Jan 31st 2023
Last week we joined a team at the AUB Innovation Studio to talk about the ethics of generative AI, and its place in the creative sphere.
There were some great discussions from a broad range of disciplines in and around design, with another round scheduled in the
coming weeks. I thought I’d post some of my current perspective on the technology, given the hype that’s going on around it in our broad industry.
So cutting to the chase, what do I think as of today? I think it has a place. Most examples I’m seeing at the moment in design are based around using words to direct software to scrape together image data, and create “unique” artefacts. I think that in this instance, at this point in time, there’s a limited use. Principally, a tool in the divergent design phase to help populate mood boards, spark aesthetic ideation in concepts, and drive some visual language options. That's not to say this doesn't go without risk. But more on that later.
It’s really important to not observe AI creating some form of media artefact and get confused with AI designing. I see a lot of people getting excited, and concerned in equal measure about the prospect of AI designing products of the future in totality. I feel that we are a long way from that given the limited evidence I see today. I’m seeing artwork generation, but I’m not seeing design. Can AI currently understand, and more importantly implement things such as form classification of volumes, proportion, layout, connotation, design measure etc. in any way? Can AI account for human physicality, habit or emotion? Can it look at what exists, critically asses its strengths, weaknesses and produce an improvement… yet?
In Midjourney or Dall.E for example I see a very clever mash-up of images, either sourced autonomously or as directed. Yes, some of these source images may have had the aforementioned considerations implemented, but that doesn’t mean that mixing them will result in anything good or usable off-the-bat. There’s probably a good cooking recipe analogy here somewhere…
Further, if there is no auditable trail of the source data, then is there risk of post-launch litigation where the unique artefact output from AI, turns out to be not so unique? The ability to back search is critical.
Beyond aesthetics, if the source data isn’t inclusive of problem statements that encompass human factors, functional requirements, manufacturing methods, or socioeconomic factors for example, how could it ever be that useful in design?
In a recent forum, Dr. Christian McLening described the difference between art and design being that design must compromise to solve problems and I agree. Design must account for limitations outlined in a specification, derived from synthesis of some form of data (observation, focus group etc), whilst still invoking feeling or emotion, whereas art must only deal with the latter. As the technology develops in the field of design and tools become available which tap into more than just images and words, I think AI's role could be interesting. When we look to generative design in engineering, the engineer focusses on input parameters navigating iterations, and testing. Design if far more nuanced than mechanical engineering. Could we ever see a machine attempt to understand subtleties such as stimulus of emotion? Could the designer eventually evolve to assume the role for director, orchestrator, or architect, where AI for design is a principle tool? It seems like a stretch… and also Marty McFly told me I’d have a flying skateboard by 2015. Can I have that first. Priorities, please.
Andy is co founder of Element Creative Ltd. A small agile UK design agency supporting in a diverse spectrum of industries from micro start-up to blue-chip.