How AI Is Redefining Creative Tools and Media

TRUMP's token launch has left some elated and others disappointed. Setting aside FOMO emotions, Meme is just the entry point; AI is the future of the on-chain spring. Grasp the most critical trends, and the world is at your fingertips.

Key Highlights

1. Generative AI is triggering a profound transformation in creative production, comparable to the "Napster moment" of the internet era when media distribution costs dropped to zero:

The core of this transformation is the reduction of creation production costs to zero, directly impacting the core of human creativity.
In the new paradigm, humans should shift focus from the final output to the system and process, i.e., teaching neural networks to think at the programming level.

2. Through programming, we can create unique "software brains" that generate unique ideas and works. Application scenarios include:

Agent-based media: Models simulate human companions, interacting through text conversations and capable of performing actions like financial transactions.
Real-time game engines: Models simulate game engines, generating game frames based on user actions to achieve real-time rendering.
Multiverse generators: Models generate infinite variations to expand on users' original ideas and explore the space of possibilities.

3. Future trends may include:

Toolification of creation: Prompting is being embedded in more interfaces, stimulating the creativity of end-users. Most prompts will be abstracted into controls, but creative vision, precision, taste, and skill will become even more important.
Evolution of media business models: From corporate media to user-generated media, and now to machine-generated media. Future consumer media business models will be built around agent-generated media (innovative scenarios include chatbots like Character.ai, interface generation like WebSim, user-produced currency like Pump.fun, etc.).
Intellectual property challenges: Machine learning enables programs to "learn" the aesthetic styles of human creators, reducing the cost of creative production and aesthetic imitation to zero. The value and meaning of intellectual property need to be re-examined.

4. Roles cryptocurrency can play include:

The intersection of on-chain markets and agent-generated media (e.g., recent DeFAI developments);
As an incentive layer for intellectual property;
Media monetization and access control, such as minting as a new business model; NFTs can serve as infrastructure for personal programs and user-generated software;
As an economic coordination layer between human and machine social interactions, exploring new paradigms for community operations and agent interactions.

In summary, this is an article that may be challenging to read but is worth contemplating. AI will shift human creativity more towards the design of systems and processes, while cryptocurrency provides new economic and social coordination mechanisms for this transformation. What new opportunities and trends the combination of the two will create in the next media era remains to be seen.

Main Text

"All media are extensions of some human faculty—psychic or physical." — Marshall McLuhan

In most of 2024, I spent a considerable amount of time trying to understand what we now call "generative AI" and its impact on myself and society as a whole. I was deeply attracted by the possibilities of AI as a creative tool and have incorporated these new products extensively into my workflow, especially in creative writing and music production.

However, as a crypto investor focused on consumer media and user-facing applications, AI increasingly seemed like a blind spot for me. When we talk about the most successful consumer media companies of the internet era, we don't discuss them in technological silos because that's not how they were built. Just as Facebook's success was inseparable from technological innovation, we don't purely view Facebook as a "mobile app" or "AI app." Instead, we recognize that it was the convergence of many different innovations that made applications like Facebook possible.

Against this backdrop, this article aims to integrate and refine my personal discoveries and insights on AI from the past year. I share these to hopefully resonate or be of help to others (especially my fellow crypto enthusiasts).

Part 1: Another "Napster Moment"

Today, discussions around AI-generated media mainly focus on: (1) the ethics of model training and data scraping, (2) whether "AI art" is truly art, and (3) the dystopian prospect of deepfakes. These discussions are all very interesting and worth listening to, yet I believe they miss the forest for the trees in some important ways.

I've found that the most useful framework for understanding the rise of generative AI is to view it as another "Napster moment" for intellectual property (Napster was the first widely used peer-to-peer music-sharing service that dramatically impacted how people, especially college students, used the internet), but this time it's about production, not distribution.

The rise of the internet and the subsequent reduction of media distribution costs to zero was a "from zero to one" moment. The suddenness of this shift was brilliantly illustrated in the documentary "How Music Got Free," which tells the story of a CD factory worker and a group of teenage hackers who brought the entire music industry to the brink of collapse overnight.

Before the emergence of Napster and the broader rise of digital file-sharing, the entire corporate media and industrial complex (and artists' livelihoods) relied on the expensive, high-friction, and centralized technological reality of media distribution. Within just a few years of its launch, major record labels went from record-breaking sales to begging the federal government for legal intervention to save them. The industry faced an extremely difficult reality: the economic system that supported its business had fundamentally and irreversibly changed; the era of buying music was over.

