How Do We Picture AI in Our Minds?
Dec 11, 2023
Illustration by Neema Iyer
Recently, when I penned down my blog on the Use of AI in Fundraising, I turned to Midjourney for some inspiration for the illustration that I would create to accompany the blog. I tried some simple prompts like ‘AI creating proposals’ or such and received images of humanoid robots writing at a desk. I felt uninspired and decided to draw up a cover illustration that had two hands with an incorrect number of fingers, a disfiguration that is still commonly produced in generative AI art.
This got me thinking about what is the usual ‘mental imagery’ when we think of AI. What is the visual representation of AI in our mind’s eye? Intelligence is, per se, a difficult concept to draw, often portrayed as a brain or a light bulb. In this blog, I’ve decided to draw out some of the usual interpretations of AI. Perhaps, understanding AI through its visualisation could play a role in simplifying or demystifying exactly what it is and opens the door to a deeper comprehension amongst a wider audience, given the increasingly ubiquitous nature of AI in our everyday lives. Is it magic? Is it just ‘spicy maths’? Is it actually the invisible labour of human beings as in the case of the ‘mechanical turk’?
Calvin and Hobbes comic strip by Bill Watterson
The Classic - Humanoid Robot: You know the one, the sleek white robot that has come to symbolise AI. I think there’s a few reasons that this one has reached such prominence such as the deep-rooted human fascination with creating entities in our own image or that it makes such technology more relatable as it looks similar to us. In this case, the typical depictions of the humanoid robot have the features of a caucasian man. The clean, white aesthetic is also similar to that which Apple has consistently marketed to us could also signify ideas of purity, elegance, precision and futuristic advancement.
However, this has me thinking about “uncanny valley”, a term coined by roboticist Masahiro Mori in the seventies which refers to the phenomenon where humanoid objects that closely resemble humans provoke eerie or unsettling feelings, especially when they are almost, but not quite, lifelike. It’s believed to be rooted in evolutionary mechanisms and ancient survival instincts of disease avoidance and threat detection. Our brains are supposedly attuned to identify anything abnormal in human-like appearances as potential risks which could be a trait that was once essential to our ancestors' survival. I feel similarly when I see some types of generative AI art.The Brain: As mentioned earlier, AI is often portrayed as a brain, sometimes fleshy and human-like, or else, digital - with gears or microchip imagery - often in fluorescent blue against a dark background. Similar to the concept of intelligence, the metaphor signifies the capacity to think and to learn, and once again, ties in to a concept that most people are already familiar with, and makes a clear connection or linkage with humanity.
The Renaissance Hands: Here, the visualisation plays on Michelangelo's iconic painting, 'Creation of Adam', which shows two hands reaching towards each other - one human and one robotic, playing on the idea of a connection between humans and AI. This painting, a part of his Sistine Chapel fresco, depicts a scene from the Bible where God ‘breathes life’ into Adam, the first man, symbolising the beginning of life and consciousness,tying in with concepts such as knowledge and power. In the case of AI, we humans are the divine and in turn are bestowing ‘life/consciousness/knowledge’ to AI systems or robots.
The Magic Wand: It seems that magic wands are, interestingly, more compact versions of rods and staff which have historically been wielded as a symbol of power, authority and magic across Ancient Egyptian and Greek mythology to European Folklore and more recently in popular culture such as Harry Potter.
I came across the use of Magic Wands to depict AI recently while using Canva. They have the Magic Studio where they house their suite of AI tools (which are all very fantastic and so affordable, tbh) and use terms such as Magic Write, Magic Switch, Magic Animate and Magic Morph. In this case, AI is magical and whimsical and a mere flick can transform text, images or videos from one thing to another, symbolising the extraordinary capabilities of modern day technology that were once only a fantasy.
The Reality - Servers: A more grounded viz would be a rack of blinking servers in a server room which is the actual physical infrastructure that powers AI. AI systems are usually located in data centres which must provide efficient cooling systems to avoid overheating, appropriate humidity to avoid static electricity build-up or condensation and reliable and uninterrupted power supply with backup generators - and lots of it. Data centres and data transmission networks each account for about 1-2% of global electricity use. To put that in perspective, Australia, 20th globally in terms of energy production, produces around 2.4% of all electricity in the world.
