I didn’t plan to have an opinion on AI. Really. It didn’t seem that much of a moral concern to me, and still doesn’t. However, there’s something I do care about: content quality.
Introduction
Even though I didn’t want to spend my time thinking about the morality of AI, I did end up building a personal opinion.
The other day, I was watching a video about AI-slop math videos (3b1b style) and AI as a helper tool. The guy said that, for him, as long as the AI use is minimal, it’s OK. He also said that, for example, generating all the animations with AI wouldn’t be okay.
I, personally, disagree with the latter. Generating the animations (i.e. the code that renders them) is to me, probably, the most legitimate use one could think of in that scenario. If the creator already knows what they want to animate, the LLM is just a tool to do it. Furthermore, since there aren’t any software security or code quality concerns, I don’t really see the bad thing (apart from eventual ecological concerns, which are beyond the scope of this post).
However, a friend mentioned that making the animations can be just like painting. There is not art in just asking an AI to generate it, and it won’t be the same as if you painted it yourself, to the best or your ability to represent what you have in mind.
AI for writing
This is where I couldn’t hold onto not having an opinion. Some articles I read are just unbearable. The repetitiveness, the void statements and the AI formulations are just too annoying.
AI formulations
You will notice that LLM have some wordings they use all the time. It is very common to see things like “it is not this, it’s that”, “Adjective. Adjective. Adjective” or three-element enumerations in general. Of course one of these doesn’t mean it’s written by an LLM, but if you see more than one you can start to have serious doubts.
The temptation
It’s tempting, isn’t it? For me, it has even become a reflex: every time I write something I just send it to an LLM to correct it; no one wants to make a mistake in a text, after all. Even if I try to only correct grammar and orthography, LLMs often recommend rewriting things or changing the style. The worst part is that they’re right. Yes, its version is “better written”, but, the thing is, I didn’t write it…
I will send this post to a friend to ask for improvements, the old-fashioned way. Maybe it won’t be as perfect, but right there lays the diversity. If we just wanted a statistically perfect text, there wouldn’t be two books on the same matter. I really hope we stop using AI for this particular task.
The signification
AI-generated text that happens to be right is void of meaning, in some sense. There’s no intention. There are lots of things to say on this topic, but I just wanted to bring it up so that the reader can think about it.
Exercice for the reader
Next time, don’t send your text to an LLM; send it to a friend. Ask real people for an opinion. And if it’s not anything serious, just make mistakes.