The Model Kit

I don’t like generative artificial intelligence as it currently exists. Part of this is just a personal reaction. As an amateur artist, I felt my stomach drop when I saw a midjourney generated image win an art contest and the person behind it responded to the backlash by saying something like get over it, art is over, the machines won. As an educator, I watched in disbelief and disgust as academic dishonesty took a quantum leap forward with the proliferation of large language models. Moreover, it feels to me like the heat death of human creativity. I’ve heard about the upside too, all of the ways this technology is going to make our lives better and unleash everyone’s potential, but I remain an A.I. malcontent.

I’m familiar with the moral debate about training these models and I know most of the arguments for and against the widespread proliferation of A.I. usage. I don’t intend to delve into any of that here. Instead, I want to offer a rationale for why I personally choose not to use generative A.I. at all, even if it could make my life a little more convenient in the short term. I have in mind a simple decision matrix, structurally similar to Pascal’s wager, pictured below.

This compares the decision to embrace or abstain from using generative A.I. under two possibilities: that A.I. takes over everything, or that A.I. turns out to be just another tool. Now by ‘A.I. takes over’, I don’t necessarily mean that some powerful, sentient artificial general intelligence like Skynet from The Terminator or the Master Control Program from Tron has seized control. I simply mean that generative A.I. will have so thoroughly transformed the economic order and everyday life that, for better or worse, no human being’s skills, knowledge or abilities make any real difference anymore. The other possibility is that A.I. is still here to stay but ends up transforming our world to a lesser degree. It becomes just another tool, albeit a ubiquitous one. Like the smart phone. I won’t bother with other unlikely possibilities such as everyone gets sick of A.I. and it goes away in a few years.

As I see it, in the case that A.I. is soon to take over everything, it won’t matter much whether I choose to adopt it or abstain from it in the interim. We will all be obsolete: scientists, coders, engineers, artists and prompt whisperers. Let’s hope we’re taken care of.

If A.I. proves instead to be a ubiquitous tool that still requires a semi-skilled user, then the problem I see is that people’s cognitive ability and skills may wither away from disuse as we become overly reliant on generated content to replace our own output. This creates an opportunity for anyone who chooses to practice any given skill unaided by generative tools. Under such a scenario, such a person’s skills would become more valuable because they are rare. On the other hand, the better and more user friendly A.I. tools get, the more everyone will be able to use them to get more or less the same results out of them, conferring nothing special for having embraced their use. Therefore, abstaining from A.I. could lead to the best outcome, while having no worse a possible outcome than embracing it.

Let me address the limitations of this model and anticipate some objections. It might be said that I am presenting a false dilemma because there are endless intermediate positions between total abstinence and the degree of reliance on A.I. that is likely to lead to the atrophy of skills and cognitive ability. I acknowledge that there is a spectrum of positions, but what I suppose I am saying is that the more one embraces A.I. usage, the more one’s skills will atrophy, and the more one abstains and practices these skills independently, the better these skills will be maintained. Some people will argue that using A.I. will not atrophy but actually enhance their skills (the underlying skill (writing, drawing, etc.) not the skills specific to prompting the A.I. or setting up an A.I. workflow). Here I think that people are confusing augmentation for improvement. A.I. may be augmenting your abilities when you use it, allowing you to output higher quality content than you would otherwise be able to produce, but is it making you any better at doing those things when the power is off? I suppose the jury is out on this one, but it seems to me that walking around all the time in a mech suit that does the heavy lifting for you is liable to make you weaker, not stronger.

Another objection is that my entire premise is wrong because in a post A.I. world, these human skills won’t be valuable anymore. Who cares if you can draw a picture, write a book, write some code, or explain the significance of Boyle’s Law to an undergraduate when a passable synthetic approximation of any of those things is available at the click of a mouse? I have several responses to this objection. First, excellence, creativity and innovation are always needed in all skill domains, and generative A.I. of the kind we have now may not ever fully surpass us in these attributes. If it does, then we’re probably headed for the scenario where A.I. takes over everything, and none of this matters. Second, the fact is that many people care about the process, about authenticity, about having something that a person crafted in some kind of time-honored way. And they may be willing to pay a premium for it. Like the Narrator said in Fight Club, “I had it all. Even the glass dishes with tiny bubbles and imperfections, proof they were crafted by the honest, simple, hard-working indigenous peoples of… wherever.” Granted, this doesn’t apply to all domains (alas, I doubt many people care about ‘hand-typed computer code’). Lastly, it seems to me that from a self-care perspective, you want to try to avoid putting too many things on Easy mode in life. Many of us must go out of our way just to get enough steps in per day to maintain our long-term health. I’ve written elsewhere on this blog about the benefits of thinking and doing things for oneself so I won’t repeat them here, but this is going to increasingly be something that we must consciously seek out. So there are intrinsic benefits to cultivating these skills regardless of the economic situation of the future.

This concludes my self-interested case for generative A.I. minimalism. I should say that I’m not trying to moralize here or shame anyone for choosing to use it. It’s hard to avoid it, frankly. And there are plenty of other things I should probably boycott but don’t out of convenience, so I have no leg to stand on anyhow. Finally, I must admit that this is an easy choice for me because as I mentioned, I find generative A.I. viscerally repellant. So, maybe I am wrong and will miss out, but this is my play and I will live with the consequences.