Responses by Pj Jones, founder, Mallard & Claret.

Background: Abstract Intelligence is an exploration into the use of AI within the world of art and culture—in particular, the ethical impact of using AI to produce artwork using prompts that specifically reference another artist’s name or style or that have been trained using copyrighted material. To demonstrate this point, we use examples that compare the results of instructing an AI to produce a piece of art “in the style of Mona Lisa” versus simply describing the artwork without referencing the artist or the original painting.

Design core: The site opens with a 3-D sculpture of Michelangelo’s David covered in graffiti, which, on scroll, breaks apart to commence the story. The site is extremely text heavy, so our goal was to make the experience of scrolling and interacting with content as engaging as possible. This included an interesting mouse AI/paintbrush interaction and adding a subtle noise to all of the images.

Challenges: Performance optimization. Despite the creativity in asset-heavy websites, we were determined to prove that they don't actually have to be heavy. In the end, we delivered a complex, asset-heavy site that loaded in just milliseconds.

Achieving page speed scores in the 90s took a while to perfect, and once we got there, Google then introduced some new parameters in its scoring algorithm that penalized Abstract Intelligence further. Despite this, our Lighthouse scores are 100/92/100/100—the 92 being a color-contrast ratio penalty.

Time constraints: Mallard & Claret is a creative website design and development agency, and Abstract Intelligence was an internal, unpaid project that we created to bring our creative and dev teams closer together, experimenting with new technologies and design trends. As such, this project had to compete for our team’s time alongside commissioned projects for clients, which ultimately took priority. This meant the project took longer to deliver than we would have liked, as we had to keep putting it on hold and dipping in and out of it.

Alternate paths: AI has evolved so quickly over the last six months. Between the time Abstract Intelligence’s design and launch, many of the topics we discussed had already evolved to a point where the copy and citations probably needed updating to make the site more relevant. However, as this was an internal project, we didn’t have the time to do this; in hindsight, we should have made the site slightly smaller so that we could have delivered it more quickly, and it would have stayed relevant for longer.

Technology: The core is built with HTML, SCSS, TS and Webpack. We also used the libraries GSAP, Lenis, Lottie and three.js.