Creativity, AI and Cubism

Following on from our previous blog “Does AI feel pain”, which discussed the extent of AI’s knowledge, this blog is concerned with its creativity and how it compares to its human counterparts.

If I ask AI for a picture of an elephant that demonstrates attributes of that elephant in a new way, it will produce an image. The question is this: is the new representation novel, or is it an amalgam of images the AI has been trained on previously; in other words, does AI ‘create’ something new, or does it reuse other material?

Consider Pablo Picasso and Georges Braque in their studios in Paris in 1907. To that point there was a tradition of displaying three-dimensional objects on a two-dimensional plane, that is, a piece of paper or canvas, by using devices like perspective and shading. However, Picasso and Braque saw the two-dimensional representation as an illusion, and not representative of the object or landscape being portrayed. Consequently, their aim was to show things as they are, so they tried to give the viewer an accurate understanding of the landscape or object by showing it from multiple angles, in a single image, suggesting three-dimensions but without using earlier techniques.

There can be no doubt that Picasso and Braque ‘invented’ cubism, but did it appear out of thin air, or was it an extension of earlier artistic trends? Well, they admired Cezanne, and his use of small planes, flat shapes, and repeated brush strokes, often seeming to be painted from slightly different viewpoints, to represent figures and landscapes. They were also influenced by African tribal masks, Oceanian artifacts, and ancient Iberian sculptures that provided an alternative to Western naturalism, and helped the artists learn how to break down the human body and objects into flat, geometric planes. So, did they just plagiarise earlier work? Of course not, but they did use earlier methods to help them create something that seemed, to them, more realistic than idealised or stylistic art.

So, can my picture of an elephant generated by AI, which uses previous images as starting points be classed as creative? This is a philosophical question, but if I can find no ‘similar’ images elsewhere, it is, undoubtedly, new, although it might be possible to suggest it is not ‘great art’. But then, not all new ideas are either immediately accepted, or considered worthy of further exploration. However, what we should focus on, rather than the artistic merit of the image, is that the AI used logic to find relevant earlier sources and combine them in a new way. Whether is truly new, or interesting, is beside the point.

There will be fierce proponents of AI, and there will be equally vociferous detractors, but to look at the middle ground, here is a tool that creates something where before that thing did not exist, which is a step forwards if not the greatest. In the last blog, we looked at the breadth of ‘learning’ of AI, and here we have considered its creative aspects. In both cases, to see the benefit of the use of AI, we should not bemoan the lack of breadth of its knowledge, nor the perceived lack of creativity. Rather we should accept what it provides, but be mindful of its limitations and be prepared to check the validity of the data, information, or images it produces, in the same way we would if that material were produced by a human. After all, if people neither check the quality of an idea or its provenance, I might have just created the square wheel and discovered relativity.

In our next blog, AI, fridges and the conservation of energy, we will continue to examine different aspects of AI.