The Reluctant Embrace of Generative Artificial Intelligence

 I originally had a pretty significant post scheduled to go out this time, but opted to hold it, as I needed to link to this post within it — albeit for a minor reason. Stay tuned for a fun (to me) “announcement.”

Capitalism requires growth for its continued existence. Growth requires more things to be done faster and for far less capital. This is, in my personal view, a vulgar and inhumane artificial construct causing unmeasurable damage and suffering — yet also something I greatly benefit from and unfortunately has no universally agreed-upon or feasible replacement.

As capitalism requires things to be done faster and cheaper, capitalists have always devised new ways to partially or fully replace workers. This dates to at least 200 BCE, when Romans invented the water wheel to automate the grinding of grains. (I’m sure a later example exists, but that one came to mind first.) A more recent example I often think of is a family friend’s cousin telling me at a barbeque a few years back at how he was the “matting expert for every photo lab in the city” whose job suddenly went away in the early ‘90s after the introduction of Photoshop. This is a fairly roundabout way of saying as time moves forward, more inventions and automations are introduced that could theoretically augment a worker’s abilities but are often used to outright replace said workers instead.

Generative artificial intelligence is no different. The best large language models — which, in my opinion, are those developed by Anthropic — have more or less the capabilities of a junior-level worker a few years out of college. Over time — and likely over the next 12 months — these capabilities will likely grow to the point where this junior-level worker will graduate to mid-level status (and/or beyond), giving anyone with internet access, a few bucks, and the ability to reason access to their own skilled non-labor worker.

While this sounds exciting — I can pinpoint dozens of hours of work and personal tasks I’d love to pass to someone else each week — it is also a Trojan Horse for those most advantaged of capitalist workers. In previous eras, where business owners and executives were able to automate labor-intensive tasks that required no diploma (farm work, manufacturing, warehouse management), this era’s introduction of generative artificial intelligence allows companies to jettison entire teams of workers with advanced degrees and replace them with an API call. This is not a hypothetical; it is already on the horizon, if not happening elsewhere. It is, by and large, one of the key reasons every investor has extensively invested in related companies and systems.

How I Was Replaced by a Large Language Model

On a personal level, my previous position from three years ago (as Head of Content at a fintech company) was made redundant, and my work was used to inform a large language model of the company’s creation to pump out bespoke content in my voice as they saw fit. I have no legal ownership of the content, as my content created over the course of two years belongs to that company. (It was also the most purposefully uninspired material I ever created, so I am not hung up specifically on their use of my ersatz voice.) 

In fact, I recently pivoted to focus entirely on PR and corporate communications — a position largely irreplaceable by generative artificial intelligence — purely because “content marketing manager” positions have almost entirely been replaced by usage of large language models. (I should note here that I am hiring a content manager replacement at Reality Defender to write, manage writers, and use generative artificial intelligence to create content all at once.)

Why I Embrace Generative Artificial Intelligence 

I use large language models regularly. While I work at a company that “uses AI to catch AI,” I have wholly and extensively embedded Claude into my personal and professional life long ago and for four key reasons. Such a decision was not one I made lightly — especially due to unchecked growth capitalism being the main driver of these systems’ existence.

Self-Preservation

Fear is a great way to drive anyone to the unnecessary capitalist creation that is general artificial intelligence. As I started before, I am a beneficiary of capitalism. The growth and popularity of generative artificial intelligence models have made it so that to continue benefitting from capitalism (as opposed to being a casualty of it, as I’ve seen firsthand), one must abide by its abiding rules. To continue to be a skilled worker, I and others like me must be highly skilled in understanding and adopting things like large language models.

This is unfortunately not a topic up for debate in any work environment. Those who prefer the old way of doing things — not using artificial intelligence — will be able to keep their moral high ground but largely find themselves replaced. This is what generative artificial intelligence was built to accomplish in the first place: make many redundant and keep the few who know how to use it until it is time to replace those people as well. More people are using it than you are led to believe, and if you are reluctant to embrace generative artificial intelligence, you are likely several steps behind them in adoption and skill.

I apologize for the frank and bleak outlook, and I fully believe that this should not be the direction we as a world head towards. Yet at the same time, there are no societal structures in place to prevent this from happening — in America or elsewhere. The endgame here from leading technologists and investors is and has always been fewer workers, more productivity, and greater profits. Thus, out of fear of not being able to earn an income in the next 25 years and benefit from capitalism (which, in my world, is simply to support my family and not die early and penniless), I have taken the time to master the usage of large language models and other major inventions born out of the generative artificial intelligence boom.

