Why Does All The AI Write Instead of Doing Something Useful?

There exists in the Pacific Ocean a kind of man-made island.

No one will admit to attempting to create this island, despite its impressive size -- larger than France. If pressed, though, or put under torture (say, having to listen to all of Barry Manilow's greatest hits), many of the business executives responsible for its creation would be quite content with its existence. After all, it represents quite a lot of cost that they do have not had to take responsibility for. I am talking about the Great Pacific Garbage Patch. Calling it an island is something of a stretch -- its mostly microplastics and so it's more of a soup, and there are actually two of them on opposite ends of the Ocean. But its existence is not just a condemnation of our environmental stewardship but a commentary on the structural issues that drive the inefficient and downright harmful usage of technological capital such as so-called AI.

Much of the hype around imitative, or generative, AI systems focuses on what we might call human creative or human centered areas. Writing, illustrating, programming, customer service, certain kinds of medical services such as psychiatry and therapy -- all areas that have been bastions of human creativity and human empathy and understanding. Now, these are not the only areas machine learning systems have been introduced, but this latest round of hype and effort has been largely focused in these areas. The question is why? Why go so hard at areas of human creativity and empathy? I think a significant part of that answer is that because, as a society, we have abandoned even the idea of societal control of the economy.

Efficient markets are not a natural phenomenon. The process of selling may arise out of our natural human cooperative spirit, but it is also one of the areas most likely to attract assholes. Not everyone, not even most, who are in business are assholes, of course. But the incentives for assholes to thrive are certainly there. What better way for an asshole to thrive than an environment where one person can, through a disparity in knowledge or power, take advantage of another person for their own personal gain? And so economic history is largely the story of assholes taking advantage of or harming more and more people until society says enough and takes steps to reign them in, to one degree of success or another. Unfortunately for us, we live in one of those times when the assholes have the advantage.

The last forty years have been an orgy of deregulation, lax oversight, fantasies about the supremacy of shareholder value and corporate personhood (as if corporations were not the explicit creation of the state meant explicitly to benefit the state), and a successful assault on the very notion that society has any obligation to control the excesses of the assholes who might be influential in the economy. As a result, we have seen the hollowing out of the middle class, a massive increase in income inequality despite large productivity gains, and environmental degradation on a planet changing scale. What, though, has any of that to do with finance bros trying to turn their computers into Maya Angelou or Carl Jung?

A lot, actually.

Because markets have been allowed to become completely disconnected from their societal purposes, we now live in a society subservient to the goal of making money. Meaning that labor is seen as a cost, not an investment, quick cash outs are the goal rather than sustainable businesses, and the more costs you can push onto the society as a business, the better. Large language models and their visual equivalents check all those boxes.

In theory, imitative AI can replace labor. They can write screenplays, they can create programs, they can draw pictures, they can listen to depressed people's problems, they can handle customer complaints. And they can do it while being trained for essentially free -- just the cost of some servers, some electricity and scraping all the data you can eat from the internet. People have already solved these problems, after all. And so a cost goes away and you are rich beyond your wildest dreams of avarice.

Of course, in reality, it's not that simple. Since these machines can only regurgitate what people have done already, and since they cannot understand context, there is a limit to the problems they can solve. No blockbuster will ever come out of one of these machines in a filmable format -- there is simply too much other stuff a writer needs to know about film production in order to make a shootable script. But companies still probably think they can get a rough draft and then underpay someone to "edit" it into shape. Similarly, if you can produce a ton of boilerplate code and then have two people instead of ten people to whip it into production ready shape, that has to be a win, right?

Of course, we already have a lot of this functionality today: no on writes boilerplate code unless they are a masochist, and almost all hospitals use some form of screening procedures to help the initial triage, for example. You might be able improve things around the edges. But any significant money is going to come from the bubble window before it becomes clear that forcing the world to meet the hype isn't going to work. But at that point, the people behind these systems will have cashed out. After all, in the modern economy, hype and grift are a faster, easier way to riches than building actual products.

And that is the rub. If we lived in a functioning society, that state of affairs would be unacceptable. We would be managing our markets to limit the damage the assholes could do. We would be trying to control externalities, we would encourage policies that forced corporations to recognize the interests of all their stakeholders, not just shareholders, we would treat labor as an investment, and we would incentivize companies and products that solved real world problems, not grift and flim-flam. None of the above is radical or unachievable -- everything that I have outlined has been done in living memory. We have just been conned by the assholes into thinking it's unnatural or inefficient or some other buzzword meant to scare us into forgetting that we live in a society, not an economy.

As long as we forgot that the assholes are going to keep winning and use genuinely useful creations like machine learning to line their pockets instead of make society better. The robots could do the things that help us rather than hurt us -- as long as we don't let the assholes control them.