Into The Weeds of Artificial Intelligence

A couple of days ago, the New York Times published an article by Emma Goldberg asking, “Will AI Kill Meaningless Jobs?” Goldberg begins the article with a techno-optimist’s perspective, suggesting that when generative AI destroys opportunities for humans to work, it’s an ultimately of benefit to human workers, because the jobs that are replaced by AI tools are meaningless drudgery. Perhaps when AI does away with these jobs, she suggests, it’s a mercy killing, because nobody enjoyed doing them anyway.

Goldberg’s article eventually considers a few of the problems with this techno-optimist outlook, but there’s a problem that she never stops to consider: Generative AI is already replacing people’s work, and much of the work that it’s replacing is meaningful work that people enjoy.

Remember the screen actors’ strike demanding protection from generative AI last year? This year, we’re seeing a similar strike from voice-over actors who have worked in the video game industry. These are not people who hate their work. They love what they do, and they find a great deal of meaning in their work. Generative AI tools are threatening their livelihoods anyway, leaving them without any reasonable path into another creative profession.

Creative writing is a kind of work for which many people feel a passionate sense of purpose. It’s never been easy to make a living as a writer, but these days, it’s nearly impossible for aspiring writers to catch a break, though, as publishers replace journalists with bots and the broader literary marketplace is being drowned in written articles, screenplays, and books that have been composed by AI chatbots. Even established writers are struggling to keep their work visible in a sea of cheap knock-offs.

Visual artistry has long been understood as the epitome of meaningful work. Sculptors chip away at blocks of marble to uncover their vision hidden within the stone. Painters take years to hone their craft, and struggle at the canvas to get it just right, but they love what they do. Now, generative AI models have been created, using unauthorized scans of human artists’ work, that are capable to creating imitations of human-made visual art in a matter of seconds.

Yes, it’s true that automation could eliminate many meaningless jobs, but the truth is generative AI is replacing the most meaningful human work first. The reason is that meaningful work takes special effort and skill. The products of meaningful human work are therefore typically expensive.

The automation of meaningful human work such as acting, writing, and producing visual art rapidly undercuts the market value of meaningful work, forcing creative workers either to quit or to accept work for a tiny fraction of the money they used to be paid.

As a result, the advent of generative artificial intelligence devastated the economic value of dedication to our work. All the rewards are going to those who take a careless, even flippant attitude toward their work. The consequences of this shift will be devastating to human ecology.

An Ecology of Artificial Weeds

Up until this point, the impact of generative AI has been discussed mostly in technical and mathematical terms. Rather crudely, economists have argued about the number of jobs that will still exist for humans in a future dominated by artificial intelligence, without much consideration for the quality of the work that people would do.

I propose a more systematic approach: To critically evaluate the ecology of artificial intelligence. While an economic or technical perspective will evaluate horses in terms of the horsepower of the work that they are capable of performing, an ecological perspective considers the kinds of settings that horses will thrive in, and the ways in which their work contributes to the thriving of other living things, interacting with the work done by other species to keep the entire system going. Ecology considers a system from multiple perspectives, with an eye toward a resilience inherent to the system as a whole, rather than the assessing the metabolism of individuals, as if they are isolated from one another.

I want to see more contemplation of this question: What are the ecological impacts of the introduction of artificial intelligence, not just for the natural world, but for human society as well?

Consider this question in the context of visual art.

I am not a visual artist. My hands are clumsy. When I try to paint, what I create is nothing like what I have seen in my mind. So, when I create an image, the results are not of high quality.

Even for the sloppy results I get, the process of painting takes a great deal of time and effort. There’s planning and revision involved. The simple act of repeatedly dabbing a brush in paint and applying the color over and over again is slow.

An example of what I get out of this process can be seen below, on the left hand side of the screen. This image took me two and a half hours to paint by hand. Aesthetically, it’s not everything I would like it to be. Even for an amateur, my skills are not great. Nonetheless, the process of creating it was enjoyable to me. I was captivated in the work. I hardly noticed the passage of time at all.

On the right hand side, for comparison, is an image that a generative artificial intelligence model was able to create, in just 20 seconds, after scanning my painting and being given the simple prompt, “woman standing in a field of robots”.

I wanted to create an image of a person standing in a field of robots to match the topic of this article, representing the feeling of humanity crowded out by automatically-produced replicas, unable compete against the sheer scale of digital production.

My painting isn’t high quality. It’s more muddled and sketchy than the generative AI replica, but it represents a specific idea that I wanted to express. Where the woman stands, there is no unoccupied ground left for her to do any work. She has desires and needs. Her work has meaning, while the robots have no purpose except to do what they are told. Yet, because the robots can be produced en masse with relatively low investment, they occupy territory quickly, and their work soon dwarves hers.

The temptation to create an image using generative AI tools is strong. It’s quick and easy to do. Yet, when it comes to this article, choosing the quick and easy option would have contradicted the central point that I want to make, which is this: The rapid reproductive rate of generative AI fosters the ecology of weeds in our society.

The ecological role of weeds is to quickly sprout and cover ground that has been disrupted. Weeds are marginal players in stable ecosystems with well-established, diverse populations of annuals, perennials, shrubs, and trees. When the botanical diversity of an acre of land is cleared away, however, and the soil is turned over by a plow, weed seeds will be the first to sprout and claim dominance, leaving no space for more slow and steady plants to grow.