The Infinite Complexity - Fractals: These are intricate and endlessly repeating patterns which are often found in nature in the pattern of plants, tree branches, leaves, lightning bolts, river deltas, tree branches, coastlines, snowflakes, Romanesco broccoli
and axons. Axons are the long, slender projections of our nerve cells (neurons) that typically conduct electrical impulses. The complex fractal branching of these axons is what enables neurons to communicate with various other cells. In AI, this fractal branching represents neural networks and the often recursive or repetitive nature of machine learning at work. As such, there are depictions of AI as trees or neurons (neural networks with nodes).
There’s also the use of microchip imagery - which could fall under the categories of ‘the brain’ or ‘fractals’ or ‘servers’. I felt that the depiction of microchips overlapped within these different categories, but did not feel that it needed its own category.The Imperfection - Disfigured Hands: This one’s a bit niche and I’ve only seen it in relevance to articles which are specifically talking about AI’s inability or struggle to accurately generate images of human hands due to current limitations of the technology and the complexity of human hands, though that has significantly improved. As an artist/illustrator, I fully understand just how difficult it is to draw hands. I understand the basic building blocks, I’ve drawn 100s of hands, I HAVE HANDS, I look at them everyday! And still, I just can’t get it right most of the time.
That said, I find it to be an apt metaphor of “artificial” intelligence and it would be interesting/funny if it stuck around as a visualisation of the early days of AI.The Cloud: Given “cloud computing”, i.e. the use of third-party computing resources such as servers, storage, databases, networking or software over the internet, without hardware infrastructure or investment by an individual or entity, the image of a cloud has increasingly become synonymous with AI and the thinking around decentralised, on-demand computing resources. I find it feels most removed from reality. First, there is the invisibility of the infrastructure - which is HIGHLY resource intensive and could negatively impact our climate as the needs grow. Secondly, it lacks form and substance, it’s just a wispy cloud - ethereal and abstract. It adds to the mystery and mystique of AI. I don’t love it.
The Cute and Helpful - Doraemon: Here I’m using the example of Doraemon from a popular manga series dating back to 1969. Doraemon is a robot (earless?) cat that is sent back to the present day from the future, the 22nd century, to help a young boy named Nobita who is “lazy, unlucky, weak, gets bad grades and is bad at sports”, and improve his quality of life. Doraemon is not explicitly represented as an AI pet but does depict AI characteristics such as learning, problem-solving and emotions, empathy, and moral reasoning - as an advanced AI might. In juxtaposition to the Terminator depicted in Western Media, Doraemon provides a friendly alternative and might help to harbour a more optimistic view of futuristic technology.
In this page below from ±44 years ago, Doraemon creators predicted the AI functionality that is most notable from this year. Read it from RIGHT TO LEFT.
The Possibilities: This one is more complicated. It’s the visualisation of what AI could create in the future, heavily rooted in sci-fi narratives of digital landscapes, virtual worlds, transparent dashboards but also futuristic future societies à la “The Line”. It’s meant to showcase the creative and constructive power of AI to elevate humanity to some beautiful future which is often very dystopian looking - overly reliant on tech, probably oversurveilled, but hey, think of the convenience! I wouldn’t mind this narrative if there was more humanity and joy in the images. We can dream of BETTER worlds.
While researching for this blog, I stumbled upon a resource called Better Images of AI run by BBC R&D, We and AI and Leverhulme CFI, which asks the question, “Have you noticed that news stories and marketing material about Artificial Intelligence are typically illustrated with clichéd and misleading images?” They offer similar examples to the ones I’ve listed above. The website states that such depictions deflect accountability of the developers of AI, misrepresent its capabilities, obscure the real societal and environmental impacts as well as hide the lowly paid human labour behind these processes.
I feel that there is room to more creatively depict AI. I imagine at some point we’ll need a simple icon to identify AI applications. Nowadays, even though young people have never seen the obsolete floppy disk, they know it’s the universal symbol for the save function. This is an example of a skeuomorph whereby a design feature (to save) mimics the form of a similar object (floppy disk) in order to make new tech (advanced storage tech such as graphical user interfaces (GUI)) feel familiar and intuitive to older, better understood tech. Similarly, the magnifying glass is used for the search function or the digital camera shutter sound mimicking mechanical shutters. I’d be curious to see what designers settle on for AI!
Have I missed any? What would you add to the collection?
Note: I would like to thank the brilliant Favour Borokoni for reviewing this blog post and for always being a fascinating and resourceful sounding board.