amazon, at best, and that offered a viable alternative, most would take it.

Perhaps the solution is another tech company should own the problem and lead the conversation.

But then how do you prevent another Amazon situation from happening, where 20 years down the road when that tech company owns the books and AI ecosystem, and book publishers are forced to operate at the whim of their AI distributor?

Maybe each book publisher is left to solve this problem on their own?

Given the fractured nature and current trajectory of the industry, I believe this is likely where things are headed, and is part of the problem “…unless. something changes…”.

Finally, the opportunity at hand is massive and should not be dismissed.

If handled correctly, this is not just about licensing book data for training AI, or becoming a footnote in a ChatGPT conversation or Google search.

The opportunity here is a complete evolution for books that could unlock limitless new applications and market opportunities.

Consider Wiley.

For the last 2 years, Wiley has made ~$22MM per year licensing their content to train AI.

If the inference opportunity is 100x that of training, then they could be sitting on a $2.2BN opportunity to license their data for inference-based applications.

True, it may cannibalize some of their current business.

But even if it cannibalized all of their current business, it would represent a 10% revenue increase from where they sit today at a much higher gross margin — which would have a massive increase on their bottom line ($17MM net income in 2023; trending negative in 2024).

Let’s now turn our attention to what type of future could unlock the 100x inference opportunity for books.

This was the third article in a four-part series discussing the Future of AI and Books. If you’d like to continue reading, click the relevant link below.

1. AI and the Future of Book Publishing
2. The Current State of AI and Publishing
3. Why the Established AI Content Licensing Model Breaks with Books
4. Vision and Call to Action for the Future of Books and AI

If you’re a bit fuzzy on some of the AI terminology used in this article, I’d recommend reading my “Beginner’s Guide to Understanding Generative AI”.

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or connect with me/send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

This second article in a four-part series analyzes the partnerships Digital Publishers have made with AI companies and sets the stage for why these partnerships won’t work for Book Publishers

In the previous article, we identified the key concerns for publishers that AI represents a loss of control of data; by disregarding consent, credit and compensation of their IP. We evaluated the validity of those concerns and then provided an outline of what this series would cover.

In this article, we will continue that conversation by discussing what’s currently happening in the industry and how those relationships are likely addressing the previously established concerns.

When the game changed

The world shook when ChatGPT was introduced in November of 2022.

It took 5 days for ChatGPT to reach 1MM users, and it reached over 100MM within 2 months.

ChatGPT was the fastest-growing application of all time, by far; the level of user interest and media attention it garnered was unlike anything the world has ever ever seen.

The fictional AI of our childhoods was no longer just a story; ChatGPT made it a reality. It reignited every conversation about AI, from futures of abundance and leisure to terrified technophobic claims regarding the rise of Skynet.

Everyone started to worry about their jobs because ChatGPT showed the world that AI was capable of doing everything we didn’t think was possible.

Hey AI companies, where’d you get that data from?

7 months later the lawsuits against AI companies began.

The first shot across the bow came from book authors Mona Awad and Paul Tremblay, asserting their IP had been used without permission to train GPT.

They had a right to be concerned.

In the early days of ChatGPT, you could ask it nearly any question about a book you wanted to, and if it had been included in the training data, ChatGPT would answer similar to how it repeated The Art of War earlier.

It was clear that books played a big part in training ChatGPT (they admitted it themselves), and the assertion by authors and publishers was that OpenAI had violated copyright law in doing so.

Dozens of similar lawsuits have since followed, with many authors and publishers of all kinds asserting their IP had been used “without consent, without credit, and without compensation”, as we discussed in the first article.

Book Publishers, however, were not alone in their complaints against OpenAI and the other AI companies.

Digital publishers like the New York Times filed lawsuits claiming Microsoft had violated copyright law, and online forums such as StackOverflow (a popular website for software engineers to find answers to technical questions) and Reddit became extremely vocal as they observed massive hits to their traffic as a result of ChatGPT.

ChatGPT represented a new channel for people to directly and quickly find answers to their questions, without having to browse endless forums for the one relevant response.

Solving for Consent, Credit and Compensation

In retaliation, Reddit made a dramatic change to their API (API stands for Application Programmable Interface. API’s are how software applications communicate with one another. We‘ll talk more about APIs later, so remember this definition for when we get there in case you haven’t heard of it before).

Data that was once available to the public for free now required payment

This change caused a massive controversy in the Reddit community, even warranting its own Wikipedia entry, but this new way of thinking had a tremendous impact on Reddit’s business model.

No longer able to access the data for free, Google struck a $60MM per year deal with Reddit to license their data. It presented one of the largest deals to date where a digital publisher explicitly licensed their data for AI use.

Reddit now provides some of the most popular data used to power Google’s AI Search Responses; this is one of the reasons why you may have started to notice Reddit showing up at the top of so many of your search results in recent months.

The three core concerns have been addressed in the partnership:

1. Google operates with Reddit’s consent to leverage their content for AI (both in training their model and through search, which is Google’s version of inference)
2. Reddit gets credit, or links to its website in the Google search results
3. Reddit is compensated by the licensing deal where Google uses its data to train their LLMs. Reddit has also likely seen a huge boost in their advertising business due to the massive traffic increase

It appears that digital publishers and AI companies have found a path forward.

Licensing data for use with AI: a new revenue model for digital publishers

This model of AI companies licensing data from providers is becoming increasingly common.

Recently, a wave of well-known digital publishers such as Conde Nast, Time, Vox Media, and The Atlantic made announcements that they have struck “strategic partnerships” with OpenAI to license their content.

NewsCorp (parent company of media giants such as The Wall Street Journal and The New Yorker) announced a huge 5-year, $250MM partnership with OpenAI for access to content across all of their properties.

The model of AI companies like OpenAI, Microsoft and Google licensing data to be used in training, and then providing credit and compensation through inference appears to have a strong foundation.

It’s working with Reddit, and while I have some concerns about the longevity of this model, I have no reason not to believe it won’t work in the near term for the other digital publishers as well.

Digital publisher content licensing for usage with AI companies

You may notice I keep saying “digital” publishers, which I have done intentionally to distinguish them from “book” publishers.

AI partnerships are happening with book publishers; Wiley recently announced a $44MM deal to license their content with AI companies, which we’ll discuss further in the next article.

