Getting to grips with ... Midjourney
A regular look at a prominent AI or algorithmic system or dataset
The yins and yangs of Midjourney’s AI image generator
Midjourney is a subscription-based AI programme created by a US-based company of the same name that creates images from textual descriptions.
Midjourney was available in open beta since its launch in July 2022 via a Discord bot on its official Discord.
Midjourney version 5 was released in March 2023 and version 6 in December 2023. Both are seen to have improved the quality and realism of images generated by the system.
The yang: Midjourney is seen to offer a range of benefits, including:
High quality images. Midjourney is renowned for generating high-quality, realistic images that can rival traditional photography and artwork in detail and artistry.
Speed and efficiency. The platform can generate images in real-time, significantly reducing the time spent on manual image creation - an efficiency gain that is particularly beneficial for industries that require rapid content production, such as marketing and advertising.
Customisation and flexibility. Users can create custom images by specifying parameters such as style, colour and content, making it easier to align visuals with branding or project themes.
Cost effectiveness. Midjourney can be a cost-effective solution, especially for individuals and organisations needing large volumes of images.
User-friendliness. Users can generate images using simple text commands, making it accessible to those without advanced technical skills.
Scalability. The AI can produce thousands of images quickly, making it useful for large-scale projects.
Community Support. Midjourney has an active community on Discord where users can share their creations, gain insights and get support.
The yin: Midjourney is seen to present numerous potential and actual harms, especially for people and organisations in a commercial context, including:
Privacy loss. Images created using Midjourney are publicly accessible, even if generated in quasi-private settings such as Discord, meaning users can inadvertently expose sensitive or proprietary work, as anyone can search for prompts or usernames to access a user's entire image history.
Discrimination. The training data for Midjourney reflects existing societal biases, which can result in the generation of images that perpetuate stereotypes or exclude certain demographics. This not only impacts the quality and inclusivity of generated content but also raises ethical concerns about representation in media.
Copyright loss. Midjourney's training data includes numerous copyrighted images. Users may unintentionally generate images that closely resemble copyrighted works without realising it, potentially leading to legal disputes.
Legal liability. Midjourney users face potential legal liability if their generated images infringe on copyrights or trademarks. While the platform claims that users can use generated images commercially, it does not provide any guarantees or support in case of legal challenges stemming from such use. This lack of protection places the burden of risk entirely on the user.
Creative skills loss. There is a risk of users becoming overly reliant on Midjourney for creative tasks, potentially stifling their own creativity and skills.
Employment. The more that groups of users become reliant on Midjourney, the more likely the system will result in the loss of creative jobs.
Incidents and issues associated with Midjourney:
Man's use of AI to "resurrect" grandmother stirs controversy
Midjourney refuses to create images of Black African doctors treating white kids
Images of Australian children are used to train major AI models
Study: Top AI image generators 'easily' produce misleading election photos
Text-to-image AI models tricked into generating violent, nude images
Generating an AI image consumes as much energy as charging a smartphone
AI image generators accept 85% of election manipulation prompts
Amnesty uses fake AI images to promote Colombia national strike
Deepfake image of Pope Francis clad in a Belanciaga puffer jacket goes viral
Vermeer 'Girl with a Pearl Earring' AI facsimile causes controversy
Artist's private medical image used to train LAION-5B dataset