Under the churn-and-burn economy fostered by digital corporations, human society has been thoroughly and purposefully disrupted for decades. Now, with the ground cleared, generative AI tools are spreading the seeds of mindless, space-filling content more quickly than any human artist ever could.

While creative human workers are slow and thoughtful about their work because their goal is to effectively express meaningful ideas, generative AI is designed only to provide plausible mimicry of human creativity, without any value placed on actual meaning. The ability of AI to generate content is practically infinite.

Consider the impact that the weed ecology of AI has had on writing. Online publishing tools like WordPress once provided a means for human writers to make their words available to others. Now, these same tools have shifted to include generative AI tools that help spammers to quickly populate entire web sites with articles written by algorithms. WordPress merely requires a few keywords and the click of a virtual button, and then almost instantly publishes an AI-generated blog article that connects those keywords together with a semblance of sense.

People can still use WordPress to publish articles that have been created solely through human effort. However, while even a brief human-made blog post might take between 30 and 60 minutes to create, a blog post of equivalent length on the same subject would take between 30 to 60 seconds to create.

In the time that a person can write and publish one article, generative AI can pump out 60 or more. There is no way to distinguish the human-made article from AI imitations except to read them, and nobody has the time to read through 60 pieces of AI slop in order to find one authentic human voice.

Regardless of the consideration and craft people put into their words, their articles disappear when they’re placed online, like rosebushes planted in fields of kudzu. In an online economy that pays writers for the advertising impressions they trigger, a weedy piece of AI slop is treated as if it is of equal value to a person’s carefully articulated thoughts.

Thus, generative AI has incentivized spammers to flood the Internet with automatically-generated slop, making the online work of human writers more difficult to find than ever.

At the same time, opportunities for writers to publish offline, in actual print, are few and far between. Newspapers and magazines have been disrupted. Book publishing has been disrupted. Their carefully-tended gardens of ideas have been roto-tilled by Amazon and social media platforms, churned into an open ground on which weeds are more likely to take hold.

Thus, while more words are available to read than ever before, the overall quality of the material that’s available is lower than ever before. The impact of generative AI has harmed both readers and writers. Similar damage has been done in the visual arts.

The communication of meaning is being lost in a cacophony of weedy noise. The ability for humans to do meaningful work is withering as a result.

Learning From Horticulture

The scale and speed of generative AI can seem impossible to resist. This impression is an illusion, however.

Every gardener knows the same feeling, witnessing the surge of weeds in midsummer that seems to completely overwhelm every plant that we have nurtured since springtime. Every experienced gardener, however, knows that the victory of the weeds is not as inevitable as it first seems to be.

The first step in regaining control is to stop continually disrupting the ground beneath our feet. Digital corporations have been undermining workers by disrupting the gardens of our work, ripping apart systems of professional and commercial relationships and replacing them with apps.

Generative AI is just the latest version of this scam. With every disruptive cycle, Silicon Valley has promised that the new tools will enable us to be more productive, and generate more wealth, if we just clear away everything that we’ve done before to make room for their new scheme. They made the same promises with social media, gig work, NFTs, and the metaverse. The promises never come true. Every time we fall for them, we just find ourselves deeper in the weeds.

So, let’s stop ripping up all of our work every time a new thing comes along. We don’t need to start our gardens back to the bare ground every year. We can focus on nurturing perennial bodies of work that keep growing over time, instead of relying solely on short-term, seasonal projects that have to be replanted over and over again.

The second step in regaining control over our work is to suppress the growth of weedy generative AI. This doesn’t mean that there can never be a place for generative AI tools in our work. Weeds have a place in a good garden, but we need to cut them down to size every now and then and disincentivize their growth.

To protect the ecological diversity and stability of human society we need to bring the rapid growth of generative AI back down to a human scale. The best way to do that will be to tax the work of generative AI, in the same way that we tax human work.

Right now, powerful digital corporations like Microsoft and Google are using massive amounts of capital to try to replace human work with generative AI. They’re investing immense amounts of money to make generative AI tools seem less expensive than they really are, offering these models at absurdly low prices, or even for free, so that human workers can’t hope to compete. Only later, after human creative workers have been eliminated from the marketplace, will users of AI be asked to bear the true cost of the technology.

We have the opportunity to interrupt this part of the digital dependence cycle by imposing an authentic cost of generative AI through taxation. What if there was a tax of ten cents on every prompt on ChatGPT, DALL-E and similar platforms?

Human artists and writers charge money for their work, and they charge much more than just a dime. A ten-cent tax would slow down the use of generative AI, leading people to be more thoughtful about their interactions with it, instead of peppering it with ridiculous requests, as if AI were an limitless resource.

The proceeds of this ten-cent tax could fund the social infrastructure that keeps the meaningful work of human creativity going. Libraries, galleries, theaters and other offline places for people to display their work could receive some of the funding. Newspapers, publishing houses, and community studio spaces could benefit as well.

Ten cents per prompt is a low price to pay for an original image, a poem, or an essay. If we’re not willing to pay even that much for the work that has traditionally been the most meaningful of all human labor, then we should not be surprised to find our social spaces overrun by weeds.

Into The Weeds of Artificial Intelligence was originally published in Becoming Human: Artificial Intelligence Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.