The licensing model digital publishers are establishing has critical differences when it comes to how it could apply to book publishing, which is where Alex’s “…unless something changes…” most directly applies.

First, let’s dig in a bit further and build a better understanding of how these digital publisher / AI partnerships work.

Disclaimer. None of the details of the aforementioned partnerships are public, so the following is my expert analysis of what’s happening.

In this section, we discuss several key concepts around AI.

If your understanding of these topics is rusty, please check out my “Beginner’s Guide to Understanding Generative AI”. In that article, one of the ideas I discuss is how the data in an LLM becomes obsolete starting on the first day of training.

From that day on, an LLM’s ability to generate responses based on its internally trained data ends, and it requires external data sources to respond about anything more recent.

For this reason, data usage in training and inference must be considered separately for license partnerships.

In the case of the Google and Reddit partnership, two parts map directly to how LLMs work:

1. Training. Google has access to historical Reddit data for use in training their LLM. Once the historical data makes its way into the LLM, just like our piano music analogy from before, it cannot be removed and is referencable forever.
2. Inference. Google has access to both historical and new Reddit data via the API for use in inference (again, inference is called “search” in Google’s case). This access allows it to provide current results via “RAG”, and it is the mechanism by which the search results reference the source data they draw from.

Distinguishing these use cases allows us to think through how the three concerns previously mentioned — consent, credit, and compensation — work their way into the partnerships. We will do so first by analyzing the 3C’s through the lens of Training, and then through Inference.

Digital Publisher Content Licensing: Training

Consent and Compensation

Data used to train an LLM is licensed on a perpetual, “per model” basis . This to me is the logical case, simply due to the impracticality of any other alternative.

Given the time and resources it takes to train a single LLM combined with the semi-permanent and persistent nature of that LLM after it has been trained, there wouldn’t be a way to “continue” to license data with time.

Since granting a perpetual use license to train multiple LLMs would be a lost business opportunity (each iterative model is essentially a new business), data is licensed every time a new LLM is trained and released.

For instance, OpenAI has released GPT-2, GPT-3, GPT-3.5-Turbo, GPT-4, GPT-4o, GPT-4o1, etc., and under this licensing model, the data would need to be re-licensed to cover each successive model, or the license adjusted to include scope for all of them.

While the publisher’s data may be used to train the LLM, this doesn’t mean the LLM can openly use the data in inference for representation or recreation.

These days, training for the closed-source LLMs takes copyright into account — like the example I provided at the beginning of this series— if they don’t have permission, they don’t disclose the details (example).

Researchers learn a lot about an LLM after it is published and they study how it is used, and continually make improvements from one generation to the next.

For instance, they may learn that some parts of the training data are repetitive and therefore unnecessary, or there may be gaps in the training data, requiring them to acquire new data.

Each new LLM leverages its unique training dataset, which suggests new licensing considerations for each successive generation.

Credit

2 scenarios using our piano analogy (from the “Beginner’s Guide to Understanding Generative AI”) to help illustrate the challenges of providing “credit” in training, and why the only real way to provide credit in training is at the model level (i.e. the digital publisher partnership announcements previously discussed).

Scenario #1: Inability to provide sources

If you were to perform the piece you memorized for your friend, they may ask you for details about the song.

You could tell them the composer, the title of the song, and perhaps the characteristics of the song, but you couldn’t provide them with a copy of the sheet music itself.

Your friend would have no way of confirming the details — they wouldn’t be able to jump to a specific page in the sheet music to see if you played a specific part correctly, check if you hit every note, or even learn it for themselves.

They could only take you for your word and would have to look up details on their own if they were sufficiently invested to do so.

Scenario #2: Training soup makes exact attribution impossible

Suppose the first song inspires you to continue to practice. You learn hundreds of other songs and eventually start to compose your own music.

The skills you learned in practicing all of the songs helped you to learn how to compose, but the music you write is completely your own.

Then one day, you perform your new music for your friend. When they ask about the song, you tell them it’s an original you wrote.

You might be able to share some of the songs that inspired your style, but ultimately, what you’ve created is something new — providing exact attribution is nearly an impossible feat.

These scenarios illustrate the challenges an LLM would have in providing credit based on the training data alone.

However, for the sake of argument, pretend an LLM could provide credit for every single word it generates by telling you exactly every piece of training data that contributed to selecting that word.

You’d still be left with the challenge that the “credit” is self-referential. Like the pianist's inability to provide source material for their music, the LLM cannot provide the source of the data because that’s just not how they work.

They can only generate accurate predictions of the data, there is no way for them to provide sources or external references to the source material.

For this reason, unless an LLM can point to an external data source, credit is unlikely to be provided via training alone — and so similar to the ChatGPT example from before — LLMs will likely continue to avoid scenarios where permission of representation or replication is required to display content.

Digital Publisher Content Licensing: Inference

This section provides an inroad to illustrate the power of pairing inference with training.

Consent and Compensation

In the partnership with Google, Reddit provides access not only to historical data but also to all the current data on their site. This is how they can offer up-to-date results to any query.

The way Reddit governs access to all of its data is via its API.

If Google violates the terms of their partnership or fails to properly compensate Reddit for their data, like the piano music, Reddit would not be able to extract their data from the LLM itself since they already licensed their data for training, and the training is now complete.

But, they would be able to restrict Google’s access to all of the historical and recent data used in search — because that’s how an API works — an API is all about controlling who has access to what data, and ultimately governing how the data consumer uses that data.

For everyone else, Reddit has a consumption model in place — where they charge for every time someone accesses their API.

Credit

This, from my perspective, is one of the greatest strengths of inference.

Because inference requires feeding source data into the context window, it makes it easy to determine the source of the response back from the LLM when it references it.

Back to the piano analogy.

From all of your practice, you’re now able to play the piano by sight — meaning you may have never played a piano piece before, but now because you’re so good, you can sit down to a new piece and play as if you’ve been practicing it for years.

Before you sit down to perform for a group of people, one of your friends gives you a piece of sheet music from an artist you’ve never heard of, and asks you to play it.

Never having practiced it, you’re able to perform it perfectly for the audience, perhaps even adding in some of your personal distinct style flourishes as you play it.

When someone comes up to you afterward and asks about the song, you’re able to pull out the sheet music and point to it for them — your source is readily at hand, as it provided the input for your performance.

For an LLM this is similar.

