Insights From The Blog

Is Midjourney actually Sentient? Should We Worry?

The recent fake image of Pope Francis, resplendent in a white puffer coat was proof for much of the tabloid press that AI is fast taking over and we should all fear it. The image which showed the pontiff in an apparently relaxed state of dress was of such high quality that few realised that it was a fake picture which had been created by the Midjourney art AI, and the Pope doesn’t actually lounge about in polyester on his days off.

Midjourney – actually the name of the Institute that developed the AI program, rather than the program itself – creates images based on natural language descriptions. Sometimes known as “prompts,” these are developed in a manner comparable to that of OpenAI’s DALL-E and Stable Diffusion, and it is a revelation that is changing the face of digital art. As with so many other fields, AI has become extraordinarily powerful, and is now generating images that are frequently finding their way into mainstream outlets. However, while many see this as an incursion too far, a growing number of artists are arguing that AI-generated art is simply another medium in a highly complex area of human endeavour. There are plenty of people who would argue that art should not be bound by rules, and anything goes, even AI-creations, while others are of the view that while anything goes, art should remain the preserve of humans alone.

An activity that is exclusively associated with humans, art is defined by anthropologists as the creation of something for the purpose of gaining the aesthetic approval of other people. But human motivations are rarely so straightforward. Artists operating in the modern day are acutely conscious of the monetary value of their work as well as the ancillary perks that come with achieving artistic achievement, such as celebrity and the possibility of worldwide adoration. According to this anthropological concept of art, the possibility that animals can create art is not one that can be considered. There have been many instances of animals held in captivity, such as cats kept as pets or chimps housed in zoos, which have been given the opportunity by humans to express themselves creatively through the medium of art. They might create works in the style of Jackson Pollock, but this is an induced activity rather than a spontaneous one, and it is unclear whether they have any creative intention in mind. The famed sign language chimpanzee Washoe would occasionally describe what each of her paintings depicted in relatively expressive detail, but couldn’t elaborate on the thought process behind it. But it is still art, so, if animals can produce art, why can’t sufficiently smart computer programs?

For many, a key aspect of this argument is that both humans and animals create from first principles and the art produced is the result of a mental exercise. If that definition becomes the accepted one, the AI programs such as Midjourney could not be considered to be true artists, but perhaps it is just a case of rigid definitions simply becoming flexible. Internet users are currently being besieged with artificial intelligence platforms that offer everything from an exact duplicate of the user’s voice to lengthy poetry about black holes that are made in a matter of seconds. These platforms offer these services and more. Yet, the origin of this entire movement can be traced back to photographs, which is also where the concept of Midjourney was born. After the tremendous success of OpenAI’s Dall-E 2, there has been an unexpected surge in the number of AI-based systems that aim to transform the average person into a contemporary version of van Gogh or Pollock.  But how does Midjourney and related programs do it?

Midjourney runs through the Discord website and relies on a series of short words (the prompts) that act as the basis for any developed artform created. The AI does this by acting effectively as a search engine that cobbles together the terms and uses them to hone the resulting image to try and best match all of the terms. Because it uses AI to search, using the same terms can generate significantly different final images, but whether that actually makes Midjourney sentient or not is currently a moot point. Current thinking is that it is very clever at what it does, but is not really considering the output in any intelligent way, and is unlikely to pass a Turing test.

The technology that enables tools such as Midjourney to function is becoming progressively more complicated and advanced. As a direct consequence of this, the tools, and consequently the results, are continuously improving. This means that it is predictable that at some point in the future, it will no longer be possible to make photographs that are usable, but it will still be possible to create images that range between good and very good with very little input work. However, it does have its downsides, and these can be amplified by a poor choice of prompts, and once developed, an image cannot currently be tweaked if you don’t like something in it.  Usually, the only option is simply to start again and hope to get similar results the way that you want them.

Don’t get me wrong; Midjourney is a powerful tool and it will help many people create eye-catching works of art, but it’s not perfect and it can only produce art based on pre-existing images. An artist uses the most powerful computer on the planet – the human brain – and imagination is limitless. Additionally, images created by Midjourney tend to have a slight fantasy feel to them, which may fit the intention but if not, can become annoying and a little twee.

AI like Midjourney will evolve to be able to produce increasingly viable and engaging images, but for the moment, professional artists don’t have a huge amount to worry about.