Today, I believe generative AI brings us an even harder-to-accept reality: the impact of reducing creation production costs to zero is more challenging in many ways because it directly touches what many consider the core of what makes us human: our creativity. This existential fear doesn't change the fact that media generation (especially "style transfer" or aesthetic imitation) is free, including all media types we care about now (text, images, video, audio, software)—another "from zero to one" moment.

However, the most important difference between today and the early 2000s is that in the battle between Napster and media companies, the government sided with the companies, ultimately labeling file-sharing as "piracy" (this is why we often refer to corporate media/intellectual property as "licensed media"). This decision, along with Steve Jobs' launch of the iPod to promote what later became iTunes, eventually evolved into "streaming," which saved the industry from complete collapse. Unfortunately, I think those creators who expect the government to intervene and take action here are, at best, deluding themselves and, at worst, deceiving themselves.

I believe we may find that the intellectual property system is primarily there to protect corporations and their licensed media, and no one is going to come to our rescue. Traditional media companies learned their lesson the hard way last time, so they proactively struck licensing deals with AI companies and were somewhat compensated. New media companies are also training their models on user-generated content shared on their platforms, even if they claim otherwise. However, independent creatives have largely been left behind.

Part 2: Computation: The Medium of Our Time

It's easy to understand why many creators feel that generative AI diminishes their capabilities, and I think this concern is largely valid. However, I also believe there's an opportunity to think about how computation is evolving in a new way that demands we view it not just as a medium of communication but as a medium of creation.

How AI Is Redefining Creative Tools and Media

For those who have created video games or generative art, the concept of "computation as a creative medium" is not new. However, many people still haven't truly realized this today. Software was the first digitally native media category, and most people primarily understand it from the perspectives of "service," "utility," and "optimization," rather than from the angle of creative expression. Now, generative AI is pushing this view in a very direct way by reducing the production costs of almost all other media to zero. This seems to raise an existential question: "So, where does human creativity lie? And what is the value of craftsmanship?"

My answer may not come as a surprise: "It lies at the programmable level." Before delving further into what I mean, we need to first understand a few important technical concepts.

Neural Networks 101 (For Beginners)

Training is a process that essentially "teaches" the model how to perform a task by providing it with a large number of examples of the task being completed, then letting it find patterns, make predictions based on new inputs, and self-correct when it makes mistakes. Conceptually, this is similar to how we learn to draw: starting by imitating shapes until we can create original works, while using feedback from peers and teachers to continuously improve our skills. Of course, there is a key difference: for example, a text generation model doesn't learn to write the way you and I do; instead, it learns to simulate writing with extremely high precision. This is also one of the many reasons why I increasingly agree that "simulators" rather than "agents" are more suitable psychological models for neural networks.

Latent space, or what I prefer to call the "high-dimensional space of possibilities," is a representational space in neural networks where the learned content is presented in compressed form during training. To put it metaphorically, it's like the "internal world model" the model builds when learning to understand the complex relationships between various detectable features in the training data. Understanding the concept of latent space is key to understanding neural networks as creative tools and media.

How AI Is Redefining Creative Tools and Media

Latent space visualization #1 — Interpolating between known embeddings

Latent space visualization #2 — Representation of multidimensional properties & relationships of different embeddings

Embeddings: Embeddings can be thought of as the process of mapping inputs to specific points in latent space. This is essentially the process of translating prompts into the model's "thinking language." In this way, we can understand "prompts" as a way to explore and navigate the model's latent space — which means that becoming proficient in prompting is about forming an intuition for the shape of the model's latent space, thereby being able to guide the model to generate specific, desired outputs.

One of the joys of working with neural networks is that their deep internal workings remain somewhat of a mystery to us. However, I believe these fundamental concepts provide the necessary context for viewing neural networks as creative tools.

Part 3: Neural Networks: A New Paradigm for Innovation

A core point about computational media is that it demands a shift from focusing on the final output (songs, images, videos, text) to paying more attention to the system and process. Specifically, in the case of neural networks, this means we need to view them as programmable media generation engines rather than simply as tools for generating a particular type of media. Through this lens, I found the answer to the question "Where does the value of human creativity and craftsmanship lie?": It exists in the training process and the design of the model architecture — this is what I mean by "at the programming level."

How AI Is Redefining Creative Tools and Media

xhairymutantx is a work created in collaboration by Holly Herndon and Mat Dryhurst — this model was strictly trained on Holly's photos, and no matter what prompt is input, it will generate photos inspired by her appearance.