With Google, when you ask a question, not only can they respond with the AI Google search result, but they can easily provide links to the source (like a post on Reddit) they drew their data from so the user can go back and reference the original content. They can do this both for new data as well as data used to train the original model.

Inference with RAG is a powerful resource when it comes to leveraging AI with private data sets. It allows the data to remain secret and separate from the LLM, yet still provides a way for the LLM to use it and generate a response.

This is likely how digital publishers and AI companies will continue to address the concern around “credit”.

The devil is in the details

I am not oblivious to the fact that this last section is starting to get into the weeds. I’d apologize, except the fact that you’ve read this far means that you must be eating this up.

That said, this is the level of detail required to understand what needs to change to enable books and AI to have a future together — so please stick with me.

The key points from this previous section have been summarized in the matrix below.

Thus far, we have seen that licensing of digital publisher content is becoming more and more common — the partnership models appear to be working.

We have examined how the primary concerns of digital publishers are likely addressed in those partnerships.

In the next article, we’ll look at the differences between digital and book publishers, and what needs to be different in their AI content licensing partnerships.

This was the second article in a four-part series discussing the Future of AI and Books. If you’d like to continue reading, click the relevant link below.

1. AI and the Future of Book Publishing
2. The Current State of AI and Publishing
3. Why the Established AI Content Licensing Model Breaks with Books
4. Vision and Call to Action for the Future of Books and AI

If you're a bit fuzzy on some of the AI terminology used in this article, I’d recommend reading my “Beginner’s Guide to Understanding Generative AI”.

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or connect with me/send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

This first article in a series of four sets the stage and outlines the key concerns that need to be addressed for AI and books to have a future together

A dark future for publishers…if nothing changes

“What do you think book publishing looks like 15–20 years from now?”

This question arose during a conversation I had a while back with a publishing consultant we’ll call “Alex”.

I remember the way Alex hollowly looked at me, then pensively gazed off of the camera while they gathered their thoughts before turning back to me and saying

“Jonathan, I fear unless something changes dramatically, AI is going to eviscerate book publishing.”

I was flabbergasted.

Alex, this incredible publishing techno-sherpa of the last 30 years, who helped the top book publishers navigate the rise of the internet, mobile phones, the dominance of Amazon, the evolution into digital publishing, and now AI, was worried if book publishers would even be around in 15–20 years.

I asked why, and Alex said,

“Well, I just am not sure people are going to want to continue to read books written by people; I think ultimately they’ll want to read ‘books’ written by AI.

Not now; AI-authored books are pretty awful.

But the technology is exponentially improving, and it won’t be long until AI-authored books are as good, if not better, than people-authored books.

At that point, why would someone want to pay $20 for something an AI could write just as well, on-demand, tailored specifically to the needs and interests of the individual, for free?”

Alex’s comment, “AI is going to eviscerate book publishing”, made this conversation stand out because it acutely captures the unspoken sentiment I’ve felt across many other conversations with people in the industry.

There’s a broad underlying fear about what AI could mean for book publishing, and I believe there is good reason to be worried.

Book Publishers’ legitimate concerns

One of the first lawsuits filed against an AI company by authors succinctly synthesized the general concerns, where they asserted their IP had been used

“[W]ithout consent, without credit and without compensation”

What does this mean?

• Consent, meaning their IP had been used without regard to the will, desire, or permission of the IP holder (i.e. the publisher or author)
• Credit, meaning where the IP plays a role in influencing how the AI works, no credit or acknowledgment is given to the IP holder to recognize their contribution
• Compensation, meaning the IP holder has not received any financial remuneration for the use of their IP, even though their work is providing value

As a test, I went to ChatGPT and created a simple prompt, asking it to give me the entire first chapter of The Art of War, word-for-word as it exists in the text.

If you compare the ChatGPT version against Project Gutenberg, you’ll find they are word-for-word identical.

Since The Art of War is in the public domain, concerns of consent, credit, or compensation are irrelevant; however, this example illustrates what is possible with the technology as it exists at this very moment.

This generative AI, or AI that generates multimedia content, used the IP without consent, without crediting the source, and without compensating the IP holder.

If you think this only works for IP in the public domain, I can assure you that is not true — and there’s a straightforward test you can try right now.

Go to ChatGPT and ask it to give you Chapter 1 of any book not in the public domain, word-for-word. If ChatGPT has been trained on it, it will tell you it can’t because of copyright. If it hasn’t, it will tell you it’s not familiar with the book.

AI won’t reproduce Harry Potter, The 7 Habits of Highly Effective People, or the Steve Jobs biography not for technical concerns, but as an attempt by the AI companies to respect the copyright of those who own the IP.

The bottom line?

For the last 5 years, I’ve been working at the intersection of books and AI technology, and over that time I’ve earned some deep insights into both sides of this conversation.

The bottom line is the concerns of the book publishers’ are valid — AI is capable of everything they fear, and more.

But just because the fears are valid does not mean the game is over. In fact, it is far from it — recall Alex said, “…unless something changes…”.

I believe Generative AI represents the next phase of evolution for books, and part of what is needed is a joint vision for Generative AI companies and book publishers to work towards.

This series intends to provide one version of that vision and address the questions of what needs to change to make it a reality.

Who is this series for, and why should I read it?

This series is for anyone who wants to understand Alex’s comment better — about what needs to change to stave off the forthcoming evisceration of book publishing — specifically targeting those in the industry.

With that common starting point, the remainder of the conversation will go as follows:

1. I shed some light on what’s happening with the many recent announcements of AI companies licensing digital publisher content, what they’re doing with it, and how it relates to Training and Inference
2. Analyze the differences between digital and book publishers, and why the opportunity for book publishers is not as clean as it is for digital publishers
3. Discuss what would need to change for book publishers to seize the biggest forthcoming opportunities in AI in a way that satiates the concerns of all interested parties
4. Paint a vision for what the future of AI + books could look like if the elements we discuss were put into place
5. Ultimately, I’ll share some suggested next steps regarding where I believe we should go from here, and how you can get involved

I’m extremely optimistic about the future of generative AI, books, and book publishers, and my hope is by the end of this series, maybe I can help influence your perspective as well.

Two critical concepts for the remainder of this series are “Training” and “Inference” in the context of Generative AI.

If you are not familiar with these two concepts, then I’d highly recommend you read my “Beginner’s Guide to Understanding Generative AI”.

