chatGPT makes me a language
exploring chatGPT’s language-making abilities
It’s no doubt that ChatGPT immediately took a hold of the human world, whether it be a student cheating on an essay or a congressman using it to write a speech to deliver. It seems like no matter what the task is, ChatGPT is able to do it (to some extent). I, for one, have wanted to create my own language, or conlang, for a while now; naturally, I was drawn to ChatGPT to help me out. Let’s see how well ChatGPT can create it’s very own conlang:
I started off with a pretty simple command: “write a conlang.” Instantly, ChatGPT spat out a name, a list of phonemes, some basic grammar and tenses, and little bit of vocab. Immediately, I was impressed at how much it provided from such a simple command.
My interest was officially piqued. I ran the command “continue developing Tersal” a couple of times, and here are all the things it came up with:
- Continuous vocab additions: It continued creating a TON of new vocab; Personally, I did not have the patience to look through and check if any of the words repeated. From some of the words I did see though, it seemed to have pulled inspiration from English, such as “pigo” for “pig”, and “kowa” for “cow.” It also created words for conjunctions, adverbs, adjectives, interrogative pronouns, prepositions, and more.
- A writing system: It was a little janky and not exactly practical, but it was definitely attempted
- Noun cases: It created a nominative, genitive, accusative, and dative case. Additionally, it later created noun declensions for each of these cases alongside plurality. It also created a diminutive suffix to indicate little things.
- Various additions to the syntax, such as changing word order and how adjectives modify nouns. It also created the passive voice, relative clauses, conditional sentences, and negative sentences.
- Noun classifiers: ChatGPT created suffixes for nouns that are animate, inanimate, long, and round. Unfortunately, these noun classifiers had similar suffixes to the noun cases, so, in reality, it would be hard to distinguish between the two.
- Verb conjugation: There was an addition of suffixes to indicate the present, past (preterite and imperfect), future, continuous, and perfective tenses. From what I could tell, it does not distinguish between who is doing the verb. It also created a subjunctive mood, indicated by the word “mi.”
- Adverbial phrases: In addition to creating adverbs, it made sure to inform me where it would go (after the verb in this case). Intensifier adverbs were also created, such as “very.”
- Adjective agreement: It created a very intricate table of how adjectives agree with the noun, as the suffixes changed depending on grammatical gender, case, and plurality. In addition to that, it created comparative and superlative forms for these adjectives (tall vs taller vs tallest in English).
- Various other pronouns: Alongside interrogative pronouns, ChatGPT created indefinite and reflexive pronouns.
To get this, I had only ran the “continue developing Tersal” message 7 times, which is incredible for the amount of output it generated.
From this, it can be said that ChatGPT does have some conlanging abilities, but obviously, as a chatbot, it’s not the best; some of the following additions contradicted rules established in previous messages, and suffixes ended up overlapping one another. I had a lot of fun looking through whatever it created, and I recommend any conlangers out there to do the same if in need of ideas.