Augmentation, Not Replacement

I mentioned above that large language models are about as skilled a good junior-level associate three years out of college. What I neglected to mention is how they maintain this skill level across most industries and use cases.

I have used large language models to help with professional work — brainstorming, strategizing, and plotting — as well as personal and creative work. Many of the blog posts we release at work start from heavily restructured outlines follow our company messaging to a T spat out by Claude, then written into final works by humans. Meanwhile, when my in-laws were here for Christmas and I had to create sweets that were lower on the glycemic index, I fed a recipe into Claude with the ask that it returned the same recipe, albeit with maple syrup instead of cane sugar. It did so quickly and expertly, reducing liquids knowing that syrup would be added.

Perhaps what impressed me most is when I fed Claude — which famously does not (re)train on inputs — my novel from a year ago and asked for discovery of plot holes. Not only was it able to differentiate between my own failings and stylistic decisions or an unreliable narrator; it had also suggested minor changes I could make and where I should make them.

These use cases are impressive but still low level in terms of what large language models and other generative tools can do. (I have a whole website dedicated to coding experiments created entirely with Claude.) As the reluctant embracer of generative artificial intelligence — one who can prompt its models with ease and skill — I am nonetheless scratching the surface in its capabilities.

Technological Curiosity

I am often an early adopter of some new technologies. My first use of the internet was via command line. I learned HTML in 1995 (and was taught by Duke Ellington’s cousin, of all people). I’ve used RSS to consume news since before it became standardized. I was even one of the first ten thousand users on Facebook, though I no longer hold an account.

I say some new technologies because not all new inventions are necessary and will be adopted widely. I think VR as a technology is as useless now as it was when Virtual Boy came out in 1995. I feel that centralized social media is largely a stain on the history of human communication and has pushed society in a largely negative direction, to put it mildly.

Yet some technologies I embrace pre-emptively knowing that they will one day be a work/living requirement. This applies to generative artificial intelligence as much as it applies to owning a smartphone. Simply put, I embrace early so as I don’t get left behind professionally (and, sure, socially.)

We’ve Already Used Them

Artificial intelligence is nothing new. If you used any modern computing device in the last decade, you used a primitive form of generative artificial intelligence. Think Siri, the object detection tool in Photoshop, an email scheduling time optimizer, a social media algorithm, an advanced spell checker, a face tuner, and a million other things in between. Large language models and other generative tools are these things condensed and amped to 11, for better or worse. Using generative AI was not a difficult decision to make, as I had been using some of these tools for years prior. 

There is no flow chart on the morally acceptable/societally appropriate use of some artificial intelligence but not others. There is either all or nothing, as it all comes from the same source.

---

Claude has some guidelines built in — you can’t ask it to give you, say, the formula for plastic explosives as you can with open-source large language models — as well as a structure that mostly pushes it in the direction to help the user without putting them in harm’s way. That said, it is a tool used for efficiency and growth — one that, like all other major large language models today, is also embraced by the military industrial complex. 

It should also be noted that powering Claude and other generative artificial intelligence models requires an insane amount of energy, to the point where it is expected that this usage will accelerate climate change and its devastating effects. This energy consumption will happen whether you do or do not use artificial intelligence, and it seems that every technology company has almost entirely removed any pledge towards a better, cleaner world in the pursuit of profit from these tools.

Such is the amoral nature of capitalism, and something to keep in the back of one’s mind when they’re using the very same tool to find patterns in a spreadsheet. Of course I wish for an alternate world where these systems were never created, especially as it would mean a safer, cleaner world where hundreds of millions of jobs are not at stake. Unfortunately, to continue living in this one and ensure a greater probability for survival within its many oppressive systems, one will need to embrace generative artificial intelligence as they have credit scores, debt, and other artificial constructs in which capitalism thrives and punishes the least fortunate. 

Leaders from all political parties in our country have mostly thumbed their nose or threw up their hands at the mention of any alternative or safeguard against the creep of generative artificial intelligence into the capitalistic structure. I truly believe this is because they will either benefit greatly or not lose much from its addition to everyday life.

While I am an idealist and optimist — one who believes change can come from wide support of measures against everything discussed above — I also realistically know that there is apathy and indifference towards making this a major issue compared to, say, climate change, housing, or the cost of living. I do hope, however, that enough people will see a sliver of the negative impact that generative artificial intelligence can/does have on their lives to rally support against its ill-intended impacts before it enmeshes itself as a permanent fixture into our lives.