If we are to defend our creative data from being sucked into large centralized AI models, we may need a new ethos and behavior.

THE AI SOFTWARE ECOSYSTEM

In a recent interview, Sam Altman, the CEO of Open AI, foretold a vision of how companies with multimillion-dollar processing power will sit at the base of a new artificial intelligence software ecosystem. (1) A gold rush of smaller companies, will use the APIs and "tune" these massive models to provide customizations for specialized markets.

For example, if I were a comedy writer...

I might query a paid service that is built upon Open AI's chatGPT. The large language model would provide the sentences, structure, and logic. The service --- tuned with, say --- 80's stand-up comedy routines, could provide the timing and personality. The prompt would return a killer routine, with little effort from the comedy writer.

This is not fantasy, this is happening. (2)

In December, a panic among visual artists in the Art Station community led to many placing an iconic "No to AI-Generated Images" within their portfolios. (3) This month, (January 2023) the first of many class action lawsuits have begun. (4) Our government may attempt to make policy, but the angry artists are already mounting an attack on large tech organizations.

Even if traditional legal actions are successful, "slaps on the wrist" will do nothing. The graphics company, Adobe, has been aggregating user data to train their own AI models. The legal on Adobe's site points to the "opt-out" function, where users could select not to have their work part of a training set.

Of course, everyone on Adobe's Creative Cloud is opted in... by default. (5)

WHAT CAN BE DONE?

It's clear to me that my days of animation labor are numbered.

The models may be imprecise and difficult to control now, but from what I see in my research, I do not doubt what is to come. I can not opt-out or go silently into a data set.

I've begun to ask myself:

As a creative, is there something more tangible I can do than simply support traditional legal action?

Could I participate in a new ethos that would allow the productivity of AI tools, without surrendering ownership of my work?

As an environmentalist practices sustainability, perhaps generative artists can practice defensibility. I am pursuing three ideas, though a solution is not readily visible.

My instinct is that, perhaps, these ideas point in the right direction.

OPEN SOURCE

Open source is the act of making an idea publicly accessible. Open Source software, distributed for free, and combined with network effects, may yield new economic models for collective production. Wikipedia, Android, Blender, Apache, Linux, Python, Rust. These are not companies, but open networks of production. (6)

Open Source built the internet. It began peer-to-peer. It was messy, creative, and unstandardized. Software, free from control, can evolve. (7) Open Source becomes the public forum, the language for us all to use, where no one entity makes all the money.

Our new AI software ecosystem is being born, not peer-to-peer, but centralized. It too is messy, creative, and unstandardized. While we are developing this new ecosystem, perhaps we could allow the technology to mature, free from control.

Emad Mostaque's vision is just this. His company, Stability AI, builds open-source alternatives (like the image generator Stable Diffusion) in a hope that it can "de-risk" the centralization of AI from doing serious harm. Imagine if AI were a "public good."(8)

Stability AI is, in effect, another centralized company. What's different, however, is that I can download the model on Git Hub. (9) Because it is open source, many others have customized it. Using community-created code, I can tune Stable Diffusion to my needs.

This democratizes AI and empowers creativity across the world. This is why I skew open-source AI whenever possible.

But even if open, the art I create still needs to be ... my art.

PERSONALIZATION

AI needs a lot of data to train, but it also needs customized data to focus. I can prompt an image and get pretty close to the desired result. However, to prompt an image and make it look like me, needs custom tuning. (10) This requires customized data sets and affordable processing power.

Companies are now rushing to satisfy this personalization for users. With the APIs of the large models available, it's up to them to determine the specific market fit and provide that service.

But the concern of data collection still applies.

With AI models, you turn yourself into a superhero, but the cost may be, that your likeness no longer belongs to you. (11) For us to patronize these new entrants into the AI ecosystem, trust, more than the product, will become the critical selling point.

Recently, I discovered a company called Personal.ai. CEO, Suman Kagnuganti founded the company to focus on data privacy and ownership. Their product, effectively a chatbot system, is built on a sort of "vanilla" language model where you add your data. It is designed to be trained with personalized "memory stacks", so your bot becomes you. (12)

As a teacher, I can potentially put everything I know about animation technology into a personal AI chatbot. I spend a large percentage of time helping students through chat, how much time would I get back if all of my most asked questions were instantly answered?

But the value of this company, for me, is not AI. If I choose to use it, it's because Personal.ai has built its application on the Oasis protocol.

