One of the changes for 3blue1brown this year has been to take on more of a team around what has traditionally been more of a solo project. I asked some of the new members if they wanted to offer any reactions or writings in conversation with all the AI in math discussion. Below is a short essay in response from Mitchell Zemil, an animator/illustrator who I’m delighted to have on board.
First: a quick disclaimer that I am not a mathematician by any means; I am an artist. Namely, I am professionally a freelance 2D animator; I am an adjunct professor teaching animation techniques to young artists; I dabble in pottery and watercolors, and I play a little music on the side. And though art and mathematics overlap in certain curious ways, they can be seen as relatively disparate fields (compared to math and physics, for example). Still, as both are being affected by the latest advancements in AI, I wanted to share my thoughts on how this technology’s rapid advancement and adoption feels from an artist’s perspective, and how that may overlap with the math world.
For what it’s worth, I’ll also give the obligatory disclaimer that “artificial intelligence” here refers to the hyper-contemporary definition: specifically the LLMs, the GPTs, and so on.
In many ways, the art world is set up almost precisely to be antagonistic to generative AI. It is a very broad set of disciplines which celebrate individuals who carve out niche skillsets and interests within the field, or find ways to connect distant disciplines in a novel manner. Though creativity of some kind is often considered a core skill of artists, and perhaps that is true, it is the artist’s labor and effort which is regularly celebrated and venerated (admittedly to a sometimes excessive “celebration of toiling”). Not only is the artist respected for dedicating impressive swaths of time and energy towards their oeuvre, but the years they spend cultivating their talents are implied to be part of the reverence they earn, and make the skills of a virtuoso appear mythically unattainable to the casual observer. Artists also have a very specific relationship to “sources of inspiration,” and there is a very fine line between creating art that builds off past works and is in conversation with them, and art that is considered plagiarism or highly derivative without intentionality or thoughtfulness.
So then, when a new tool comes along that promises anyone the ability to produce work with a high level of mastery, to synthesize any subject matter, aesthetic qualities, and sources of inspiration, with inhuman speed (and with minimal human toil), it is natural to expect some pushback. This is compounded by various factors beyond the essential abilities of the technology; many artists already experience their work becoming devalued and face professional precarity, and it is clear that some others see one promise of AI-generated artwork being “never having to pay an artist for anything ever again.” This also ignores the ethical can-of-worms of how the training data for these AI models originally came from, among other sources, human artists who may now be finding their careers threatened by the technology. This also ignores the damage to the environment, infrastructural issues, and arguments of subjective taste (whether or not any AI-generated art has the potential to be aesthetically compelling or meritorious). Still, in a world where AI could ethically produce aesthetically impressive or interesting works, without economically displacing anyone’s livelihoods, would we invite it to supplant all of our musicians, writers, painters, and actors? Or would we demand these artists ‘keep up with the times’ and incorporate AI tools into their processes? My gut-reaction is “no,” though it’s not easy to articulate exactly why.
It’s worth appreciating, in the meantime, that a big part of the story of art has revolved around access: who gets to be an artist, what tools and training were requisite to becoming one, and even who gets to own or experience art are longstanding questions with ever-evolving answers. Beyond shifting cultural norms, technology has played a big role in democratizing access to the art world. These advances can introduce new crises to the art world, which reacts accordingly. When cameras became an affordable alternative to paying a skilled painter to capture a scene with lifelike accuracy, new styles and traditions sprang forth that rejected realism in favor of what a camera couldn’t do (and photographers, in their own right, have found many ways to produce work that goes beyond a straightforward realism). But even as an ever-evolving field shakes up the status quo, what has remained consistent is the dedication an artist makes towards learning and understanding their craft, the intentionality and care they put into their practice, and the recognition they receive for the contributions they make to their discipline(s) and the journey they took to get there.
So what makes generative AI a different kind of tool than digital painting software, a cell phone camera, or a 3D printer? To the AI ‘optimist,’ it is simply another step in advancing what kinds of sounds, images, words, or other outputs can be achieved in a relatively short amount of time by a smaller team, or a less practiced individual. But to the skeptic, it turns artmaking into a practice of tastemaking: asking for a result and choosing the output which you enjoy most, perhaps fiddling with a few settings or details of the prompt until landing on a satisfying answer. They may argue that the artist’s “DNA” is a recognizably important part of the final result: the work is informed by their background, the certain relationship they’ve developed with their tools (and their own bodies, such as in acting or singing), the relationship they have with contemporary artists and past artwork they’ve encountered, and the techniques and unique solutions they’ve discovered to synthesize these skills and concepts (or at least their attempts to do so). Likewise, when one imagines an “artist at work,” whether or not that behind-the-scenes work is being documented, one can picture the artist engaging with the process in a highly intentional and interactive manner, seeing the gears turn in their head, and observing how their unique combination of instinct and intent guides their hand from one brush stroke to the next. And we tend to admire that quality about artists and their work.
Jumping over to mathematics (as an outsider), it seems to be a field whose outputs are celebrated less as examples of subjective expression, and more as advancements in knowledge or understanding. The revered mathematician’s work may be informed by their individual background, their own familiarity with old theorems and new discoveries, and the different areas of mathematics that might matter to them personally, but this “subjective DNA” is not what gives their contributions value as much as the ‘natural truth’ they imbue; mathematical knowledge remains objectively true (and functional) whether it is observed by humans or sea sponges or computer programs. This perceived “purity” of math sometimes gives the false impression that it is a discipline that can be distilled into practical tasks and to-dos, that simply “doing enough math” will answer all of the questions we could possibly ask of it, and the main task of the field is to crunch enough numbers until the answers reveal themselves. Therefore, would we not be happy to have a hypothetical machine that could answer any mathematical problem you handed to it?
Again, this is an opinion that may be cynically held by those who see math as a means to a specific pragmatic end. In fact, one of the complicating matters, and why controversy is so often a byproduct of AI’s achievements, is that there are several parties with massive vested interest in AI’s mass adoption, and there are open questions about whose autonomy is expressed in the adoption of AI tools by various industries. For both evangelists of this technology, and also its most pessimistic opponents, AI is seen as a tool to replace the work done by humans, rather than one to empower those humans.
But much like in art, people seem to like that mathematicians exist. We enjoy that, even in the age of calculators and supercomputers, we have living stewards of this knowledge, that someone can point at the machine’s outputs and explain what it means and, with a genuine passion, explain what it means to them. For those who care about math, it is a deeply human pursuit and is thereby informed by the human curiosity and ever-evolving understanding of its disciples. Even if advances in AI represent a trivializing of certain mathematical tasks, much in the same way the computer trivialized arithmetic, a demand will exist for those who follow in the field’s traditions, gain footing at its frontiers, challenge the boundaries of human understanding, and share their gifts with contemporaries and pass them down to future generations. Just as long as humanity appreciates the value in attending the ballet, there will be dancers.
-Mitchell Zemil
Adjunct Professor, Montclair State University