Otherwise, please continue reading! I’ve included the next article links below in sequence of my recommended reading path.

1. AI and the Future of Book Publishing
2. The Current State of AI and Publishing
3. Why the Established AI Content Licensing Model Breaks with Books
4. Vision and Call to Action for the Future of Books and AI

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or connect with me/send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

Key AI concepts like “training”, “inference”, “hallucination”, “fine-tuning” and “RAG” are put into simple terms for anyone to understand and remember

A quick introduction

This is a companion article to my series on AI and the future of Book Publishing. It is written specifically for those in the book industry and provides critical context for the ideas I discuss in that series.

That said, the concepts here should be relatable to anyone. I wrote it without assuming the reader has any prior understanding of AI other than colloquial exposure to some concepts.

I focus specifically on how Generative Text AI works and the implications for the book industry, using examples and simple analogies to help the reader connect with and remember the content.

Enjoy!

Let’s first align on some basic vocabulary

One of my principles comes from The 7 Habits of Highly Effective People, “Seek first to understand before being understood”.

For this reason, we need to first seek an understanding of generative AI technology before understanding its implications for book publishing.

You may or may not be as excited to read this article as I was to write it, but I cannot sufficiently emphasize how critical it is for anyone interested in this space to have a fundamental grasp of how generative AI works.

For that reason, I’ve tried to make these ideas fun and easy to relate to.

I’ve taken liberties to omit key details for the sake of relatability — so to anyone reading this with a deeper understanding of AI, send me a note and we can discuss it further.

Some key vocabulary for this section:

• AI — Artificial Intelligence. A general “catch-all” term that refers to a system that can understand, generate, and interact via text, audio, video, etc.
• Foundational Model, aka “Model” — Set of mathematical instructions that have been programmed to accomplish a particular task. For instance, there are different types of “models” that generate text, images, video, audio, etc.
• LLM — A type of Foundational Model, called Large Language Model. Specifically works with understanding and generating text. The model that ChatGPT uses (called GPT3, GPT4, etc.) is an example of an LLM
• ChatGPT — A popular AI chatbot application many people recognize when they think of “AI”
• OpenAI — The parent company that built ChatGPT

There are many different fields and types of AI, but for this article, we will be primarily focused on LLMs as they have the most direct relevance to current state book publishing.

Artificial Intelligence today is not actual “intelligence”

Let’s be clear about something.

Let’s be clear about something: When you go to ChatGPT and ask it a question, ChatGPT isn’t “thinking.” Instead, it’s predicting the sequence of letters to respond with — based on the words you initially provided it (the “input”).

Someday we might have true “thinking, reasoning, considering” AI, but that day is not today.

Instead of thinking, ChatGPT uses complex statistical methods to take the prompt you provide, and then predict which letter makes the most sense to follow the previous one.

It turns out that people are pretty good at this as well; let’s look at some examples to illustrate what I mean.

Predicting Common Phrases

Let’s try a quick exercise. I’ll provide you with the first part of an English colloquialism with some minimal context. See if you can figure out how to complete each phrase (the answers are at the end of this article):

1. Your friend is about to star in a play. You say to them, “Break a ____”
2. You easily complete a complicated task. When asked if it was hard, you say “It was a piece of ____”
3. Someone tells you a secret, and you assure them, “My lips are ____”
4. You’re struggling with a project, but a friend reassures you, “Don’t worry, two heads are better than ___”
5. A friend offers to help you move, and you reply, “That would be a load off my ____”

Not knowing anything about you (and assuming you speak English), I’d be willing to bet you answered most if not all of them correctly.

At its core, an LLM does the same thing as your brain did in this example.

The main difference is instead of predicting the next single word like in these examples, they’re capable of extending the prediction into sentences, paragraphs, and entire narratives.

How is an LLM able to create its predictions?

The short answer is complex math.

I’m not yet qualified to give a longer answer (although this resource provides a fantastic interactive visualization of what’s going on under the hood), but I can give some simple examples to help conceptualize what’s happening.

The Magic Box Analogy

Imagine we have a magic box. When you feed a number into this magic box, it does something to it and then creates a different number as an output.

I’ll provide you with a couple of magic boxes with examples. Let’s see if you can figure out what the “rule” is for each of the magic boxes (answers at the end of the article).

Using the Magic Box #1 as an example, the way to think about these is “When ‘1’ is put in the box, you get ‘2’ out. When ‘2’ is put in, you get ‘4’ out.”, and so forth.

For each of the magic boxes, you need to figure out what the output would be when you put “5” into them.

📦 Magic Box #1

📦 Magic Box #2

📦 Magic Box #3

📦 Magic Box #4

How’d you do? Again, not knowing anything about you, I’d bet you did very well.

These examples are very basic, and even if you didn’t know the exact mathematical function, you were probably able to pick up the pattern easily enough that you could confidently answer the output from the magic box for nearly any input value.

What we’ve done here is simply illustrate the two major fields of science around LLMs: Training and Inference.

“Training” is where an LLM learns the rules of a system

Training is the process of teaching an LLM the rules of a system so that when given an input, it can accurately predict the output (which is called “inference,” and we’ll talk about it shortly).

Referring to our magic box example above, the box represents the LLM itself, and you, the reader, represent the engineer training the LLM.

The first few examples you feed the box represent “training” data.

You know the input and the output, and the magic box is trying to figure out the rules of the system so it can predict the output when it’s not provided.

Once the LLM has been trained and the engineer is satisfied with its performance, the LLM can generate a response for any given input — even if it hasn’t seen that specific input before — which is where Inference comes into play.

“Inference” is where the LLM applies its Training

Inference is the process of using the patterns an LLM learned during training to generate predictions based on new input.

Your brain already did this in our magic box example.

Once you learned the rules for any of the magic boxes, you could quickly predict an accurate response for any number I gave you — even if that number wasn’t in the training data.

“Hallucinations” are when an LLM makes up an answer during inference

When an LLM is trained, the training data cannot possibly cover every possible scenario.

In our example above, “5” was not part of the training data, but given the historical values, we can probably safely assume we know the rule for what the answer should be.

But what if that’s not how reality works?

For instance, what if we put “-1” into our magic box? Based on our training, we would assume the output should probably be “-2”.

But what if in reality, when inputs are below 0, the outputs look like this?

