Tudor emailed the employee communications team in March to share a selection of his AI-generated images. “I saw an intranet article about the Pictet Innovation Challenge and then attended a talk organised by Pictet Academy on ChatGPT. I thought by sharing my passion and experience I might inspire colleagues and help them to better understand the topic of generative AI.” Tudor’s timing couldn’t be better. In the team we’d been exploring how we might use ChatGPT and AI-powered image generators to become better communicators.
Tudor started tinkering with AI-generated images in April 2022, as new programs were released and started going viral. He initially uploaded sketches he had drawn as a teenager on an AI-powered image generator called Stable Diffusion and then modified them by tweaking settings in the program and adjusting his text prompts. “I loved drawing and painting as a kid, especially imaginary creatures. I also got into 3D designs and almost studied architecture, but finally opted for math and finance instead.” But art has continued to be a part of his life and a way to complement his job as a business insights manager.
From images based on sketches, next he turned to his own image, uploading photos of himself and reimagining himself in different forms as a private detective, Mr. Bean or even Captain America. Around this time, Tudor began posting his art to his Instagram account and in tandem with the technology, his exploration took off.
Now he creates visual stories featuring distant alien worlds, abstract Baroque paintings and cyberpunk Vikings, to name a few examples. Since the launch of ChatGPT, he refines his prompts on the program, before pasting them into the AI-imaging software to create new images. “Then I let the program do its thing while I make a coffee. Five minutes later I have up to 50 images to choose from. But it takes a lot of trial and error.”
“My English vocabulary has expanded as well. Precise words and descriptions are essential to generate the visual output you’re looking for as programs can recreate nearly-infinite variations of an image,” ‘Prompt engineering’, the fastidious act of writing a detailed prompt is becoming a highly sought after skillset among tech recruiters. And with the recent boom of attention on the AI-generative space, that interest has extended to ‘civilians’ beyond the tech industry…
As Tudor discovered when a web-based furniture company reached out to him for help providing interior design recommendations to their clients. They asked him to develop 30 prompts to transform their clients’ living room into futuristic, vintage or artsy lounges. “The difference between generating imaginary characters and hyper realistic lifestyle photography depends on the words you use. I had to learn to differentiate different design styles, from art deco to art nouveau.” Only after learning the nuances of design could Tudor get the program to generate the images the company was looking for.
Though the potential feels limitless, Tudor is eager to raise awareness of the risks associated with these powerful tools. “I’ve experienced doomerism, a pessimistic outlook on the development and impact of AI on the future of jobs, society and the economy, more than once this year. As programs improve, it will become increasingly hard to know what’s real and what’s not.” Realistic images of Pope Francis in a Balenciaga puffer jacket or Donald Trump in handcuffs recently went viral. “Seeing well-known figures in unexpected, comical situations is funny, but also very scary if you imagine the potential for impersonation and misinformation.”
In a little more than a year, Tudor has produced over 70,000 images and started posting a daily selection on his @house.of.draco Instagram account entirely dedicated to his art. I ask him how he sees his future. “I love my job here at Pictet and only see these tools as add-ons that can help me express my creativity.”Quand il n’est pas en train de créer des images grâce à une IA, Tudor est Business Insights Manager chez Pictet Asset Management.
At the time of writing, as journalists, bloggers, and financial analysts debate daily about the threats and opportunities of AI it’s tempting (and natural) to feel overwhelmed. But looking back at my exchange with Tudor, I remember his passion beyond the words recounted in this article. I see his smile and kind, twinkling eyes. I remember him using his hands animatedly as he talks. I hear the lilting tone of his voice. And I look forward to grabbing another coffee with him, human to human.
When he’s not creating images with AI, Tudor is a Business Insights Manager at Pictet Asset Management.
ChatGPT is a natural language processing tool driven by AI technology that allows users to have human-like conversations with a chatbot. It can answer questions and assist with tasks like composing emails, essays, and code. But beware of occasional factual errors and biases. The technology has regularly made headlines since its public release on 30 November 2022 and is considered the fastest-growing app of all time according to UBS, with over 100 million users within three months of launching.
About AI-generated images
Regarding visuals, programs like Midjourney, Dall-E and Stable Diffusion generate images based on text prompts or other images. Similarly to ChatGPT, these tools were trained with massive amounts of data (in this case images) from the internet, fuelling copyright debates among artists, creators and photographers. A text prompt is the text description or set of instructions to describe the image a user wishes to create.