Creativity and Imitation

As part of our post-apocalyptic recovery, my wife Nancy and I drove up to Los Angeles to visit the Getty.  Some people find refuge in music; for us it’s art.  We were especially interested in an exhibit featuring a chromatic analysis of Van Gogh’s Irises, the most important work in their permanent collection.  Other great painters are there—Manet, Courbet, Pissarro, Turner, and Monet, for example—but generally with their second level work.

Not Irises.  This was Van Gogh at the peak of his craft.  But the museum’s conservators had realized for some time that colors may have faded over the years, and decided to try to discover how it appeared when Van Gogh first completed it in 1889.

For the exhibit, Irises was set in room of its own.  Diagonally across the room was a short film on how the analysis was conducted.  A technician scanned the painting with a stereo-microscope, using non-invasive techniques including X-ray fluorescence and reflectance imaging spectroscopy—whatever they are—to discover the trace elements that would reveal the pigments Van Gogh employed. 

For example, although the flower petals now appear blue, Van Gogh wrote in a letter to his brother Theo that they were to be violet.  The analysis revealed that the red he mixed with the blue to create violet had almost completely faded, leaving the flowers to appear blue to the modern viewer.

When the conservators had completed their task, they created “a digital reconstruction of the painting to show how Irises might have originally looked,” in which “the contrast between the violet petals and their golden stamens rendered all the bolder.”  The digital reconstruction was on the wall opposite the genuine painting, about thirty feet away.

When Nancy and I entered the room, we first walked across to see the original and, as whenever we see anything produced by this monumental talent, spent a good bit of time gazing at it with awe.  We then walked across to see the film, which explained how detection of the trace elements of chromium, bromine, cobalt, and copper revealed Van Gogh’s intention and how ultraviolet light was used to enhance the underlying imagery.  (We did not actually understand precisely how this was done, but we took their word for it.)

Then we looked at the reconstruction, expecting that now, with Van Gogh’s bold colors restored, we would be even more awestruck.

We were not.

The colors in the reconstruction were indeed accurate and bold, and the contrast that Van Gogh was looking for stood out, but still…

The original, faded reds and all, was more powerful, more alive.  Much more.  Thinking it was me, I walked back across the room and the distinction was unmistakable.  When I returned to the reconstruction, Nancy said, “The original is better.”  Then she added, “You can’t recreate genius.”

Indeed you cannot.  Which brings us to AI…

That AI will aid or supplant many functions now performed by humans is not in doubt.  It has already established incredible utility in the hard sciences and medicine and the threat of AI replacing, say, screenwriters, was at the center of their recent strike against the studios.  But for all its virtues, artificial intelligence is not actual intelligence, any more than an artificial Van Gogh is an actual Van Gogh.

AI can access and compile existing information with staggering speed and then produce an almost infinite number of permutations or, if the input parameters are sufficiently detailed and specific, can produce limited outputs to solve a specific problem.  The application to medicine and hard science is obvious.  AI can create prototype drugs, for instance, much faster than human researchers, or provide interpretations of data observed by subatomic- and astro-physicists.  In the film business, AI can write a buddy movie every bit as commercial as can a screenwriter, or, in publishing, produce formulaic mysteries or romance novels faster than even rapid-fire authors can turn them out.

What they cannot do, as with the Irises example, is create.  No AI program would produce a Citizen Kane, although it would be more than capable of spitting out a film like Citizen Kane once the first version existed.  There are intangibles in the creative process that AI cannot replicate, nor was it meant to.

That distinction is likely to be lost on many educators, and, as a result, their students as well.  Many teachers will almost surely employ AI as a substitute for engaging students by seeing what works and what doesn’t and altering their interactions with them accordingly; students will employ AI, whether allowed to in theory or not, as a substitute for finding their own paths to learning.  Thus, papers or lesson plans can easily degenerate into an imitation of actual learning, just as the Irises recreation was an imitation of actual art.

In other words, both groups will use AI for what is at its core.

A shortcut.

This is not to say, of course, that both teachers and students do not employ shortcuts now. They have done so since before the internet or computers existed.  But AI is perhaps the most sophisticated shortcut that has appeared in decades, or perhaps ever.  And this is a shortcut that works against both critical thinking and creativity, both of which are in increasingly short supply in a nation that thinks reality television is the same as reality.

There is, therefore, the opportunity to use AI to indoctrinate students to a particular point of view, creating persuasive material to, for example, demonstrate the United States Constitution was inspired by the Bible or that gay classmates are perverse rather than just different.  Both fit neatly into the stated Trump agenda.  Without critical thinking skills, these students will be a good deal easier to manipulate.

AI can be a powerful tool to improve the lives and fortunes of tens of millions of people across the globe.  But, like all tools, it has specific uses and limitations and if used for purposes other than those intended, it can tear down that which it was meant to build up.

We have, alas, become a nation of shortcut seekers, but we need real Van Goghs, not imitations, and there are no shortcuts to getting them.