On Not Keeping Up-To-Date

Being new to Substack, and as I get back into the groove of writing again, I’ve been thinking a lot about the different topics I want to post about. But there are also the logistical issues of keeping a blog alive — things like: How often should I post? Should I write one post at a time, or work on several simultaneously? And, how topical do I want to be?

The topical question was the trigger for this post. I had just read “How to Stop AI FOMO” and it resonated with me. I thought it could be interesting to write about how, as an artist, I try to avoid getting caught up in the endless hype of the new. It’s a theme that underlays much of my work — in part because I’ve been working with technology for a long time and I’ve seen plenty of fads, hypes, and boom and bust cycles. I thought that a post about the joys of not being up-to-date would be a nice way to talk about my creative philosophy.

But then I did something awful. Probably because I haven’t yet settled into a writing routine. I asked ChatGPT to write the post for me. And that made me want to quit writing, quit Substack, quit everything. Almost.

As I watched the text appear, as if being typed in real-time by ChatGPT, I read what felt like a legitimate blog post that covered most of the themes I wanted to write about. Yikes! It made me think of Albert Romero’s post “No Writer Is Safe From AI” where he laments how “GPT-4 has increased ChatGPT’s linguistic ability so much that its output is ‘good enough’ for most writers and editors.”

Where does that leave me, trying to write about my own work? Should I outsource it to an AI system that doesn’t actually know anything about me or my work? Should I accept its “good enough” output and hope that you, my readers, won’t mind? (Or, more depressingly, wouldn’t realize when it was wrong?)

I thought I had come up with a clever solution. I’d take ChatGPT’s post and annotate it — noting where it was correct, where was incorrect. Maybe ChatGPT could be my writing partner as I go forward. — But even that solution wasn’t so original… as I realized that just this morning I read “When A.I. Chatbots Hallucinate,” in the New York Times, which did exactly that: fact-checking various chat systems’ output.

Well, as Oscar Wild wrote, “Imitation is the sincerest form of flattery” (except the full quote was: “Imitation is the sincerest form of flattery that mediocrity can pay to greatness“). So let me go forth with this experiment in writing. Apologies in advance for this long preamble. And apolgies if the end results are less than perfect. But hey - this is just a blog, so there’s room for experimentation. Plus - I’m doing something which is contrary to the topic of this post… I’m using new, not old, technology.

(Note: I’ll put ChatGPT’s text as a block quote, and I’ll bold the text that I’ll comment on.)

Prompt: Write a blog post about the joys of not keeping up with the latest AI technology, and instead using with my old GAN code to create images. I believe that old code can help give us a more balanced perspective on the latest developments.

As an artist, I find immense joy [1] in not always keeping up with the latest AI technology and instead relying on my old GAN (Generative Adversarial Network) code to create mesmerizing images. While the art world is constantly buzzing with the newest tools and techniques, I have discovered that my trusty old [2] GAN code allows me to express my creativity in a unique and balanced way.

1. While it probably works well as a hook to draw the reader in, I think that “immense joy” is weird and strangely absolute. If anything, it’s really difficult to not get pulled into the hype of the new, or to worry that by not staying up to date I will be left behind. I’m not some some wise old man, living at the top of a mountain, immune to the pulls of the world. Instead, I think that one needs to remain vigilant, and keep perspective, on the latest trends. And to remember that what may appear new most likely has precedent.
2. My “trusty old” code sounds so sweet. Like a faithful dog that follows me everywhere. Although, when I asked GPT to make it sound more intellectual I got: “I have discovered that my steadfast allegiance to antiquated methodology allows for a measured and cultivated expression of creativity.” Oiks. My dog sounds like a nicer companion.

In the realm of artificial intelligence, staying up to date with the latest trends can be tempting. Exciting advancements in deep learning frameworks and cutting-edge algorithms captivate the attention of many artists. However, I firmly believe that there is something magical about embracing the old while exploring the new.

There is a lot of hype in this sentence. Probably the result of all the technological hype and optimism the system has been trained on. But, generally, it’s not too far off.

My old GAN code holds a special place in my artistic journey [1]. Although it may not be the most advanced AI technology available today, it continues to be an incredibly powerful tool [2] for generating stunning [3] images. It has stood the test of time and remains just as relevant [4] and valuable as when I first discovered it.