The rules that we trained ourselves on based on the original training data no longer hold, and we would have to retrain the model.

The key thing to understand here is that hallucination is the LLM's best attempt to respond to every question, even if it doesn’t know the answer.

Does an LLM recall training data for inference?

It is important to note that the LLM does not retain a copy of the data used to train it.

And by “copy”, I mean, when you’re writing a document on your computer, you save it on your computer or your cloud, and there’s an actual file that contains all the text of your document. LLMs do not do this.

Since the LLM generates a prediction of the next word (or letter) one at a time and it does not keep a copy of the training data, there is no way for the LLM to check its results to see if they’re accurate.

However, since the LLM used the training data to learn the rules of the system, it often accurately predicts responses that are extremely close, if not identical, to the training data itself.

Remember what ChatGPT did with The Art of War example from earlier?

In the Art of War example, ChatGPT was not retrieving the text from a file stored in memory somewhere; instead, it was predicting what the text of the book should be, based on its training data.

This is a critical point to understand as it pertains to book publishers and their concerns, which we will dig into further shortly.

An analogy to help illustrate the significance of this point.

Imagine you’re learning to play a song on the piano.

You purchase sheet music of your favorite artists’ top hits and practice it over and over again, committing the notes to memory.

After a while, you can play the piece without looking at the sheet music because your brain has internalized the pattern of notes, chords, and timing — in effect, you’re predicting each of the next notes in their proper sequence.

You can now perform the piece for your family or as a fun “party trick” with friends. You become known as the pianist in your social circles, reproducing the music from memory.

This is similar to how an LLM works once it has been trained on the data; once the model “knows” information, it cannot be deleted or removed.

Just like you learning to play the piano, you can remove the sheet music, but your brain still retains the training that will always allow you to recreate the music.

For an LLM, once the training process is complete, the data is now part of the LLM. There is no file sitting in a folder somewhere waiting to be hacked or deleted — it’s now part of the LLM just like music is now part of your brain.

How does an LLM stay current with new data?

The process of training an LLM is a distinct event — it has a beginning, a middle, and an end.

Once the training is completed, tested, and vetted, developers release the LLM to the public with much publicity and fanfare.

Since training an LLM has a “start date”, by default this means the training data can only be as current as the data available when it started training. So if an LLM takes 3 months to train, that means it will have a 3-month knowledge gap by the time the LLM finishes training.

In the early days of ChatGPT, there was a well-known issue with the knowledge cut-off date of September 2021. That was the most recent date for information to have made it into the LLM’s training data.

Training a foundational model n LLM like GPT4 costs tens, if not hundreds of millions of dollars and consumes thousands of GPU processing years (which is shortened into weeks and months through parallel computing).

For this reason, it’s impractical to retrain an LLM just to get new information into it.

Fortunately, there are multiple faster, lower-cost options for LLMs to work with data that was not part of the original training data.

The two main ways of updating a model after it has been trained are call “Fine Tuning” and “Results-Augmented Generation” (RAG)

What is “Fine-Tuning”?

Fine-tuning allows you to update an existing LLM without retraining it from scratch — think of it as giving the AI a focused lesson on a very specific subject.

While fine-tuning can generate new “knowledge” for an LLM, it is most often used to alter the model’s behavior.

For instance, in the earlier piano analogy, what if you wanted to change how you played the piece? Maybe adjust the tempo, improve the dynamics, or update the mood?

These changes can take place without you having to learn the entire song from scratch again, it just takes some time for you to practice playing it with the actual changes you want to make.

With an LLM, it’s similar. Perhaps you want it to understand or speak legal jargon better, or better align the LLM’s tone with that of your brand.

Fine-tuning allows a model to be updated in specific areas without retraining it from scratch.

What is “Results-Augmented Generation” (RAG)?

All LLMs have what’s called the “context window,” which is basically how much information you can pass as input into the LLM for inference.

An easy way to think about a context window is the short-term memory of an LLM.

In the early days, the context window was so small that if you talked to ChatGPT for longer than a few messages, it would start “forgetting” the first things you discussed with it.

Today, context windows are much larger and less of an issue than they were in the early days. Some of the latest models are so big they can fit multiple books’ worth of text into them.

The way RAG works is you feed the information you need the LLM to analyze into the context window, and then it provides a response based only on the information it received.

In most instances where people talk about “training a model” for some specific use case (e.g., a customer support client), this is what they’re usually referring to.

The model isn’t actually trained on the support data, but rather it has access to all of the information, and just like you do when you’re looking for an answer on Google, it searches through the data to find what it’s looking for, and then pulls it into the context window and generates a synthesized response.

A simple analogy to illustrate how RAG works is like a human customer support representative.

They may not know all of the answers off the top of their head, but they have access to a massive database of troubleshooting issues.

When presented with an issue, they can search it, find the relevant information, and then synthesize it back as a response to help customers answer their problems.

In this example, the customer support rep is the LLM, the database of troubleshooting issues is the data you feed into the context window, and the response to the customer is the response you get back from the LLM.

What did we learn about LLMs?

Understanding how LLMs process and generate text is essential to a productive conversation addressing the opportunity AI could play for book publishers.

In this section, we discussed:

• AI today isn’t actually “intelligent”, but rather good at predictions
• Two main fields of science govern how AI LLMs are built and operate: Training and Inference
• Training is the process of teaching an LLM the rules of a system
• Inference is how the LLM is used once it has been trained
• Hallucinations are when an LLM imagines a response based on its training data, yet does not match reality
• Training data is not stored by the LLM but recreated through advanced predictions
• A context window is like short-term memory and is the data used for inference
• After an LLM has been trained, Fine-Tuning can be used to update its behavior, and RAG (Results-Augmented Generation) can provide it with access to new information

By demystifying the technology, we are now ready to explore better how AI can be a powerful ally for book publishers rather than something to be feared.

I hope you’ve enjoyed this article! If you’d like to learn more about the implications of AI on the future of Book Publishing, please check out the rest of this series through the links below.

1. AI and the Future of Book Publishing
2. The Current State of AI and Publishing
3. Why the Established AI Content Licensing Model Breaks with Books
4. Vision and Call to Action for the Future of Books and AI

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or connect with me/send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

One day it happens—the Epiphany.

The idea you know in your core is going to change the world.

You can’t stop thinking about the idea. You lose sleep and become anxious that someone else is going to do it first.

