This Needs Fixin'

Boosters of the generative AI slop that’s taking over everywhere these days like to call everyone who objects “Luddites”.

Which is actually true, but not how they think.

The popular conception of the Luddites is they they simply objected to technological advances, but this is far from the truth.

The Luddite movement objected to the deployment of new technology in a way that impoverished those who had worked with the previous generation of machines and enriched only the owners of the new machines. They did not interfere with businesses who set up schemes for compensating workers, for example.

Their issue was not, in general, with the technology itself but with the mill owners and what the Luddites saw as an obligation to share the benefits of the new technology with the workers.

So, yeah, there’s a hint of Luddism here.

Today, employers anticipating great productivity gains from generative AI deployments are firing large swaths of workers whose work they think will soon be made surplus by the new machines.

The Luddite response to new technology is that productivity gains should be shared with the workers by decreasing the expected working hours rather than being hoarded for the benefit of the already-wealthy owners.

So call me a Luddite, fine; I am that.

But that’s not all that’s going on here.

The Luddites were responding to a new technology that actually worked.

The new shearing frame really did significantly reduce the labor (and expertise) required to produce the same quantity of goods; it was a genuine leap ahead in technology.

Whereas generative AI, for the most part, is not that.

Generative AI is not actually “Artificial Intelligence”; there is no real “Artificial Intelligence”.

Seriously: it’s not a real thing.

Yes, the machine can produce large blocks of plausible-seeming text; this is true.

But since it has no concept of utility, or truth, or functionality, or anything you really can’t trust that to be useful, or factually accurate or even compilable code.

You still have to spend the time, and have the expertise, to evaluate the machine’s output.

So, generative AI offers no productivity gain and no reduction in expertise.

Unless you skip the evaluation and just go with the slop.

But that’s not a gain in productivity; that’s a reduction in quality. And the less expert oversight you have on it, the bigger the reduction in quality will be.

The hacker and professional information security communities, for example, are anticipating a significant decrease in overall computer security as a result on introducing generative AI “assistants” into coding shops. The consensus among the serious practitioners is that this will lead to a huge boom time for their businesses.

And they’re totally correct: “vibe coding” may give you something that looks right on the surface, but under the hood will be loaded with unintended behaviors with potentially catastrophic consequences.

So, a better model for thinking about generative AI is tulip mania.

It’s taking over the business markets ludicrously fast, but it’s not built on anything fundamentally real.

And like tulip mania, it’s going to just wreck any business that gets involved in it.

And the Luddites are just trying to limit the damage.

•