Skip to content

Anthony McGuigan

3 posts by Anthony McGuigan

AI Art has a Compounding Impact

AI uses human-made art

Human-made art remains essential—even as we explore and embrace AI-generated work. That’s because AI models are trained on art created by people, not other AIs. For AI art to continue evolving in meaningful ways, we must continue to value and support human creativity. Unfortunately, AI’s rapid rise is making this less common.

How AI Demotivates Artists

One of the most troubling effects of the AI art boom is its demotivating impact on human artists. Many are losing the drive to improve their craft or are abandoning art altogether. This is harmful—not only for those of us who cherish human-created work, but also for the most passionate advocates of AI art. Without fresh, original human input (which, importantly, must be used ethically and with the artist’s explicit consent), AI models have nothing meaningful to learn from. AI doesn’t innovate on its own—it relies on human creativity as its fuel. Without that, progress stagnates.

AI’s Compounding Effect

As more artists become discouraged and produce less work, AI-generated art continues to flood the digital space. This creates a dangerous feedback loop: the most visible content becomes increasingly homogeneous and derivative, eroding the diversity and quality of art available online. This is a loss for both critics and supporters of AI-generated art.

Conclusion

Supporting and protecting human artists is vital—no matter where you stand in the AI art debate.

Poisoning AI art

Protecting art from being used to train AI

Many artists are frustrated their work being used to train generative AI without their consent. Ben Zhao’s group have come up with two solutions artists can apply to their art.

Defensive measures from AI

To help achieve this, Ben Zhao’s lab at UChicago came up with Glaze, a protective filter which you can apply to your art. While not making much difference how the art appears to a human eye, the idea is that it adds a subtle amount of noise which disrupts the ability for AI models to learn from the glazed artwork. This is an example of ‘Adversarial Perturbations’. See the paper here: Glaze

Offensive measures for AI

Rather than trying to protect your art from being stolen, Nightshade attempts to poison the model. Again, while looking very similar to the human-eye, this filter can have a large effect on the AI image output. Adding noise in an attempt to hamper the models ability accurately recreate images from prompts, it worsens the image output by the AI. See the paper here: Nightshade

What next?

This will continue to be a cat-and-mouse game for the foreseeable future. AI models will be improved to evade these measures, and in response, these measures will be updated to become effective again. Ben Zhao mentions this in an article here, in response to a paper where the authors overcome Glaze.

Studio Ghibli in seconds?

Studio Ghibli going viral

Studio Ghibli, the Japanese animation studio famous for their lovingly handcrafted movies and distinct watercolour style, are currently going viral on social media. You may be familiar with them from movies such as My Neighbour Totoro and Spirited Away, but many people are hearing of Studio Ghibli for the first time due to a viral trend of applying a ‘Ghibli’ style filter to images using OpenAI. The Google Trends data below gives a picture of the recent spike in interest.

Google trends

Twitter is now awash with posts attempting to emulate Studio Ghibli’s style in a massive variety of images. This ranges from simple selfies to the White House tweeting a controversial picture of a person being arrested by immigration officers.

Viral X post

White house X post

This trend raises some interesting questions, which I’ll broadly group into three categories:

  1. The integrity of art
  2. The intentions of the artist
  3. Stolen images for training data

The integrity of art

Studio Ghibli is famous for painstaking attention to detail and love of their craft; all frames are hand-drawn and painted with watercolour. Four seconds from The Wind Rises, a frame of which is shown below, took Studio Ghibli 13 months to animate (Hayao Miyazaki: 10 Years With the Master). At 24 fps, 4 seconds corresponds to 96 watercolour images handpainted just for one scene. Compare that to the mere minutes it takes to get AI to produce ‘Ghibli-style’ images – how can artists compete against this?

The Wind Rises scene

However, the studio Ghibli style is not actually what is being recreated. Just like how a tomato and an apple may look similar side-by-side at a squint, closer inspection would quickly reveal differences between the two. Similarly, this filter is perhaps what someone who is unfamiliar with Ghibli might consider to be Ghibli style. As Sindu argues on her substack post AI filters don’t replace art any more than instant ramen replaces food:

“We’re seeing cultural flattening where “Studio Ghibli style” has become a catch-all term for any anime-adjacent art with soft colours or watercolour effects. This surface-level imitation… reminds me of what happened with Van Gogh’s style—people often reduce his work to swirly strokes without acknowledging how those techniques expressed his unique vision of the world and his emotional state.”

While at a glance these images may pass as ‘Ghibli’ like, there are subtleties to Ghibli art being missed by these AI models averaging out features across millions of training images. Somewhere along the way the intricacies of Ghibli’s art are lost and we are left with an empty imitation.

The intentions of the artist

Hayao Miyazaki, the head artist of Studio Ghibli, has vocalised his dislike for the use of AI in art. In a 2016 video when asked about his thoughts on AI, he stated “I would never want to incorporate this technology into my work at all … I strongly feel it is an insult to life itself.”

I find it hard to believe many artists would appreciate their style being used in this way by AI image models. As with anything on the internet, people have been quick to take things to the extreme and have created ‘Ghibli’ images of events such as the assassination of John F Kennedy. This, I imagine, is not what the artists at Ghibli would want their name associated with.

Stolen images for training data

To train these models, vast amounts of art are scraped off the web. Much of this training data will be art that the artists did not consent to be used for this purpose. This has big implications for copyright, as now anyone can use AI to try and emulate the style of artists who have skilful honed their craft over a career, undercutting them with AI which stole their work. Ben Zhao, author of the paper from our last blog post, said he’s disappointed to see OpenAI take advantage of Studio Ghibli’s beloved style to promote its products. He co-created the tool Glaze, which helps artists protect their work from being used as training data in these AI models.