Legal & ethical aspects of using DALL-E, Midjourney, & Stable Diffusion

In this newsletter, read about:

• 🕵️‍♀️ Top Image Generators: Legal & Ethical Considerations

• 🗞 News and Top Reads

• 📌 AI Art Tutorial: Midjourney V5

• 🎨 Featured Artist: Julian van Dieken

• 🖼 AI-Assisted Artwork of the Week

• 🤓 How to Get Started with Generative AI?

🕵️‍♀️ Top Image Generators: Legal & Ethical Considerations

In my previous two posts, I have been looking into the terms and conditions of various stock image sites and art marketplaces to discover their position on AI art. But what about AI image generators themselves?

Today, I want to compare major AI image generators – DALL-E, Midjourney, and Stable Diffusion – with regard to their legal, ethical, and financial approaches:

• Who built the AI model?

• How much does it cost to generate images with this tool?

• What images have been used for training?

• What are the in-built restrictions on generated content?

• Can you sell the images you generated?

Of course, these three tools also differ in terms of their performance, strengths, and weaknesses, but let’s leave the creative part for another time.

DALL-E 2

Who: DALL-E 2 was introduced by OpenAI, one of the leading AI research and deployment companies, which is also behind ChatGPT. OpenAI was co-founded by Elon Musk in 2015, but he resigned from OpenAI’s board in 2018, citing a conflict of interest with his AI-related work at Tesla. Recently, OpenAI has been actively cooperating with Microsoft, which gets priority access to all of OpenAI’s products in return for financial support and computing resources.

Costs: You get 15 free credits that replenish monthly. One credit will be debited every time you generate or make variations or edits. When you run out of free credits, you can buy more credits for a minimum of $15 USD per 115 credits. Since each request generates 4 images, you can basically generate 460 images for $15.

Training data: OpenAI claims that DALL·E 2 was trained on a database drawn from a combination of publicly available sources and sources that they licensed. They’ve also made an effort to filter the most explicit content from the training data.

Restrictions: DALL-E has a content policy in place that prohibits creating, uploading, or sharing images that can be harmful. This includes content related to violence, harassment, illegal activity, and hate as well as sexual and political content. There are corresponding filters set by the DALL-E team, and you might be banned for trying to generate restricted content.

Also, OpenAI requires that you proactively disclose AI involvement in your work and don’t mislead your audience.

Commercial use: Yes, you are allowed to sell images generated with DALL-E 2.

Midjourney

Who: The company was founded by David Holz, a researcher and entrepreneur based in San Francisco. According to his LinkedIn profile, David has over 110 patents and publications with over 4,000 citations.

Costs: Midjourney has a free trial plan and three paid plans, ranging from $10 to $60 a month. The higher the costs, the more GPU time you have for generating your images (3.3 – 30 hr/month). Note that by default, your images are publicly viewable and remixable, but if you opt for the most expensive Pro Plan, you may bypass some of these public sharing defaults.

Training data: In the Forbes interview, David comments on the training dataset: “It’s just a big scrape of the Internet. We use the open data sets that are published and train across those. And I’d say that’s something that 100% of people do. We weren’t picky.” He also admits that they were not able to seek permission from every artist and copyright owner: “There isn’t really a way to get a hundred million images and know where they’re coming from.”

Restrictions: Terms of service discourage “creating images or using text prompts that are inherently disrespectful, aggressive, or otherwise abusive.” Adult content and gore are also prohibited. Some of the corresponding text inputs will be blocked automatically. Violation of these rules is likely to result in a ban from the Midjourney services.

Commercial use: As long as you are a paid member of the Midjourney community, you own the assets you create. Hence, you are allowed to use these images for commercial purposes (i.e, sell, reprint, merchandise, etc.).

Stable Diffusion

Who: Stable Diffusion was built by the Stability AI startup in collaboration with academic researchers from the Ludwig Maximilian University of Munich and non-profit organizations (i.e., Eleuther AI and LAION).

Costs: Stable Diffusion is an open-source project and you can actually use this tool for free on your own machine if it has sufficient computing power.

If this sounds too complicated for you, Stability AI has also launched Dream Studio for image generation. You’ll start with 100 free DreamStudio credits, enough to create 500 standard images. When they are all used, you can purchase 1,000 credits for $10.