The more you think about it, the more you talk about it. The more you talk about it, the more social reinforcement you get that it has to exist.

So you launch the Startup.

You get early traction. Your excitement attracts investors. You recruit a team.

Everything is going great until one day you realize it feels like the momentum is slowing.

Sales aren’t scaling as quickly as you’d hoped. Customers say they like your product, but then repeat sales are low. All the while, the bank account is draining.

Your team starts to spin its wheels. Frustration mounts. The energy you had at first seems to have disappeared.

What’s going on?

You Fell in Love with your Solution

Who can blame you? It was an incredibly compelling idea. But here’s the challenge.

Solutions are easy and fun—they make us comfortable. They give us hope and help us feel like we are progressing.

Frankly, they’re inspirational and can quickly rally others to your cause.

But the problem is, generating solutions without understanding the real problem is like treating medical symptoms without understanding what’s creating the symptoms. It may alleviate the pain in the near term, but it will never make the pain go away.

Instead, here’s what you need to do.

You Need to Fall in Love with the Problem

But here’s the biggest challenge with falling in love with the problem.

Falling in love with problems is really hard.

It’s the startup equivalent of committing to a long-term relationship, which requires tremendous work, versus having a shallow weekend fling, which hardly requires any investment at all.

Falling in love with the problem requires you to be comfortable with discomfort. It requires you to be ok saying “I don’t know” to investors, employees, and customers alike because you know admitting this doesn’t mean you can’t eventually figure it out.

Falling in love with the problem requires asking yourself and your customers the hard questions—no delusions. It requires getting closer to the problem than anybody could ever possibly understand—even your customers.

Stop the Struggle

Once you understand the problem better than your customers, then you’re in a position where you can educate them about the problem, the drivers, and the implication of what solving the problem could mean for them. This finally puts you in a place where you can deliver real solutions.

And here’s what’s incredible. Few things create greater team camaraderie and energy than solving real problems. And when you start solving real problems instead of symptoms, everyone knows it and can feel it—from your team to your investors to your customers.

So if you’re ready to stop the struggle, release your idea and focus on your customer and their problems. If you can fall in love with them, I can promise you won’t regret it.

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or connect with me/send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

Why Your Startup is Struggling was originally published in The KickStarter on Medium, where people are continuing the conversation by highlighting and responding to this story.

Henry Ford is famously attributed with saying,

If I would have asked people what they wanted, they would have said “a faster horse”

While it’s dubious that Henry Ford ever uttered these words, this quote continues to live on in the mouths of visionary entrepreneurs worldwide.

They commonly cite it to protect the invalidated vision of their venture—for the unspoken implication of the quote is “the customer doesn’t know what they want until they see it”.

For true “visionaries”, if they ask customers what they want, or try to sell them their vision, the heathens will reject their vision because “customers don’t know what they want until they see it”.

I say that’s horse shit.

The startup battlefield is littered with countless corpses of companies that tried to “innovate” a business into the market without talking to their customers.

The #2 reason startups fail is because there was no market for their idea.

The #1 reason is they ran out of cash and were unable to raise new capital. In my experience, this is just another way of saying they didn’t try to understand their market or didn’t pivot fast enough with what their market was telling them.

The people who use the Henry Ford quote to defend against their lack of market validation believe real innovation happens in a vacuum.

They believe they’re like Steve Jobs, who had such a strong vision of what the market needed he could will it into existence without talking to anyone.

Even Steve Jobs wasn’t THAT Steve Jobs.

He had plenty of failures, where he believed in something so strongly that he willed it into existence, only to see its dead husk fall amongst the other corpses in the startup battlefield.

Remember the Lisa? NeXT Computers? Apple III? Yeah, most people don’t. Jobs failed plenty of times, but his successes were so meteoric that it’s hard to see or remember the failures.

It’s much sexier to think of Jobs as being this singular visionary who could tell customers what they wanted, and billions of people would flock to that vision.

This interpretation of innovation hurts the field. It costs investors (and entrepreneurs) billions of dollars and countless years of people's lives.

How can you truly understand your customers well enough to solve their problems if you don’t ever talk to them?

Maybe people did say they wanted a faster horse

If Henry Ford actually did ask people what they wanted, it’s highly plausible they said “a faster horse”.

It matches their worldview and contextual history. They knew no better. Horses were their main mode of transportation.

But any entrepreneur who hears that response and thinks “all I have to do is build a faster horse” isn’t worth their salt and should find a new career.

One of an entrepreneur's superpowers is the ability to interpret pain.

If you ask 20 people how to improve transportation and they all respond with some variant of “a faster horse”, it is up to the entrepreneur to get to the root of the unmet need.

Focus on understanding unmet needs

In this instance, the unmet need is the desire to get from point A to point B faster.

Once you understand this is the real problem to solve, it completely changes the game—because THIS is the problem to innovate around—to come up with a solution that allows people to travel faster.

It’s also up to the entrepreneur to understand what other key factors are valuable to the consumer. Affordability? Comfort? Quality? Support?

In some instances, you may very well find that what customers say they want and the solution to their actual problem is the same thing—a faster horse.

However, most of the time, you’ll find that what they need is something they’d never dreamed of—like an automobile.

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

Henry Ford Was Full of Manure was originally published in Entrepreneurship Handbook on Medium, where people are continuing the conversation by highlighting and responding to this story.

I got sucked into my first Ponzi Scheme in 1996. I was 13 and had been on AOL for 2 months. I had no idea was spam was.

For all I knew, that email was deliberate, thoughtful, individually addressed — letting me in on this incredible opportunity that seemed too good to be true.

They said to participate, all I had to do was send a crisp $1 bill in an envelope to the 6 addresses in the email.

Upon completion, I was to add my name and address to the bottom of the list, knock the top person off, and then send the updated list to as many people as I knew.

This individual testified of the millions they had made, and all it had cost them was $6.

The opportunity seemed bulletproof with virtually no downside, so I went for it. I sent out my $1 to the 6 people and then forwarded the email to hundreds of people.

Needless to say, this story is the only thing I have to show for it.

In retrospect, the scam here was so obvious it’s painful — they may as well have put the subject line as “Opportunity to invest in a Ponzi Scheme!”

Years later I realized there’s a much bigger, more subtle Ponzi Scheme that millions of people, including myself, have devoted our entire lives to.

It’s called “The American Dream”.

