Seeing beyond the clichéd images of AI

This screenshot shows The results of a Google Image Search for AI are clichéd and fail to accurately illustrate its applications. (Image credit: Google Image Search)

The results of a Google Image Search for AI are hackneyed and fail to accurately illustrate its applications. (Image credit: Google Image Search/screenshot)

Artificial intelligence is the most transformative technology of our time, yet much of the imagery used to illustrate AI is so clichéd, it’s laughable. Put the term “artificial intelligence” or “AI” into Google Image Search and you’ll get a sea of glowing blue pictures that vaguely resembles something out of The Minority Report (2002) or The Matrix (1999), back when science fiction writers and designers were busy imagining the immersive digital environments of the future. Now, in 2018, artificial intelligence has arrived: Machine learning, neural nets, and other advances are transforming medicine, transportation, manufacturing, and virtually every area of human endeavor. Why, then, is the stock AI imagery so lame? And why does it matter?

Those are some of the issues Molly Wright Steenson explores in her essay “AI Needs New Clichés,” which is also the subject of her upcoming talk at the Chicago Humanities Festival. The outdated imagery that pops up at the top of a Google search is “ridiculous,” said Steenson, whose work as a writer and professor in Carnegie Mellon’s School of Design deals with the intersection and implications of design, architecture, and artificial intelligence. The images “are all in shades of blue, green and black and semi-translucent. They all have circuits and weird brains with little network designs behind them, with a Minority Report feeling or like computer generated graphics from the ‘80s,” she told All Turtles in an interview. “They don’t explain or clarify the technology in a useful way at all.”

This is a photo of Molly Wright Steenson is the author of Architectural Intelligence: How Designers and Architects Created the Digital Landscape (MIT Press, 2017). (Image credit: Carnegie Mellon Design)

Molly Wright Steenson is the author of Architectural Intelligence: How Designers and Architects Created the Digital Landscape (MIT Press, 2017). (Image credit: Carnegie Mellon Design)

Of course, artificial intelligence is famously challenging to represent and explain, especially given the opacity of deep learning algorithms. Citing researcher Jenna Burrell, Steenson writes that there are three reasons for the opacity: “the need to protect algorithms that are state or corporate secrets; the fact that A.I.-related coding is still the territory of specialists; and a ‘mismatch’ between the mathematical ways algorithms process information and the way humans think.“

An entire field of “explainable AI” has cropped up to deal with this “black box” problem and the need to be able to understand and reverse engineer algorithmic outcomes so that society can trust the decisions and actions taken by these systems, whether an oncology diagnosis or a home loan acceptance or rejection.

The visual clichés for AI, however, are not advancing the cause.  They are inert, deadening, and seem to hint, unhelpfully, at otherworldly, magical powers. “They show that AI is not of our world at all, and that we don’t have any agency over it and won’t be able to really grasp it,” Steenson said. “In fact, in these images, AI disintegrates and dematerializes.”

Steenson acknowledges creating imagery that helps with “algorithmic literacy” is really, really hard. But given the real world impacts of AI, we’ve got to do better, “because what we don’t know can actually hurt us,” Steenson said.

The movie Her, about a man named Theodore Twombly (Joaquin Phoenix) who is dating his intelligent operating system, introduces a different palette of AI cliches. As he falls harder and harder for his virtual OS lover, Samantha, he also avoids the messiness of embodied human relationship. Instead of the usual translucent greens and blues, the look of Her is rosy and everyone dresses in red or yellow. “It’s color-corrected like gauzy Instagram photos of Coachella fans,” Steenson writes, “an interface of images in which nothing is not a computer.”

This picture shows Theatrical release poster for the movie "Her" (2013). (Image credit: Warner Bros.)

Steenson hopes that designers and people of all ilk—those who aren’t necessarily tech companies and “programmer gods”—will throw themselves into the fray. Otherwise, we’ll remain stuck in a time like that of  “30 or 50 years ago when computing was beginning to boom and pursuing something that shut out a lot of people.”

One central fallacy embedded in today’s AI clichés is that AI is somehow disconnected from our world. “But AI is actually embodied, it’s physical, it’s connected,” Steenson said. “It’s within things and on things and around things and we’re a part of those things. It’s vital to make that materiality and connectedness clear.”

Recently on All Turtles we wrote about the Anatomy of an AI System, a graphic by researchers Kate Crawford and Vladan Joler that shows the vast interconnected web of natural resources, labor, and data that powers the Amazon Echo. Unlike the corporate AI visual clichés, Crawford and Joler’s graphic uses black and white, which signals that, like other industries, AI also runs on elements mined from the earth. Hopefully, this graphic and others by more humanistic data artists, such as Giorgia Lupi and Stefanie Posavec (“Dear Data”), will start bubbling up to the top of Google Image searches.  

Part of the issue may be that artificial intelligence is itself a cliché, Steenson said. “Nobody studies artificial intelligence. They work on machine learning or neural nets or other sub-questions. When we say AI it’s really just a cliché for computation and interactivity.”