Technology & Innovation - Issue 11

Can AI do artistry? Hannah Day considers the role that AI could eventually play in art and design education – and the problems it’s causing in the here and now... T his term, as part of the graphics A Level, we looked at colour theory, for which students built colour palettes. One student, rather than selecting and refining his collections, sat back and let the computer do it. I asked himhow. “Oh, I just built some code...” he responded, casually. We then watched as his screen filled square after square with various tones. On the spot, I had to decide what I thought and how I should react to what I was witnessing. Consider this My issue with this student’s work was the overall lack of selection or refinement . He was merely accepting a result that some highly advanced software was giving him, rather than evaluating work he had produced by himself and changing it accordingly. In the event, the colour palettes ‘he’ produced were ultimately bland, and the work his code had generated garnered the poorest outcomes in the class. Artificial intelligence, as applied to the field of art, currently works by taking all the information it can access, and using this to produce ‘new’ creations when prompted to by human input – but it’s a process that ensures the end results will only ever be average. In this case, the student’s colour palettes weren’t as good as they would have been had he simply done the work himself. That said, a student of art will typically explore how different paints, brushes and marks can change a painting, If that student then uses different AI tools and prompts to create a range of outcomes, selects the best and uses them to further develop their own work – how is that process any different? But then there’s the matter of referencing. Any text that’s repurposed and used in an essay must be clearly referenced and cited. It therefore follows that any use of AI in image creation should be similarly highlighted and explained. What tool(s) were used in the work’s production?What prompt(s) did the student enter? Howmany outcomes were generated, and how did the student approach those outcomes? If those elements of a work directly ascribable to the students’ actions, and those that are the direct result of AI technology could be separately and clearly recorded, that would at least leave open the possibility of assessing that all-important reflection and refinement. ‘Dead’media Finally, there are the implications for research. AI can easily generate false information – as neatly illustrated by a photograph I recently saw picturing Picasso, Basquiat, Dali and Warhol all sharing a drink together. I knew immediately that this was an AI-generated image, but students just starting out on their art history journey likely won’t know that these four artists never met together. To prevent false information appearing in essays and being cited as genuine research, we’ll need to better educate our students in the skill of checking sources, and the importance of locating multiple examples of research to back up their assertions. In contrast, there’s the humble, yet also mighty book. The thorough editing and fact checking undertaken by publishers before releasing any title far outstrips the comparative ‘anything goes’ wild west of online publishing. With the internet becoming ever more swamped with unreliable sources – which are now feeding AI tools that in turn produce yet more unreliable content – perhaps in time we’ll see a true resurgence of books and other printed matter. Manipulating the paintbrush So what does an actual artist make of all this? I wanted to talk to someone working across both the tech and art sectors and get their perspective – which is how I found myself in the studio of Dan Catt, who considers himself to be both an engineer and an artist. Dan writes code, which he sends to a mechanical arm that he built himself. Said armholds a fountain pen, the movements of which produce beautiful lined drawings. This, however, is not an example of AI art, because Dan is actively telling the armwhat do to. The process is functionally the same as what your own armwould do whenmanipulating a paintbrush. The only difference is the use of technology being placed front and centre. In Dan’s view, we’re all trained to varying degrees in the history of art, just as an AI would be, since nothing or no one can work fully independently of the canon of imagery we have. The difference is the speed at which AI can absorb and assimilate this data, and the immediacy and range of its outputs. He suggests that this can be seen by setting an AI a relatively simple task: “ Take light, for example. AI can generate an image of each individual student’s home town, and explore light in the style of 10 different artists. Suddenly, you see the same place specified by the same prompt, but rendered through very different approaches, making comparisons and analysis much simpler. ” Never go‘whole code’ Another of Dan’s observations concerns idea generation. If you have a student wanting to create a poster for a music festival, they could ask an AI to generate 100 different “Any text repurposedandused in anessaymust be clearly cited; it follows that anyuse ofAI should be similarlyhighlighted” 50 teachwire.net