LLMs, Mechanical Work, Craft, and You

Are LLMs Evil, or Just a Tool that Can Be Abused?

Not all work needs to be done by a human (Photo by Simon Kadula on Unsplash)

Yesterday, I was talking with a copywriter who also uses LLMs in his work. I commented that I hated LLMs, and later, in email, he referred to the LLM he uses as “the evil AI man.” Of course, both of these are exaggerations.

There are things I hate about LLMs, but I also recognize potential value. My acquaintance obviously doesn’t really think AI is evil; it’s just a convenient way of expressing a more nuanced point.

Casual conversation tends to simplifications and exaggerations, which is fine in many contexts. But simplifications and exaggerations are not great as the basis for decisions. Decisions want nuance. We don’t want a broad condemnation to blind us to positive potential. Nor do we want a broad appreciation to blind us to real problems.

In our attention-limited society at present, we get simplifications and exaggerations, and that’s a big problem.

What Can LLMs Do?

When I wrote about LLMs previously, I wrote: “an LLM… is literally a device to re-create that which is already commonplace and conventional.” And there is a lot of work that is commonplace and conventional. In my work, there are two areas where LLMs might be of use: updating my website and writing R code for simple data analyses (I say R because I’m familiar with R, but I’m sure that if an LLM generated python code for simple data analysis, that would be fine).

Simple Web Design

My copywriter friend uses AI to make websites, which are mostly conventional. My website, for example, is conventional in everything but the content — there’s nothing structurally or technically special. I’ve used WordPress in its most basic functions for going on 20 years.

If I update my website, everything I do will be conventional, excepting only the writing, which is where my professional skill is centered. Format, layout, color scheme, structure — all that is going to be conventional. I have some opinions, but realistically, I’m not particularly into the appearance as long as it doesn’t look silly. If people are going to be interested in my work, it’s because I have good ideas expressed well, convincing people that I will be a good resource. I know the appearance matters to a lot of people, and want to have something good for that reason, but the visual experience just doesn’t really move me. Something visually conventional will work for me. Letting an LLM generate that seems entirely reasonable, since I don’t want to mess around making a new website, and conventional is absolutely acceptable.

Once the design is there, I can add my own personal touch using my skill as a writer.

Simple Data Analysis

Another area where I consider using LLMs is with data analysis. I don’t do much data analysis, but I sometimes use R for simple stuff. When I have worked with R, the vast majority of my work is to generate code that does entirely conventional stuff — data input and output, mostly, along with simple analyses.

Beyond choices about which data to analyze, everything I want to do is conventional and commonplace. To do anything, I have previously just looked for examples of code I could copy. An LLM could do that. And all I lose is development of my skill with R, which is not an important skill for me.

Frederick Taylor and Scientific Management

In the late 19th century, the idea of scientific management, spearheaded by Taylor, was popular (among academics and business executives, at least). Taylor is known for carefully documenting the motions of each worker at different tasks, and analyzing those motions. When his conclusions (e.g., “this activity should take X seconds,”) were applied to human workers they were dehumanizing.

But we’re talking about AI. To the extent that we can reduce a task to simple motions that need no human intelligence or adaptability, then dehumanizing understanding is fine and entirely workable.

That kind of simple understanding of the parts of a simple task are exactly what makes LLMs suitable for basic website design or R coding.

Special Sauce

I believe in diversity.

Diversity is good for ecological systems, as it allows for adaptation to changing conditions: a diverse ecosystem has beings that can fill new ecological niches. A species with diversity may survive where homogeneity won’t.

In some economic theory, diversity (i.e., consumer choice and producer competition) is good as competition between different approaches leads to efficient use of resources and innovation. )This is the basic theory, anyway; reality isn’t so simple.)

Diversity is good for societies. Or at least some people think diversity is good for societies. The basic idea of Democracy is that there should be a diversity of ideas that are debated before a single plan of action is chosen.

Although some valuable diversity can come from remixing old ideas (which LLMs do), a lot of valuable diversity — the real novelties — come from unexpected shifts, “mutations” of the old material that are not simple remixes. New perspectives on processes, for example, won’t come from simply remixing old ideas.

We, as humans need to cultivate that which makes us unique. I work on my writing and my thinking because I think that I approach those in a special way.

LLMs Are Not Evil; People Sometimes Are

Using an LLM is not evil. Misusing one is. LLMs don’t think; they don’t feel; they have no moral compass or compassion for other living beings. They are just a tool. They are not inherently evil.

Make your own art! As your skill grows, the struggle and frustration lessen and the pleasure increases. (Photo by Olya P on Unsplash)

AI and You

If LLMs destroyed our individuality, that would be evil. If they stop each of us from seeking our own special sauce and they deny society healthy diversity — if they keep us from learning and growing — that would be evil.

If you want to develop your own special sauce as a writer, that’s a task only you can do alone. If LLMs take away the challenge, they will also take away the growth and the satisfaction that come with trying to do your own thing.

Writing and helping other writers, researchers, and philosophers is special to me, so I work at it. I’m going to keep working at it because it’s my special sauce. Website design isn’t. I wasn’t going to redesign my website myself, so I’m not losing personal growth if I have an LLM do it. (There’s a whole question of transferring work from humans to LLMs, but that’s outside the scope of this piece.)

Do your thing. Develop your special sauce. Don’t let the simplistic conventional competence of an LLM steal your chance for growth. Grow your craft, your skill, and your insight. And outside that craft, there may be tasks that could reasonably be done using an LLM without stealing your potential for growth.