Suicide by LLM

I'm not Cory Doctorow

This means, among other things, that I do not get paid to talk in public, or to write books, and it means that I do not have a compulsion to publish a blog post every day (and it shows).

However, it does also mean that I won't have committed “suicide by LLM”, trying to defend the indefensible for the stupidest possible reason ever: spellchecking.

I'll admit it: my posts have typos. Sometimes, when I go over one of my previous posts, I spot all the typing errors, misspelt words, doubled words, forgotten upper cases and all the other writing issues that abound in everybody's long-form writing undertakings: the reason why serious publishers have editors and reviewers.

For us bloggers, however, these are issues that for the most part are caused by us forgetting to enable spellchecking in our text editor (vim in my case). And sure, if you're Jerrold Zar you can come up with a funny and poetic way to remark the limits of these tools, but these limits are there regardless of the specific shape and form of the tool, its energy requirements both in the preparation and deployment phases, and of the ethical underpinning of its development and use.

Trying to justify the use of an LLM for this particular task, something that word processors (and standalone programs) have already been able to do for decades, with quality steadily improving over time without any of the ethical issues surrounding the tools, really takes a very special type of mind, probably the same that fell for the questionable idea that it's possible to “install our own fire exits” on a billionaire's network.

Cory Doctorow is far from the only one falling for these ruses1, but with the news of Google and Microsoft paying influencers to promote adoption of , it's inevitable to question the motives behind such a choice. Doubling down on its defense against criticism sounds particularly hollow from someone who was so keen on clarifying the true nature of the Luddite movement. Flash news: running the model(s) locally has little to no bearing on the ethical issues concerning their origins, motives, and uses.

Adafruit: the poisoned apple

Cory Doctorow is not the only significant figure in tech that seems to have decided to commit suicide by LLM. Shortly before his manifestation on the use of LLMs as spellcheckers, Adafruit Industries revealed they had started using them to develop their board designs.

This revelation suffered such a massive pushback on the Fediverse that it was inevitable to compare it to when Raspberry Pi decided to commit suicide by cop (for once, not in the usual sense of intentionally misclassifying murder by law enforcement, but an actual mediatic suicide of the enterprise' reputation).

And as it usually happens in these cases (remember when Framework Computer decided to commit [suicide by Nazi], pardon, by big tent?), instead of coming out with an apology, they doubled down with one of most inane reversals ever seen on the Internet, considering that basically the entire reason of existence for the LLMs used by Adafruit is specifically to accelerate the extraction of value (stealing) from the disenfranchised in favour of the rentier class.

Generative Large Brainworm

I have long ceased to idolise prominent figures in tech, so seeing them fall to the LLM/genAI brainworm doesn't really make me question my beliefs; however, for sure it does not spark joy to see them drift away like this.

And it is a brainworm; or rather, more appropriately, a mindworm, because it has no physical body (although it's quite possible that its influence does affect the brain biochemistry —I haven't looked into that yet). It is a mindworm, because it's parasitic in nature, infectious, and damaging to the host. It latches on the same mechanisms of our pareidolia, pushed through by the filthiest of psychological manipulations that marketing has been developing for decades, and leads to a general loss of cognitive functions, a decline that goes largely unnoticed by those who get trapped in it, but is clear as day to anyone surrounding them.

I'm guessing I'll stack that with all the other reasons why this grift is so damaging for the Internet, combined with the indiscriminate scraping I've finally had to defend my server from, the pollution of the commons that will make older and genuinely man-made web pages as precious as low-background steel, the massive jacking up of memory prices, the premature storage shortage, and all the other general attacks to personal computing.

And I'm well aware of how First World problem such complaints may be, but after all, let's be honest here: the root lies in the same extractive, colonial mindset that has dominated the last centuries, keeping under the heel people with much worse problems.