Training data: The Stable Diffusion image generator was built using a huge dataset of a few billion image-text pairs, scraped from the web. The justification used by AI researchers and startups for using copyrighted data is that taking these images for AI model training is covered by the fair use doctrine.

Interestingly, the data to train Stable Diffusion was officially acquired by researchers from LAION non-profit organization (funded by Stability AI). But having researchers collect copyrighted data and train the model, strengthens fair use defenses, and startups often use this kind of AI data laundering to keep a legal distance between data collection and their commercial products.

Restrictions: The latest Stability Diffusion update includes the adult content filter, limiting the generation of NSFW content. Dream Studio, accordingly, has a very long list of prohibited uses for its image generator, where sexual content is just one of many points.

But since the Stable DIffusion code is open-sourced, these restrictions can be easily circumvented, as you have possibly noticed from the tone of sexually-explicit content, generated with Stable Diffusion and now published online.

Commercial use: Dream Studio Terms of Service have very careful language in this regard, but it looks like you are allowed to sell the images you generated with Dream Studio.

Let’s Sum Up

Most of the AI image generators use copyrighted images but try to protect themselves legally, claiming that training AI models on copyrighted data falls under the fair use doctrine. DALL-E 2 by OpenAI seems to be in the safest position as they claim to use only publicly available sources and sources that they licensed.

Also, all top image generators try to restrict the generation of violent, sexually explicit, or otherwise harmful content. But the open-sourced code of Stable Diffusion allows many AI creators to circumvent these restrictions.

Finally, you own the images you generate with DALL-E 2, Midjourney, and Stable Diffusion, and hence, you are free to sell them or commercialize them otherwise.

🗞 News and Top Reads

• Bias and stereotyping are still huge problems for AI image generators like DALL-E 2 and Stable Diffusion:

  • The new tools built by researchers at Hugging Face and Leipzig University allow everyone to see how these image generators amplify harmful biases and stereotypes.

  • For example, adding adjectives such as “compassionate,” “emotional,” or “sensitive” to a prompt describing a profession will more often make the AI model generate a woman. In contrast, specifying the adjectives “stubborn,” “intellectual,” or “unreasonable” will in most cases lead to images of men.

• Images of the Pope wearing a white puffy jacket have gone viral online.

  • The image was first posted to a subreddit for the AI image generator Midjourney, but over the next few days, it spread on Twitter and other social networks.

  • Lots of people were tricked into thinking that this is a real Pope photo.

• 🥳 And here are images of Pope Francis taking over the Burning Man festival

📌 AI Art Tutorial: Midjourney V5

In this video, Matt Wolfe talks about all the changes introduced by Midjourney with the release of V5.

For example, did you know that you'll get better results in V5 if prompting in natural language rather than throwing a bunch of words and phrases?

And by the way, V6 is only two months away!💥

🎨 Featured Artist: Julian van Dieken

Julian van Dieken is a German educational filmmaker and photographer. He got lots of public attention after his series of AI-generated images depicting Superman, Batman, Wonder Woman, and the Joker riding the New York City subway. Also, his AI-generated interpretation of the Girl with a Pearl Earring took a place of Vermeer’s original painting in the Mauritshuis Museum, while the original was on a loan to a different museum.

🖼 AI-Assisted Artwork of the Week

🤓 How to Get Started with AI Art?

1. DALL-E: Creating Images from Text – introduction to text-to-image generation.

2. The DALL-E 2 Prompt Book – a guidebook by OpenAI that explains how to effectively right prompts to generate images across different domains (e.g., photography, illustration, art history, 3D artwork).

3. Best Midjourney Prompts – a guide that covers the basics of Midjourney prompts (e.g., which keywords to use to create abstract art, surreal art, minimalism, etc) as well as some more advanced options (e.g., keywords related to camera lenses and filters, imitating certain artists and photographers without using their names). Finally, they provide a list of 600+ creative text prompts for image generation.

4. Stable Diffusion Prompt Book – a prompt book prepared by OpenArt. The book discusses ideal prompt format, using modifiers to change the style, format, or perspective of the image, applying ”magic words” to improve image quality, adding negative prompts, and adjusting Stable Diffusion parameters.

Share Kiki and Mozart

If you like this newsletter and know somebody who might also like it, feel free to share this newsletter. Let’s have more people learn about AI art!

If you have been forwarded this email and you like it, please subscribe below. And welcome to the world of AI art!