AI-Generated Texts and the Progress Of Language
After observing the outcome of several texts generated by Chat GPT (some following my prompt, others not), I got to think about the effectiveness of those texts as communication tools. There was something in those texts, an aspect that was there bugging me, that I found to be ultimately interesting. To go into that, let’s first talk about social protocols and how this relates to Chat GPT and similar tools.
1. Historical protocols
We as humans use language and the many indirect protocols related to the language to communicate properly, or at least to aggregate some sort of meta-value to our communication that goes beyond the meaning of individual words or sentences.
Back in the day people used letters to communicate. Letters were at times very indirect and full of flourishing and bits that could be easily discarded in a social media message. Part of this protocol is the way of filling the envelope info, date and local usually mentioned on top of the letter and the introduction and endings (“Dear John, … Best regards, …”). Another part is the level of indirectness and the formality demanded.
When letters were a thing outside of the legal realm (and Germany), the difference between a formal approach and an informal approach could be a matter of life and death, literally. Flourishing texts with unnecessary language constructs was a requirement, in some cases more than others, but this was, in fact, an essential part of writing a letter, as if it was a “test” for the social skills of the writer.
Fast forward to today and we live in a much different world, one in which email made communication get instantaneous and, a bit later, social media services like Twitter almost required people to be more direct and less dependent on protocols. I found it funny to see the first messages my dad (in his 70s) used to write on social media, he actually used to write some letter protocol bits like “Best regards” to his WhatsApp messages.
As my dad eventually did, people are dropping the unnecessary bits of information and politeness and being more transparent and direct. In my view, this is a natural outcome, a match for our fast-paced days. To improve our productivity and knowledge we tend to go towards removing communication barriers and this is part of it (although other parts can be more polemic, like the overall disregard for feelings in written communication and the tendency to drop smaller languages in favour of bigger ones).
2. AI texts and protocols
Going back to ChatGPT, one of the first contacts I had with people actually using it was when a friend of mine used it to generate text for an e-mail. She used it to generate a text that was polite and flourished, full of the protocols which I mentioned above we’re distancing ourselves from.
In the case of an e-mail, many people still follow rules similar to the rules people followed for letters, maybe a little less incisively, and AI tools will almost necessarily generate a text that follows that, especially if prompted to write an email text explicitly.
In other words, what my friend did was use an AI tool heavily “trained” in formal texts like articles and academic texts, to generate socially-accepted formal text from what was probably a quick social media message like “do this” (with the frequent “please”, of course).
It was a translation, not between two languages but between two protocols (even if one of those protocols is somehow a “non-protocol”). The effectiveness and utility of the service are obvious, as people still unconsciously demand some flourishing on e-mails, especially on an e-mail asking for something.
3. AI as a language influence
There’s no problem at all in using AI to do the boring bits you don’t want. The conceptual problem with that, though, which leads me to the point at the very beginning of this piece, is that AI language tools, by mimicking the way how people have written so far, will possibly slow the progress of language towards a more direct communication, because it provides the bits people don’t actually like to write in a painless way.
If writing a flourished message with AI help is easy that seems okay, but what about being easy to read? By the way, it would be very funny to see one person generating flourished texts with AI on one end and another using AI to translate those same texts into more direct messages.
Also, there’s the issue of keeping the garbage there for no good reason. I have seen articles pointing out how easy it is (at least now) to identify AI-generated articles. Supposedly, you just look for actual relevant content versus the sheer number of words, and if the proportion is too small, you would pinpoint there’s an AI in that. I don’t know if that’s true all the time, as we still have politicians and lawyers, but ChatGPT and others when asked for an article, for example, do indeed generate a lot of filler content, a sort of content that has no value whatsoever other than showing one can write filler stuff and, indirectly, abide by social rules (that tend to disappear).
Considering the projected growth in AI-generated text though, will we still see actual value in following those language protocols automatically in the years to come or will it change when most people start to use those tools? Maybe, going ahead, the concept of a “better writer” will sway more and more towards the ones with the ability to pack more content into less text, at least while the ideas will still be in our human heads.
I have made this [letter] longer than usual because I have not had time to make it shorter. (Blaise Pascal)
In the end, if I had to bet on something, it would be on writers thriving with short, more direct texts, which is something not so simple, as writing concisely is an art in itself, as the famous quote above illustrates. Ironically, for that, we can also have AI help: you write a huge article like this one and ask ChatGPT to summarise it. I clearly didn’t do it for this article, maybe I should have.
Thinking about what’s important in a text, going out of this “protocol age” we can go straight to the message pretty much in the same way (and different scale) we largely went away from etiquette rules at the table, for example. In that we may think overly formal and flourished language like the “spoons, forks and knifes protocol” of tomorrow.