1. This is true. The code that I work with today is the same I began with in 2018. I took something “off the shelf” (ie. from GitHub), got it running, and started making images with it. It was an important transition for me — for it made tangible a shift in my own thinking. Namely that one didn’t need to be technically proficient in order to engage with new technologies. And that approach would allow more diverse voices to participate in their development.
2. Again, lots of hype. I probably wouldn’t have written “incredibly.” And “powerful” is tricky because, in the context of technology, it sounds like my GAN is technically powerful. Which it is not. It’s just some code that runs on my PC.
3. I was once told that, as an artist, I’m not allowed to describe my work as “beautiful.” I’ve since moved past that so-called rule and actively embrace the idea of beauty. But “stunning” is probably going too far. Instead I might have written “engaging.”
4. I’d break this into two separate thoughts. The first is that the issues and problems of older systems also exist for newer ones. Meaning that the problems of bias and lack of diversity are unchanged. The second is, by continuing to use the same tool, I can take the time to be more methodical and to go deeper with my investigations.

One of the most rewarding aspects of working with older technology as an artist is gaining a more holistic perspective on the field. While it's important to keep abreast of the latest developments, understanding the foundational concepts [1] and techniques is equally crucial. By continuing to employ my old GAN code, I maintain a deeper understanding of the underlying principles and limitations [2]. This empowers me to approach new technologies with a discerning eye, assessing their true value and how they fit into the broader artistic landscape.

1. I like this idea, although, again, I would have phrased it differently. A lot of my work with AI uses very small data sets to train the system — in contrast to the millions or billions of images used to train the latest tools. Using small training sets, in order to break the system, reveals fundamental aspects of the code — of AI — and how it works. I call it the “materiality” of AI.
2. One of the limitations that bothered me the most when I started with my GAN code was the low resolution of the images that were created. (That, and the incredibly slow pace at which the machine learns.) Over time I’ve found creative ways around that — including writing my own software.

Moreover, embracing older technology fosters a more nuanced exploration of artistic possibilities. The allure of new tools can sometimes lead artists to conform to the prevailing trends and techniques. However, by working with older technology, I am compelled to think more creatively and discover innovative ways to express my artistic vision. This process often yields unexpected and captivating results that push the boundaries of traditional artistic forms.

At the risk of sounding like a jerk who doesn’t like the work of my contemporaries, I like this paragraph. Thank you ChatGPT for softening this and including the word “sometimes.”

Naturally, there are downsides to relying on older technology as an artist. It can be challenging to find support and resources [1] for outdated tools, as the focus of the community shifts towards newer advancements. Additionally, using older tools may mean missing out on the latest enhancements and refinements. However, I firmly believe that these limitations pale in comparison [2] to the benefits of maintaining a deep understanding of the technology and nurturing a distinct artistic voice.

1. I realized from the day that I finally got my code to run, that the system was incredibly fragile: if I updated any one software library then everything would stop working. And so it has sat in my studio, the “update” button incessantly asking for a click, but always ignored. And I’m quite aware that at any moment things could stop for good.
2. ChatGPT really wants me to sound absolute in my conviction of sticking with the old. But I also experiment with more contemporary tools, and find that using them both gives new perspectives for each.

Ultimately, whether as an artist or a creator in any field, it is essential to continually learn and grow. Embracing both the old and the new allows us to bridge the gap between traditional craftsmanship and groundbreaking innovation. By remaining curious, asking thought-provoking questions, and fearlessly experimenting with new ideas, we can create art that transcends boundaries and contributes to the ever-evolving world of artistic expression.

It’s not a bad end for a blog post. A bit trite, and more hypey than is my usual style, but maybe I need to be less humble. This was the alternative “intellectual” ending:

In essence, irrespective of one's artistic or creative endeavors, the ceaseless pursuit of knowledge and growth remains pivotal. The harmonization of antiquarian praxis with groundbreaking innovation bridges the chasm between time-honored artisanship and progressive ingenuity. By perpetually cultivating curiosity, posing incisive inquiries, and fearlessly embarking upon uncharted artistic trajectories, we, as artists, aspire to forge masterpieces that transcend conventional boundaries and indelibly enrich the ever-evolving realm of artistic expression.

Overall, I think ChatGPT did write a “good enough” post. But is that okay? I’m certain that if I had written it myself, you would be reading something very different. This process of commenting on what ChatGPT wrote gives this post a reactionary feeling, not really under my own control. As though I were correcting the writing of someone who interviewed me but wasn’t listening very carefully.

Having tried this, I can say that I don’t want to outsource more of my writing. I want to struggle to find the words. I want to write and then realize my thoughts were wrong. I want to write and discover something new about my work and myself. And I’m fine that the process may not be very efficient.

What do you think?