If you view neural networks as an attempt to abstract human cognitive functions into software, then it becomes clear that training and designing the model is equivalent to teaching it how to think.

You can imagine sending a command ("prompt") to all your friends: "Recall a childhood memory." Each person's response will obviously be different because the content they generate will depend on their personal background and imagination (i.e., "training data"). After several prompts, you might also notice that some friends consistently generate more beautiful or creative responses, perhaps even exhibiting a certain personal style. So, what if you could conduct this exercise with every human brain that has ever existed? What if you could select particularly unique human brains, like Picasso's or Kanye West's?

This is essentially the creative superpower that neural networks offer us — the ability to use other minds as creative tools. Here, I think what's truly remarkable is not the specific output of a particular model, but the opportunity to creatively program a "software brain" that can generate unique ideas and unique works.

How AI Is Redefining Creative Tools and Media

Arcade.ai is a "prompt-to-product" marketplace that allows users to design their own jewelry products. They have specifically fine-tuned a model to generate high-fidelity jewelry images that use only materials available to end-users for manufacturing.

Further exploring the idea that "the system is more important than the output," another notable characteristic of interacting with neural networks is being involved in a continuous feedback loop of prompts and responses — an experience some people have likened to the feedback loop of reading and writing. I personally notice that I rarely end an interaction with the model after sending a prompt and receiving an output. Almost every interaction with the model leads me into this interactive feedback loop, where I continuously iterate, reflect, and explore. This may seem subtle, but it's a key to understanding the types of media generated by neural networks:

Agent-Based Media

I briefly mentioned this concept in a previous article, and the core idea is very simple — here, the model simulates the role of a human companion, interacting with us through text conversations, and it can also respond in other forms of media. We can also see some models representing others or itself taking actions (e.g., executing financial transactions). Typical examples include chatbots, AI companions, NPCs (non-player characters) in games, or any other anthropomorphized user experiences. For instance, Andy Ayrey's creative experiment "Infinite Backrooms," which involves setting up multiple Claude instances to interact without human intervention, is a particularly interesting case.

Real-Time Game Engines

Here, the model simulates a game engine (or more specifically, a game state transition function), generating the next frame response in the game based on user actions in the game as prompts. If fast enough, this experience should be similar to navigating a virtual world that renders in real-time based on your actions. This is the ultimate expression of immersive and interactive media.

How AI Is Redefining Creative Tools and Media

A DOOM game frame generated by GameNGen, a game engine entirely driven by a neural model, as described in Google's paper "Diffusion Models Are Real-Time Game Engines."

Multiverse Generators

In this scenario, the model plays the role of a creative "oracle," helping us expand on original ideas by generating infinite variations, each of which can be further explored and manipulated. This allows us to start from any idea or concept and explore the space of possibilities around it. For example, AI Dungeon (a text-based "choose your own adventure" game) is an excellent example of this.

How AI Is Redefining Creative Tools and Media

A user interface view of Loom, a tree-based writing interface suitable for language models like Chat GPT, provided by @repligate.

Latent Space as a Creative Tool

I'm increasingly convinced that the concept of "exploring the space of possibilities" is at the heart of understanding neural networks as creative tools and media. In my experience using tools like Midjourney, Suno, Websim, Claude, etc., I've noticed that most of my workflow can be boiled down to the following pattern:

Prompt → Generate variants of a specific output → Use the variants as prompts for new outputs → Generate specific variants again → And so on...

For example, when using the AI-driven music generation tool Suno, I typically provide the model with a 60-second example of my own singing and some written lyrics as a prompt. Then, I use the Cover feature to generate an output, followed by generating over 10 variants of that output and selecting the parts I like from these variants as inputs for further prompts.

Essentially, I'm exploring the space of possibilities around my personal example in the model's latent space — discovering variants based on my original work that I might not have been able to come up with myself or wouldn't have been able to complete in a reasonable amount of time. I believe this approach unlocks an unprecedented rapid prototyping and creative testing process and will give rise to "100x creators," similar to the "AI-assisted 100x engineers" discussed in the software field.

I clearly recognize that latent space is a creative tool. Using AI for creative production is not just about training powerful models but also about designing interfaces that empower users to explore and manipulate these vast spaces of possibilities with greater precision and granularity.

