top of page

About> the garden

What's going on here?


My new art series relies on neural network programming, a technology covering a range of applications related to machine learning and artificial intelligence.


In the conventional approach to programming, a computer is told what to do by breaking big problems up into many small, precisely defined tasks it can easily perform. By contrast, in a neural network we don’t tell the computer how to solve our problem. Instead, it learns from observational data, figuring out its own solution to the problem at hand.


For my artwork, I process Renaissance painting files with Google’s open-source software, DeepDream, via the Deep Dream Generator site, which allows me to transform my source images in ways that are only partly predictable, and often quite astounding.


Studio photo of In the Garden. 133cm x 90cm.

Deeper dreams


While based on neural network software originally designed to detect faces and other patterns in images, DeepDream runs this same software in reverse: it over-interprets and enhances the patterns it sees in an image. In other words, it scans an image and identifies patterns that have slight correlations with specific image categories (faces, animals, or any other designated imagery) in its ever-expanding visual database. It then – through repeated processing – builds ever more nuanced and amplified visual patterns based on the type of imagery it detects on these nodal points.


This reversal procedure is never perfectly predictable, which makes it particularly compelling for my own artistic explorations. Through DeepDream’s wide array of inputs – each of which affects how a selected starting image is altered – the final image slowly emerges over repeated reprocessing. It has been pointed out that this process is somewhat similar to lucid dreaming, originally an ancient Buddist technique for developing control over the content and direction of one’s dreams.


Just as with lucid dreaming, DeepDream’s software combines a rich mix of masterly control with chaotic unpredictability. For example, even when precisely the same inputs are applied to an image file, the result can vary from one processing session to the next.

In my experience, good art tends to emerge from a mixture of precise control and surrendering some of that control, which is what continues to draw me deeper into this way of working. It is puzzling to me that while Google’s application is quite popular, there has been relatively little interest in the concept from contemporary experimental artists.



The resin-coating process...

...and result.

Why Renaissance paintings?

For several reasons, I chose to produce this continuing series – The Garden – with Renaissance-era paintings as my key, though not exclusive, source imagery. I wanted to ground my explorations in a larger historical arc of art and technology. I also think that this process – and our era generally – share a strong affinity with elements of the Renaissance era, specifically as a time of intense advances in technology, art, and a blurring of the boundaries between the two.


Similarly, the Renaissance gave us artists concerned with timeless beauty and truth – Botticelli comes to mind – but also artists like Bosch, whose love for the surreal and the absurd seems rather perfectly in step with the DeepDream sensibility. The pictural lushness and compositional rigor of classical painting, even in its overblown and neo- forms, delights me.  But The Garden is clearly not intended as homage. The relationship of these earlier works of art to my own is as armatures or agar medium for creating my own quite distinct, radically divergent visions. 

In recent years we've learned that many Renaissance masters used technological tricks, including optically projecting their subject matter directly onto their canvases for tracing. While this Renaissance 'trickery' may deeply offend modern-day purists, it also silences critics who oddly still reject visual media such as photography and video, or my own explorations here – as being somehow less for acknowledging technology.


My early, often clumsy, fascination with Renaissance art also made it a natural choice for this series: its visual complexity, structural brilliance, pop-star power, and reliance on obscure and often concealed symbolism fits me just right.



My way of working


I choose my source images with care, and typically generate dozens of variations from a single picture. I repeatedly reprocess a selected image as it slowly transforms, pursuing and refining the most compelling visual ideas as they emerge. In addition to my primary source image, I also add one or more secondary “seed images” – such as pictures of apes, guns, flowers, or automobiles.

After selecting my final work from a folder of anywhere from 10 to 100 variations of the original image. 
The final artwork is then carefully color-balanced and archivally printed to a print size of more than one-meter square, by Artproof.  I then mount it on a stiff, honeycombed aluminum composite board and carefully resin-coat the image by hand.


Mark Maher
Helsinki, September 2020

bottom of page