Oasis is built on the blockchain. (13)

BLOCKCHAIN

The blockchain is currently home to speculative collectors and retail investing "apes." It might soon mature into an automated lawyer that tracks our transactions. (14) There is a misconception that the cryptographically secure blockchain will allow for the "protection of a digital file," like a super un-hackable database. This is incorrect.

Its value is the immutable record that proves exactly, where and when, we added our creative labor. The dream is that we can defend our contribution on a transaction-by-transaction basis. An artist brought something to life, and an artist should get a piece of that something's intellectual property.

Artists are selling NFTs now, but beyond speculation, there is little economic utility. Theoretically, protocols are being developed that trigger when the virtual asset discovers its utility value. (15)

Ideas take time to gestate. The software takes time to mature. If art is not "ready," the components could be distributed and tracked until the time that it is. And should that rare collective "genius" strike, the contract could trigger at that moment and retroactively monetize the entire development process.

The blockchain space is a massive essay in itself. They are potentially more complicated to understand than the artificial intelligence models I hope they can decentralize. Despite the complexity, I have watched with optimistic hope, as the blockchain marketplaces on Ethereum and Solana argue over smart contract protocols that defend the royalty traits on Non-Fungible Tokens. (16)

I don't understand the specifics, but I know that somewhere within this debate lies a new way for artists to monetize their ideas. Within this debate will form new collaborations and royalty splits, and the collectivism in production will potentially decentralize our data away from singular interest.

If we figure it out in time.

WE NEED TO HURRY

Open Source. Personalization. Blockchain.

These are guesses, but we need to start somewhere... and we need to hurry.

No one has fully automated themselves - or have they? But as soon as we see others turn into productive superhumans, I am fairly sure, we will all be quickly motivated. No one wishes to be left behind.

And if we have no choice but to use AI to accelerate ourselves, we must behave in a way where we don't give our creative data to a centralized entity to do it.

If you believe this to be alarmist speculation, then I thank you for your time. If you believe, as I do, that this is a transformative period in human history, I welcome you to reach out. These are merely the guesses of a single animator.

With open debate, perhaps together, we can create a new ethos for a beneficial future -- if it exists.

Nye Warburton is an animation technologist and educator. This post was written in participation with the Bankless Academy Writers Sprint. This essay was written and drawn with the human labor of the author. (January 2023)

Visit online @ http://nyewarburton.com

Notes and Attribution:

1. Sam Altman Interview -

   https://youtu.be/WHoWGNQRXb0

2. "This is happening." - NovelAI -

   https://novelai.net/

3. "No to AI images on Art Station" -

   https://dataconomy.com/2022/12/no-to-ai-generated-images-artstation/

4. "The first of many class action lawsuits" - https://www.theverge.com/2023/1/17/23558516/ai-art-copyright-stable-diffusion-getty-images-lawsuit

5. "Adobe legal opting in and out" -

   https://www.artificialintelligence-news.com/2023/01/09/adobe-train-algorithms-your-work-opt-out/

6. “open networks of production” - The Wealth of Networks by Yokai Benkler -

   https://www.benkler.org/Benkler_Wealth_Of_Networks.pdf

7. “Software, free from control, can evolve.” - The Future of Ideas by Lawrence Lessig

   https://en.wikipedia.org/wiki/The_Future_of_Ideas

8. “Emad Mostaque's vision is just this.” -

   https://www.realvision.com/blog/emad-mostaque-talks-stability-ai-with-raoul-pal

9. “I can download the model on Git Hub.” -

   https://github.com/AUTOMATIC1111/stable-diffusion-webui

10. “However, to prompt an image and make it look like me, needs custom tuning.” -

    http://nyewarburton.com/2022/02/09/machine-learning-experiments/

11. “but the cost may be, that your likeness no longer belongs to you.” -

    https://www.businessinsider.com/lensa-ai-raises-serious-concerns-sexualization-art-theft-data-2023-1

12. “It is designed to be trained with personalized "memory stacks", so your bot becomes you.“

    https://www.personal.ai/how-it-works

13. Oasis Protocol -

    https://oasisprotocol.org/

14. “It might mature into an automated lawyer that tracks our transactions.” - Nick Szabo -

    https://en.wikipedia.org/wiki/Nick_Szabo

15. “Theoretically, protocols are being developed that trigger when the virtual asset discovers its utility value.”

    https://academic.oup.com/ijlit/article/26/4/311/5106727

16. “I have watched with optimistic hope.” - https://twitter.com/MagicEden