State of the Art
Hundreds of years ago, the expert painters of the Renaissance had apprentices. The expert was free to direct the act and flow of creation to service the whims of the culture and their patrons, as they delegated work to the apprentice, who learned their techniques and skills from these mentors.
In many ways, art has changed. The tools professional artists use are different, and the work itself is often ethereal. Digital stylus instead of brush; .png instead of framed canvas. In a matter of months, we’ve seen our perspective shift dramatically on what is possible — again. With the developments in machine learning (ML), creating visual media has been made as trivial as asking for a picture.
Our Machine Apprentices
The current wave of applied ML has already disrupted the way that art is created. In professions requiring the the creative substantiation of concepts and ideas, their craft will increasingly be facilitated and assisted by AI tooling. Yet, how the technology develops is still being debated in the words and actions of those who build it.
The materials used to train the models powering this technology include content that many argue was included unethically, without consent. Yet, the world of intellectual property from the 20th century is a poor construct to understand the world of AI - in practice and in truth. As evidenced by the recent decision of the US Supreme Court in the Andy Warhol Foundation v. Goldsmith case, we have very poor guidance and consensus on what “fair use” means.
Arguing against fair use in the training of foundational models could actually work against many artists. Professional artists, in the course of commercial work, are routinely required to transfer copyrights to other parties. The biggest winners in a world where model creation is limited to copyright holders is not artists, but the large corporate entities who acquired as much content as possible before anyone, including artists, knew what would be possible with technology.
In many ways, despite the controversy still playing out in courts, open-source models provide a path forward for artists that protects them from having their access to these revolutionary tools restricted. These open models are largely considered to be “sub-par” on their own - They need additional training and fine-tuning to effectively achieve the highest quality results. The technology exists to provide professionals the ability to train these models further, on their own work, such that they serve as just building blocks to personalized tooling - And most importantly, provide the ability for those same artists to claim ownership over the generative model itself, allowing them to license it and use it freely for their own work.
Times are changing, and so too does our relationship with art. Yet, while we create differently, the art itself remains, and we find ourselves once again teaching apprentices to create our work. This time, those apprentices are our machines.
The Future of Art
In conversations with artists, one thing is clear - few understand how the technology works. They are not alone. Many misunderstandings of how “AI” works, generally, are leading to broad and misguided narratives taking hold in public discourse.
Many who have had their work used in the creation of these models will tell you that their style is only replicated in a shallow, hollow way by the unspecialized general tooling available - even when explicitly invoked in the prompt. While some may write this off as evidence of the forever subpar nature of AI artwork, I would propose that this is instead proof that to truly understand a specific perspective on art, these tools need an artist to train it directly.
If creatives are provided the affordance to train their own model, they have the power to produce a true apprentice. Further, despite the concern for the risk of their artwork being stolen for the same purpose, artists are the best equipped with the capability of truly teaching it their style. They have the ability to generate new work to better instruct it, and can use these proprietary protected datasets to distill their perspective, treating the model as a lens that reflects their style as the AI generates new work.
The possibility of exclusive ownership over these capabilities is afforded only by open-source models. This makes it absolutely baffling that legislation in Europe has proposed attempting to punish open-sourced technology in this space. We have an opportunity to change the future of technology to ensure equitable access to these capabilities. Change is painful, and there is no doubt that this technology is disruptive. However, there is a path towards a better outcome for artists, where human creativity is channelled and harnessed in a new way.
Our Responsibility
As a builder working in this industry, I will unabashedly state that the world of “AI” is imperfect.
There have been projects incorporating stolen code with disregard, and fine-tuned models created based on artists works that are distributed without consent or compensation. I’m committed to fixing those problems systemically.
I have long desired to put my effort toward building technology the right way, choosing the path of conscience over convenience in building a team and product that serves users well. I want to enable artists and designers to benefit from their work, rather than have them exploited by the system as commodities. It is our collective responsibility to ensure that technology, particularly AI, is not just a tool for progress, but a beacon for ethical standards, inclusivity, and respect for individual creativity and rights being designed into these emerging systems.
We are at the emergence of a dramatic societal shift. As we venture deeper into the age of artificial intelligence, we must not lose sight of the human element that lies at the heart of creativity. The journey towards better leads to a future where art and technology coexist, not as competitors, but as partners in a dance that celebrates the boundless potential of human creativity. The work will be difficult and challenging, but that end is more than worth endeavoring toward.