Capability Brown Meets AI

Daniel Ambrosi is recognised as one of the founding creators of the emerging Artificial Intelligence (AI) art movement. He is particularly noted for the nuanced balance he achieves in human-AI hybrid art.

Based near Silicon Valley in Half Moon Bay, California, Ambrosi has been exploring novel methods of visual presentation for over 40 years. His skills were initially honed at Cornell University when he was enrolled on the Program of Computer Graphics, resulting in a Bachelor of Architecture degree and Masters in 3D Graphics. This would prove to be a compelling combination as his work flourished and developed.

For Ambrosi, computational photography is more than just creating a sense of place. It’s about conveying how we feel about the environment – viscerally, cognitively, and emotionally.
Here’s how Ambrosi experienced the landscape at Bowood on his recent visit.

Bowood, 2023 by Daniel Ambrosi
Ambient-lit Dye-sub Fabric Print
121.9 × 121.9 cm (48 × 48 in.)

In the early morning of 15 April 2023, I set out from my accommodation near Oxford to continue my month-long quest to capture photographic source material of English landscape gardens designed or inspired by Capability Brown. The weather forecast predicted a partly cloudy, rain-free day, which is the ideal mix for generating the kind of luminous landscape experiences that inspire me.

My first target of the day was Prior Park in Somerset, which was to be followed by a visit to Bowood House and Gardens. I chose to conduct my expedition in this order after studying maps and aerial imagery; I wanted to be at these sites at times of day that would present the most optimal sun angles for my intended vistas.

When I arrived at Bowood, despite very much wanting to explore the interior of the main house, my time inside was limited to getting an answer from one of the staff members as to how I could get to the other side of the lake. My online research indicated that this was where I would find the most feature-filled views of the estate; a classic combination that included foreground trees and grasses, water, rolling hills, and the house itself.

While I consider myself a landscape artist whose work is deeply informed by the 400-year history of landscape painting in the Western world, my paints and brushes happen to be cameras, computers, and artificial intelligence. Starting with my digital camera, I capture dozens of individual overlapping views, which I stitch and blend together into a seamless high dynamic range panorama using a variety of graphics software packages.

This particular artwork consists of 24 photographs (4 shots wide × 2 shots high × 3 exposures ‘deep’), although the original capture from which this composition was cropped was actually 36 shots. Using a Sony RX1 Rii full frame 42MP compact camera with a fixed 35mm lens mounted on a panoramic tripod head, it takes me about a minute to capture the entire scene.

Back in my studio, I then pass my completed photographic panorama to my collaborative ‘partner’; a custom image recognition AI based on Google’s ‘DeepDream’ open-source software that attempts to interpret my scene. I can set the direction of that interpretation, but I can’t control the details, which is always a wonderful source of surprise. In this artwork, I employed two passes of ‘dreaming’; in two different styles at two different scales, which is how the intricate patterns (or ‘hallucinations’) that you see are generated.

In a sense, I’m using my AI as an intelligent paint brush that I push to make the kinds of marks that I feel are most compatible with the underlying source imagery. What makes my AI especially sophisticated is that, since it was originally tasked to do image recognition, it refuses to lose the details in the scene that it’s interpreting; this is why the fine lines in the foreground grasses and tree branches remain crisp despite the relatively large scale of the hallucinations in the final artwork.

My motivation behind this artwork and the collection to which it belongs, is to share with others as fully as I possibly can the experiences that I’ve had in these special places, not just visually, but also viscerally and cognitively. 

Admittedly, the act of translating that four-dimensional experience into a two-dimensional art object requires bounds, a point of view, and the use of visual metaphors. But my hope is that the experience of these translated landscapes yields the same sense of joy, wonder and rebirth that I felt when I visited historic English heritage sites like Bowood for the first time in the spring of 2023.

The resulting work is a mesmerising landscape which appears to be an accurate photograph but is in fact filled with vibrant details that reveal themselves on closer inspection.

Ambrosi’s picture will be on display in the Laboratory where 250 years ago this year, Joseph Priestly discovered Oxygen.