What is a Ponzi Scheme?

Let’s quickly align our language here. A Ponzi Scheme is a form of fraud that builds on the promises of early investors to make a huge, rapid return on their investment.

The initial “returns” and testimonies may be legitimate, but the system is doomed to fail since it can only work so long as there is an additional market to draw from. Once a market is saturated, the system implodes on itself as there’s no new money to feed returns for the investor base.

I get that may be a little complicated, so let me simplify using my “Mail for Money” example above (or M4M for short).

Suppose you were the first person to come up with M4M (We’ll call you “L1” for Level 1). You pitch the idea to 20 people (whom we’ll call “L2”), they all “buy-in” and each sends you $1 in the mail.

As the first person (L1), you’ve now made $20 and your only investment has been the time it took to persuade L2 to each send you $1.

Now, suppose each L2 gets 20 additional people (L3) to send $1 to everyone in L2 and L1. This costs each L3 $2; $1 goes to L2, and $1 goes to L1.

At this point, L1 has made $20 from L2, and $400 from L3, for $420 total. The scheme continues in this manner until there are no new people to bring into the scheme.

Assuming there are ~333MM people in the United States and you could get everyone involved in M4M, the Ponzi Scheme only supports 8 total levels.

In this example, the most money any single individual lost was “only” $6. However, the TOTAL amount lost in L8 was $1.6Bn, or nearly 80% of the total money invested into the scheme. The bottom line?

It’s a zero-sum game. The total money in the system doesn’t change — only how it is distributed across the participants.

In a Ponzi Scheme — a few people win, a handful win HUGE, and the vast majority lose nearly everything

The few people who win end up being extremely visible. Their “success” drives them to either be incredibly vocal about it or they’re so “successful” they can’t be ignored.

The American Dream as a Ponzi Scheme

The idea of the American Dream started as a land of promise. If you worked hard enough, you could be anyone you wanted to be.

We love the narrative of the “rags-to-riches” stories; not only is our country’s history full of them, it is one of them.

Fundamentally, the drive to achieve the American Dream is based on the idea of accumulation and “more”; more money, more wealth, more cars, more houses, more technology, more subscribers, more followers, more likes, more luxurious lifestyles — in the words of Jerry Seinfeld — more everything!

We follow celebrities and influencers in droves.

We glorify the lifestyles of the rich and famous.

We worship top athletes.

We deify elite entrepreneurs.

From the eyes of the masses, these individuals are living “the dream”, and by following them on social media, television, movies, music, etc.; we too can live The Dream via proxy.

But here’s the rub.

The American Dream is unsustainable at scale.

The American Dream is a zero-sum competition

Similar to a Ponzi Scheme, the American Dream is a zero-sum, “winner takes all competition”, survival of the fittest.

In order for some to win, the rest have to lose.

In order for Kim Kardashian and Dwayne Johnson to hold the titles of top influencers with hundreds of millions of followers, the other billion Instagrammers cannot.

For Lebron James, Tom Brady, and Roger Federer to make hundreds of millions in their respective sports, the other thousands of athletes they compete against cannot.

For Jeff Bezos, Elon Musk, and Mark Zuckerburg to have net worths competing with the GDP of many countries, the rest of us cannot.

Of necessity, the unbelievable wealth and success of those living the American Dream is only possible at the expense of the many.

The American Dream is mathematically impossible at scale

While the American Dream is an ideal that encompasses more than just finances, fundamentally it is heavily based on one’s financial well-being — or stated differently — financial independence.

The question is, how much wealth is needed to achieve the American Dream?

So as not to dwell too long on this question, let’s suppose the true amount to live the “American Dream” is somewhere between the generic financial advisor 4% retirement rule on the US median income of $67.5k ($1.7MM) and the “Wealthy” amount from Charles Schwabs 2021 Wealth Survey ($1.9MM).

Stated simply, the “American Dream” lives somewhere in the $1.7–1.9MM household net worth range.

The US has $134.7 trillion in household net worth (as of this writing). If we divided that wealth up evenly among the 132 million households in this country, that works out to just over $1MM per household.

Immediately we have a problem.

According to my math, even if we distribute all the wealth in the US evenly across the entire country, that’s still $700–900k less than needed to achieve the “American Dream”.

Thus, in the richest country in the world, from a purely mathematical standpoint, the American Dream is impossible for all.

Relatively few live the American Dream

As you’d anticipate from a zero-sum, mathematically impossible scenario — the reality is, very few people in the US actually live the American Dream.

Less than 10% of the households in the US have net worths >$1.7MM.

This represents the snapshot of the current state.

However, were we to map this same analysis out over time, we’d find that the concentration of wealth to the top of the population is only accelerating.

We’re in the late stages of Capitalism, and unless we change something, what belief is there that that trend will slow down?

Competition is a good thing though, right?

Competition is one of the fundamental tenets of Capitalism.

Free markets plus competition drive the invisible hand which can inarguably generate significant benefits both for society and the consumer.

America rose from obscurity to the “greatest country in the world” on the back of Capitalism.

Lest you believe otherwise, I am a capitalist at heart.

However, I also believe in compassion, equal opportunity, and that an unbelievably significant portion of our station in life boils down to PURE, unadulturated luck.

I believe it behooves those whose luck coin fell right side up to support those whose luck coin did not.

And unfortunately, due to the way we currently interpret the American Dream, the odds of it falling right side up are continually diminishing.

Some even claim the odds are so low, the American Dream is effectively dead.

Nevertheless, I can’t help but wonder…

What if there were a way for Capitalism to play its part while simultaneously avoiding the massive inherent inequality that stems from the “winner-takes-all” iteration of the American Dream we currently live?

When everyone is a winner, nobody is — I get that — and that’s not what I’m positing.

Pure socialism — where everyone is a “winner” — is where innovation goes to die.

Competition is necessary, and therefore so are winners.

But just because there are winners, do they have to take it all in a zero-sum sort of way? Why have we inextricably tied “winning” with “take it all”?

What if the “winners” are just the ones who get to the proverbial finish line first, without precluding others from getting there at all?

What if there were a way to leverage an individual’s inherent self-interest and capitalistic tendencies that enabled them to win first, and then utilize the proceeds of their winning to accelerate the betterment of those around them?

A vision for the new “American Dream”

This is the world I want to live in.