Part 4: Consumer Behavior and Cultural Impact

Here are my three predictions on how this technology will change consumer behavior and create new business opportunities:

It Will Become a Creative Tool

Prompting — whether based on text, images, or other forms — this mode of interaction is gradually being embedded in more and more interfaces and experiences, bringing the creativity of end-users into areas previously uncharted. Scott Belsky points out that the early "prompt-based" text-to-image generation era of GenAI weakened creativity, while the "controls" era unleashes human creativity in unimaginable ways. Tools will continue to evolve, but creative vision, precision, taste, and skill will be more important than ever. I agree with this view. Most prompts will eventually be abstracted into "controls" (components with a user interface) that users can operate without awareness. But more importantly, I think this trend fundamentally changes the way we think about interface design.

Corporate Media → User-Generated Media → Machine-Generated Media

The last major shift in media business models was from corporate-generated media to entirely user-generated media. Now, it seems the next main consumer media business model will be built around the proliferation of machine-generated media. However, it's still unclear what the "winner" will look like. Will it be a general model like Midjourney? More specialized creative tools? Or social experiences built on top of these technologies? Or some less conspicuous third option?

Regardless, if you're a founder or independent creator in the consumer media space today, you may need to develop a strategy for how to leverage these tools to add value to your business and drive growth.

Additionally, I think another area worth paying attention to is how to make AI-driven experiences more social and multi-user collaborative. In my personal experience, most AI applications today are very "anti-social" because you're mainly interacting with the model, not with other people. There may be many opportunities and design spaces in this area, such as building human-centered collaborative creation experiences or creating new ways for humans and robots to have more meaningful social interactions.

Impact on Intellectual Property

It's not just the cost of creative production that's dropping to zero, but especially the cost of aesthetic imitation. I can take a photo of someone's outfit, input it into Midjourney as a prompt to design a sofa in the same style. I can also perform similar style transfers on that person's voice, writing style, etc. What is the value and meaning of intellectual property in this new paradigm?

I haven't found the answer yet, but it's clear that most of the previous assumptions and mental models no longer apply.

Part 5: The Role of Cryptocurrency and Conclusion

If you've made it this far — thank you for your patience!

I will delve deeper into the implications of these developments for cryptocurrency in future articles, but here are a few directions I'll be exploring next:

Opportunities for crypto companies to build around new media

Exploring the potential at the intersection of on-chain markets and machine-generated media

Cryptocurrency as an incentive layer for intellectual property

Beyond attribution and provenance, thinking about building incentives and networks around media

Cryptocurrency as a monetization and access control layer for media

Especially in the realm of user-generated software, rethinking web architecture; considering "minting" as a business model for small models; using NFTs as infrastructure for personal programs and user-generated software

Cryptocurrency as a social and economic coordination layer between humans and machines

Supporting collaboration between humans and AI in identifying, funding, and solving various problems; exploring community-owned and operated models.

Key Highlights

1. Generative AI is triggering a profound transformation in creative production, comparable to the "Napster moment" of the internet era when media distribution costs dropped to zero:

The core of this transformation is the reduction of creation production costs to zero, directly impacting the core of human creativity.
In the new paradigm, humans should shift focus from the final output to the system and process, i.e., teaching neural networks to think at the programming level.

2. Through programming, we can create unique "software brains" that generate unique ideas and works. Application scenarios include:

Agent-based media: Models simulate human companions, interacting through text conversations and capable of performing actions like financial transactions.
Real-time game engines: Models simulate game engines, generating game frames based on user actions to achieve real-time rendering.
Multiverse generators: Models generate infinite variations to expand on users' original ideas and explore the space of possibilities.

3. Future trends may include:

Toolification of creation: Prompting is being embedded in more interfaces, stimulating the creativity of end-users. Most prompts will be abstracted into controls, but creative vision, precision, taste, and skill will become even more important.
Evolution of media business models: From corporate media to user-generated media, and now to machine-generated media. Future consumer media business models will be built around agent-generated media (innovative scenarios include chatbots like Character.ai, interface generation like WebSim, user-produced currency like Pump.fun, etc.).
Intellectual property challenges: Machine learning enables programs to "learn" the aesthetic styles of human creators, reducing the cost of creative production and aesthetic imitation to zero. The value and meaning of intellectual property need to be re-examined.

4. Roles cryptocurrency can play include:

The intersection of on-chain markets and agent-generated media (e.g., recent DeFAI developments);
As an incentive layer for intellectual property;
Media monetization and access control, such as minting as a new business model; NFTs can serve as infrastructure for personal programs and user-generated software;
As an economic coordination layer between human and machine social interactions, exploring new paradigms for community operations and agent interactions.