LLMs and Marx

Preface / LLM bollocks

Firstly, I’m not much of a clanker. I don’t use LLMs to write my posts, nor do I paste its output into a chat to ‘answer’ the questions of colleagues or even use it much at work in an agentic way. I like to deeply understand the tech stack I’m using, and I still create and review every line of code before it goes in front of a human being.

There’s probably 5 posts on LLMs I could create regarding things that are bugging me, but most of them have been covered elsewhere and better, from “Feeling sad about AI” to (various) posts about the rules of engagement on software projects when contributors are writing code in vastly different and opposing ways.

LLMisms and spotting clankers

I want to scream when I spot LLMisms. If you can’t be arsed to write it, I certainly won’t take the time to read it.

However, it’s arguably even worse that models like Gemini Pro 3.1 (which this boomer uses for research / Q&A etc.) doesn’t seem to have such obvious tells. With my small user prompt (thrown into the bottomless pit of “prompts that claim to do something but probably don’t”), its prose is plain and easy to read:

Always write using UK English. Do not use “Americanisations” (including US English spellings or US cultural phrases). Be matter of fact, not overly enthusiastic. Avoid using sycophantic-sounding phrases like “you’re absolutely right” or exaggerating results with “this changes everything”. Be boring and factual. Where possible, provide citations and base confidence levels on the level of evidence available. In cases where you’re critiquing a take or analysing evidence, do not assume I am correct. Critically evaluate the results and consider alternative views if there is evidence for them.

So I ask you: what’s worse, an LLM that breathlessly mangles the English language (making it easy to spot), or an LLM that writes plainly and blends into the background?

I thought I was pretty good at spotting LLM-authored dross, but how many AI articles have I unknowingly devoured because the models write plainly like Gemini? How long until the RLHF starts to prioritise ‘sounding human’ for a few dedicated rounds vs. going all-in on ‘intelligence’? Urgh.

On Marx and Marxism and the Overton window

This was a swift turn, wasn’t it? While I’m sure this post will have “Stringer Bell just took an intro to economics class and is a fucking idiot” vibes, I don’t care. It’s my blog, not yours!

“Y’all heard of Worldcom?”

I find the whole Marx/Marxism thing interesting because there’s a giant band of pitchfork-brandishing maniacs – who’ve never read a book in their life – who enthusiastically describe anyone left of centre-right in politics as “a woke Marxist”. This particularly applies to hard-right “take our country back” types both here in the UK and across the pond. They confidently, and proudly, know nothing about the subject matter. Ignorance is Strength and all that.

I genuinely wonder if this has subtly moved the Overton window (as in, it is now unacceptable to even mention Marx because of who’s in the White House) until self-censorship kicked in. How often do you actually read or watch anything about Marx in the media, even when some of the themes are undeniably on-the-nose?

My dad is avowedly not a Christian, but he read the Old Testament cover to cover a few times just to see what it’s all about (I’m reliably informed that there’s a lot of smiting). He’s read the bible more times than a sizeable chunk of Christians. And yet he’s not a Christian.

Hopefully the analogy is clear.

On (trying to) read Marx

If you’re like me, you’ll find much of the material difficult to read because:

1. It’s originally written in German
2. It’s 150+ years old
3. It’s full of philosophical musings that reference Hegel & co. (who I have not read)

So while plebs like you and I might struggle to read Das Kapital, we can can read The Communist Manifesto and understand most of it. Indeed, it was written for the masses.

If you find it hard going or don’t understand something, you can just try reading another chapter, another book or even getting an annotated version.

Anyway, on to …

Marx’s theory of alienation

Given the hype and the predicted upheaval of society (brain rot, educational decline, loss of jobs, people starting to hate their jobs, techno-feudalism), I’m surprised that barely anyone in the mainstream press seems to be writing about Marx’s theory of alienation.

The part that really interested me was the second point of alienation:

Second, they are alienated from the activity of production itself, which is experienced not as a fulfilling expression of creativity but as coerced, meaningless toil.

History Does Not Repeat Itself, But It Rhymes. It’s like a re-telling of the industrial revolution and the effect it had on tradespeople being forced out of skilled jobs. 200 years ago, weavers smashed up power-looms. Then assembly line production turned that up to 11 (ironically, Ford allegedly took inspiration from slaughterhouses). Someone who used to weld an entire car together was reduced to fastening bolts – goodbye to mental invigoration and job security.

No no no, we are the ones who automate your job, remember?!

Most knowledge workers (IT and programmers in particular) thought their jobs would be the last to be automated, materially changed or reduced in scope. After all, we’ve been hearing about Fourth Generation Languages for decades, watching so-called miraculous managed systems drag companies into the mud and witnessing cheap outsourcing blow up in people’s faces.

“Hey guys, we’re the smart people who automate other people’s jobs!”

Poetic justice, non? We didn’t put 2 + 2 together and realise tools that provided a scaffold for making sustained progress (e.g. compilers, linters, automated tests etc.) would be the very same tools that made our jobs amenable to automation.

There’s more on this (the RAM crisis caused by hyperscalers, loss of personal computing, the threat of techno-feudalism, OpenAI’s potential ‘borrowing’ of results from someone using their service, USA! USA! USA! tech hegemony, data centres and climate change) but that’s enough chit-chat for today.

Anyway, there’s no real, neat punchline to this post. I just think more people should be reading this stuff and talking about it more because it seems highly relevant.