What does AI look like?

In the context of AI development, public administration plays a vital role in communicating about AI to citizens, and other civil servants. It is essential to clarify what artificial intelligence is and is not, how to use it or when it becomes ineffective, the speed of its development, and the phenomena that accompany this growth. And that is neither artificial nor intelligent. Additionally, regulating AI in sectors like business, education, and the judiciary is crucial. Effective communication requires not only text but also images that capture attention, evoke emotions, encourage engagement, and can also instill fear, promise, or hope.

Daily, on social media, and in mass media, I predominantly encounter images of humanoid robots against blue backgrounds, metallic brains with intricate networks, and stars. This portrayal exaggerates AI’s nature, presenting it as more than it truly is—a technology incapable of ethical decisions, operating on probability principles, and algorithmizing sentences, events, behaviors, and judgments. Often, AI is personified as a beautiful, shapely woman to make it more relatable and empathetic. However, have you ever seen a non-slim robot? Or robots in wheelchairs? Conversely, some images depict AI as dangerous terminators aiming to overthrow humanity. Other representations include white-cuffed robots solving business problems or abstract, digital, and dark forms that suggest complexity and inaccessibility, discouraging public exploration, explainability, and understanding. These portrayals can lead to discrimination and perpetuate stereotypes. Public administration should uphold exactly the opposite things and keep much higher standards in AI representation!

Humanizing AI can make us forget the people and intentions behind the technology. Abstract and mysterious visuals can make AI seem too complex, while the reality involves large corporations, significant financial investments, data, politics, power, and essential resources like water and energy that states and businesses compete over. AI also aims to streamline processes, enhance communication, increase service accessibility, and drive breakthroughs in health, education, and climate sectors—benefits we all need! Unfortunately, overly humanoid or abstract representations hinder our ability to integrate AI into our lives and articulate its presence accurately. They may also discourage us, both as humans and citizens, from confidently expressing our expectations of AI. But understanding these needs could inspire researchers to develop groundbreaking solutions!

Amidst the disillusionment caused by inappropriate AI visualizations, I discovered betterimagesofai.org. This platform, created by a network of public, private, and non-governmental organizations and authors, is collaboratively building a library of AI images that avoid superficial portrayals. The library is complemented by a guide on effectively presenting AI in project communications and promoting responsible and ethical discourse. The guide is based on interviews with various individuals, capturing their associations and feelings to make AI more relatable. Additionally, the site encourages others to contribute their photos and images under a CC-BY 4.0 license.

To encourage thoughtful use of images and benefit from the guide's tips, I invite you—following the authors Dr. Kanta Dihal and Tania Duarte—to ask yourself the following questions whenever you add an AI image or graphic to a report, presentation, poster, or post:

• Be honest: does the image accurately represent what the AI system can do? Could it mislead viewers?
• Keep people in mind: Does the photo convey that AI is created by and for humans? Does it avoid suggesting AI is developed in isolation from real-world human problems, such as in inaccessible offices based somewhere in space? Does the image avoid perpetuating stereotypes related to gender or race?
• Choose relevance: Is an AI-specific image necessary for the context? For instance, could an article about a new chatbot use a photo of an app interface instead? Could a climate change modeling app be represented with a realistic photo of a swamp or glacier?
• Be specific: What specific AI technology are you showcasing? Is it a language model that generates text, or an implementation that automates administrative decision-making?

Sometimes, simply seek others' impressions, feelings, and thoughts before including an AI image in a report or on a slide. Remember, it's not just a picture—it's a representation of the responsible work we do in administration and a visual signpost that provides people with a sense of security.