One where the role of luck in regards to one’s financial station is neutralized or mitigated.

One where “winner” is separated from “takes all”, and replaced with “got there first”.

One where the American Dream is no longer measured by the number of zeroes in your bank account, but instead by the number of people you helped achieve their Dream.

One where you don’t have to be an American to achieve the Dream.

One where the quest for Ikigai overshadows the quest for financial independence.

For me, it’s not enough to strive for my financial freedom if my closest family and friends still struggle. I’ll only be able to Dream when those closest to me can as well.

What it is not

To be clear, I am not talking about a socialist state.

I am not talking about Universal Basic Income.

I am not talking about free handouts, charity, or government entitlement programs.

What I’m talking about is true Benevolent Capitalism.

The place to start is with “enough”

How does one action this?

As with most things, it starts by looking inward.

Instead of trying to “achieve” the American Dream and win with “How can I make more?”, one’s focus needs to become “How much is enough?”

This very simple shift in focus has tremendous implications.

Learn Lean Thinking

Rooted in Japan, the fundamentals of “enough” thinking started with Lean.

Whether you are aware of this, this style of thinking has already changed entire industries dramatically and continues to do so as society finds new ways to apply it.

From Lean Production in manufacturing to Lean Leadership in business; from Lean Startups in entrepreneurship to lean living via ideologies such as the FIRE movement and Marie Kondo and her bestselling book, “The Life-Changing Magic of Tidying Up”.

Companies such as Patagonia have found tremendous success pursuing this ideology.

Recognizing our planet's resources are finite, they aim not to sell the most products they can, but to produce the best products that last consumers a lifetime, thereby minimizing waste.

Their brilliant marketing campaigns such as “Don’t Buy This” work to educate the consumer on these facts.

What’s your number?

How many of us have actually sat down, mapped out our lives, and figured out where we want to be 10, 20, or 40 years from now?

How many of us have put serious thought into considering what we want our lifestyle to look like, who we want to live it with, where we want to live it, and then backed into the resources — and thereby the financial number — that would enable us to get there?

I like to think of this number as a “cup”.

Once a cup is full, there’s enough. And no matter how much more liquid you continue to pour into the cup, the cup will never be able to carry more than it does once it’s at capacity.

So what good does it do the owner to continue to try and put more in the cup?

Once a cup is full, it’s full.

Or in other words, what good does gaining additional wealth do you if you’re already satiated?

Identifying the financial number that’s enough for you is like determining the size of your cup, or the magnitude of your thirst. Some of us are satiated after 16 oz, others can’t get by without 128oz.

At the end of the day, I believe there is a cup that’s big enough to satiate most everyone’s thirst.

Similar to the impact Lean Thinking has had in the examples above, imagine what could be possible if we could bring “enough” thinking to the American Dream; the Lean American Dream, if you will.

A practical example: Jeff Bezos

Explore with me a simple thought experiment of what having “enough” could look like, and how it could change things.

As of this writing, Jeff Bezos is the 2nd wealthiest person in the world, with an unfathomable $202Bn net worth to his name.

Recognizing full well there are numerous legal, regulatory, logistical, and economic complications to what I’m about to share, suspend your objections for the moment and imagine the following scenario.

What if Bezos woke up one day and said:

You know what? $202Bn is way more than any single human can possibly need or begin to justify.

After thinking through the lifestyle I want to live and mission I want to pursue, I only need $10Bn in my “cup”.

That way I can put $8-9Bn towards Blue Origin and still have $1–2Bn to live my life. That’s more than “enough”.

Moreover, Amazon is nearly a $2 trillion-dollar company, and I didn’t make that happen by myself.

I’d like to further reward those most pivotal who helped make it possible.

So, Jeff fills his cup to the tune of $10Bn, and the remaining $192Bn flows over into the cups of the next 10 most instrumental people who helped make Amazon possible.

Likewise, suppose those 10 decide $100MM is “enough”, and once their cups are full they pass the remainder on to the next 10, and those 100 individuals (10 x 10) each decide $10MM is “enough” for them.

For those keeping score at home, that means there’s still $189Bn of the original $202Bn available for distribution.

Amazon has nearly 900k employees globally, and while $189Bn couldn’t do it for all of them, it could help 111k of them “financially happy” millionaires ($1.7MM, as previously referenced).

One-hundred and eleven thousand people. That’s a life-changing amount of “F-You” money for a lot of people.

Would Bezos’ life look tremendously different if his net worth dropped from $202Bn to $10Bn? Maybe…or maybe not?

Would the 111k individual’s lives look tremendously different if he decided to focus on having “enough” and to act on it?

Undoubtedly yes.

Then the question becomes, what good could those 111k individuals do in the world if each of them also lived with “enough”?

“My cup floweth over”

The Fountain of Champagne glasses is a wonderful visual representation of this concept.

Yes, I recognize the irony of the pyramid shape and that of a Ponzi Scheme.

However, there’s one enormous difference.

In a Ponzi Scheme, the massive “earnings” are funneled up to the top level of the pyramid and stay there.

With “Enough Thinking”, once the cup at the top of the fountain is full — no matter how much more is poured into it — it physically can’t hold anymore.

The excess has to go somewhere, and so it gets shared with those around it.

In Conclusion: “Financial Freedom with Friends, Family, and Fellows”

There are only so many “world-changing” ideas an individual will have in their life.

Focusing on “enough” is mine.

I call it “Financial Freedom with Friends, Family, and Fellows” or “F⁵” for short, and it’s one I’ve been thinking about for a while.

My intention for writing this article has been to get this idea out into the wild, with the hope that it spawns discussion, commenting, sharing, and ideally “hole poking”.

That’s the only way it’ll get stronger and stand a chance of actually making it into reality.

As this idea sits in your brain, you may find your thoughts steering you into the natural next question of “How do we actually make this happen?”

If so, stay tuned — because I have some ideas I’d like to test with you.

Thanks for reading! If you have any questions or would like to hear more about something specific, feel free to leave comments below or send me a note on LinkedIn.

Click here to book time with me during office hours if you’d like to chat.

This article is for informational purposes only. It should not be considered Financial or Legal Advice. Not all information will be accurate. Consult a financial professional before making any major financial decisions.

How to Fix the Ponzi Scheme We Call the American Dream was originally published in Making of a Millionaire on Medium, where people are continuing the conversation by highlighting and responding to this story.