Like many others around the world, I have been playing around with AI image generators Midjourney and DALL-E the last couple of weeks. If you have not heard about these new technologies, they are digital tools used to generate digital images via text prompts. For instance, the image above is how Midjourney imagines “someone thinking about data ethics”, which appears to be a somewhat gloomy subject according to the generator.
In recent years, these type of generators have become increasingly sophisticated and impressive. They have now reached a point where they can imitate particular artists and genres to a high degree, and generate credible photorealistic images of real world objects and environments. This development has spurred online communities of AI artists, large-scale digital galleries, and a considerable amount of discussion about the nature and limits of art. Can generated images be considered genuine art? If so, then who is the artist? The programmer who made the AI or the user who wrote the prompt?
Besides delivering fascinating images and questioning the meaning of art, however, I also think these generators provide something ethically important to the world, which has largely been missed in discussions surrounding their development and pending release to the wider public.
AI image generators somewhat level the playing field between naturally gifted artists and people who lack natural artistic talent. Some people are just naturally brilliant at expressing their artistic ideas; some people cannot draw a stick figure even if their life depended on it.
Artistic talent is part of what the philosopher John Rawls referred to as the natural and social lottery. The innate talents that we are born with and the resources we are provided by our social environment presumably all play a role when it comes to possessing artistic talent. Consequently, there are people with good and bad fortune in the artistic lottery.
This disparity in artistic fortune not only creates inequality in terms of career and job opportunities but also severely limits the satisfaction that many (most?) people can achieve from the process and products of expressing themselves artistically. AI image generators are a game changer in this regard.
Moreover, art and illustrations tend to be expensive. AI image generators give more people the ability to create and use images, allowing a much richer use of art and illustration.
There are no roses without thorns, however. Like with most new technologies, AI image generators also come with a number of ethical concerns.
As mentioned in Wired, for instance, AI image generators may contribute to discrimination by reproducing harmful stereotypes acquired through data collections containing real life biases. To some extent, this concern can be mitigated through technological means. For instance, biases can be limited through supervised machine learning (e.g. by weighting particular data more or less), certain words can be banned from prompts, and certain words can be suffixed more or less randomly to inputs (such as –woman or –person of color) to achieve greater representation.
Adding particular words to prompts require some degree of sophistication, however, as the same words can sometimes perpetuate harmful stereotypes rather than mitigating them. Consider for instance an automatic addition of –woman to a picture of a nurse, or an automatic suffix of –person of color to an image of a gang member.
Another worry is that AI image generators can be used to generate misinformation. This is, at least to some extent, why DALL-E, for instance, does not allow generating images with famous people or celebrities. However, fake images are not just a concern for rich and famous people. Consider for instance how fake images can be used by scorned individuals to harass their former partners.
Perhaps even more worrying than any particular piece of misinformation is the general loss of trust in the authenticity of images and the plausible deniability of facts that comes along with the advancement of technologies capable of producing convincing fake material.
It is imperative to mitigate these concerns through oversight, bias mitigating tools, and continuous sophistication of technologies capable of detecting fake images. We should arguably also require some degree of declaration and transparency on the use of generated images, especially when it comes to photorealistic pictures.
Although such concerns should definitely be taken seriously, it is also important to celebrate the positive aspects of new technological advances. It does not have to be all doom and gloom. As I see it, there is a huge potential for AI image generators in terms of limiting the effects of the natural and social lottery of artistic talents, and in terms of providing more people with the joy and satisfaction associated with artistic expression.