From a very early age you were likely categorized as being good in Math/Science or Language/Art. We're taught that these subjects are from different worlds, requiring a different brain to live in one or the other. The truth is, Art and Science are two sides of the same coin.
Both pursuits seek to understand our world and then communicate our perception. They both look to engage an audience and influence their thoughts. It's not enough to continually tell the same old stories, Science and Art are always exploring what is new. If an artist or scientist has nothing new to say or have trouble communicating their ideas, they will fail.
As technology becomes more and more pervasive, the lines between Science and Art are blurring. For centuries artists had to study for many years at the best studios so that their images would be as realistic as possible – just like being there. Can you imagine what it was like when the camera was invented? At the push of a button, anyone could make an image more realistic than any artist. Did artists become obsolete? Absolutely not. Photography became an art form in itself, while it caused other artists to change their perspective.
It was no coincidence that the first photograph was taken in Paris in 1826 and then the Impressionist Movement in art started in Paris in the 1850s. The first photos were blurry, but by 1850 they were looking pretty sharp. Some artists might have decided if they were going to compete, they needed to paint things that could not be photographed. They started painting their impressions of things, instead of highly realistic images. Impressionism led to several modern art movements, so indirectly, the technology of the camera profoundly affected all art.
Today, scientists need more compelling scientific illustrations to get their points across, while artists are using fractals and 3D models more often than paint or clay. The computer is integral to bringing science and art together and is proving to be just as transformational as the camera.
I use Artificial Intelligence to create new images. More precisely, I use DeepDream Generator's
(deepdreamgenerator.com) DeepStyle, which belongs to a class of algorithms called Neural Style Transfer (NST). This optimization technique is used to take two images -a composition image and a style image- and blend them together. The output image has the overall look of the composition image, but is “painted” in the style of the style image.
Google made its DeepDream algorithm popular back in 2015, then made this software open source and various implementations have spun off of it. DeepDream is characterized by images with a thousand eyes and various strange creatures appearing in strange places, like a hallucination. These creatures were “trained” into DeepDream to create something like an image “toolbox.” DeepStyle, on the other hand, uses a single image as its “toolbox,” borrowing various puzzle pieces from the style image and then applying them to the composition image. This was first used to apply the style from a famous artist to a photo of yourself or your cat. This kitsch fad lasted a couple of years and then faded.
My approach is sort of the reverse of that. I usually use 100+ year old public domain images as the composition and apply a modern style to them from abstract computer graphics, fractals, glass, minerals, flowing paint, stained glass,... Sometimes multiple images are combined in a collage to provide more variety of effects.
Rather than describing this with too many words, it's more obvious to see the images:
Computer generated mosaic tiles were shaped, sized and fit onto the Ford Trimotor plane.
Any image can be used as a composition or style, but they need to be higher resolution than your planned output. Currently, DeepDreamGenerator tops out at an output of 1MP, which looks OK on a computer screen, but is far too low resolution for printing. A new version of the software will soon support 10MP output. This is borderline printable, where small prints will look fine, but anything poster size or larger will show its limitations. Eventually, 20MP or higher will provide more professional results.
DeepDream and DeepStyle are still evolving, with exciting developments in video, virtual reality, 3D sculpting and more. It's already showing up in some music videos and video games, so this approach is going mainstream.
So how does this relate to the work of scientists? Science isn't all facts and figures and endless trial and error – Science needs Creativity. To do anything new you need a creative “spark” and art can help supply it. AI Art can be very abstract, fueling different trains of thought each time you look at it. It also has its roots in technology, requiring heavy-duty computers for both AI and Graphics, with multiple CPUs/GPUs, terabytes of storage, gigabytes of RAM,... and what science nerd doesn't love a solid dose of technology? It's so resource intensive that it's best done on a hosted server, not a workstation. For those who've had a hand in developing this related technology, there is a direct connection that pulsates through these images. Can you imagine these images evolving before your eyes in videos or real-time interactions in virtual reality? They'll be like creative thought generators, helping build new neural pathways in your brain, through the use of AI neural network pathways.
Technical papers and articles can be a little dry, with precise, realistic images providing limited energy to the reader's mind. Placing a more abstract image on the title page and even interspersed in the text can help draw attention to your work and set a mood or foster new ideas. After all, some scientific concepts are also pretty abstract. These images can also serve as a trigger or attachment point for memory. Art can be a tool, just like words, graphs and diagrams. It is a type of language that helps get your point across and stimulates deeper thought processes. It may even stimulate a deep dream, from which you wake up to a new scientific discovery.
When I start printing higher resolution images, many will likely be printed on glass, polycarbonate and other transparent and opaque surfaces. Since I often use glass and glass-like styles, they resemble stained glass and will look great hung in a window or with colored LEDs mounted behind them and programmed for different sequences.
Besides ties to Science and technology in general, DeepStyle can team up with Advanced Materials. When 3D structures are possible, we will 3D print them or print 3D molds and cast. To make some complex shapes possible or practical, advanced materials will make them stronger, lighter, more flexible or provide some other attribute. For example, cast panels used as the facade for buildings. Molds will be 3D printed and the panels cast, incorporating advanced materials, such as graphene. Panels could also be cast for interior walls and floor/ceiling slabs. All of these panels would be attached to a steel girder frame. By making them thinner/lighter through Advanced Materials, the steel frame won't carry as much weight and can use smaller girders, further reducing weight. These panels could be produced on-site, using local materials to reduce cost and environmental impact.
Since I was a kid, it has bothered me that as modern as we are, we can't build what the Romans did, such as the Pantheon in Rome. After 1,894 years it remains the largest unsupported dome in the world and it has no steel reinforcement. Another disappointment is how beautiful Art Nouveau and Art Deco buildings are, but we can't find or afford the artisans needed to build them today. Instead, we're stuck with metal and glass boxes, who's primary attribute is the amount of floor space they contain. 3D Printing and Advanced Materials will solve these issues and usher in a new era where Science and Art collide to accomplish great things. We can finally surpass the great historical accomplishments and start becoming the envy of future generations.
Steve Stout is a technologist who enjoys art. He's worked on many groundbreaking projects, including:
➢ Air Products: one of the first Programmable Logic Controller systems, managing cryogenic gases in all 28 nuclear power plants;
➢ Bell Labs: Integrated Circuit manufacturability;
➢ IBM: Software development environment planning and introduced HTML to NASA for the Space Station information access;
➢ Sun Microsystems: Market Development for middleware and Java;
➢ WorldCom: defining the cloud.
He's now working on new ways to marry art and technology to create a better world for everyone to live in.
July 6, 2020