Arrival is one of my favorite movies! It raises so many interested questions about language, culture, biology, and the mind. So I made a video about it! Check it out! 👇
Alien Linguistics — The Science of Arrival via
@YouTube
Re-reading early Chomsky, he repeatedly makes the claim that a grammar with a finite alphabet that generates sentences of finite length may produce an infinite set of sentences. Is this true, mathematically?
I'd think finite elements combined into finite-length strings would produce a large set (maybe uncountably large), but not infinite. Very happy to be wrong about this!
This is such a thorny issue, because it intersects bad ideas about language limiting thought (which have driven a lot of linguistic erasure) and valid ideas about languages encoding cultural knowledge (which motivates language preservation).
But a few clarifying points 🧵
Why is it a wild claim that languages indigenous to a land can describe it more specifically, facilitating more precise monitoring?
There’s words for flavors in my mother tongue that don’t have English equivalents. Plenty of my mom’s cooking advice is predicated on knowing them
Linguistics is a modeling science. The purpose is to create formal models that describe the behavior of a system. The models can be falsified by checking the outputs of the model against the outputs of the system.
Which is exactly how theoretical physics happens to operate!
My mom came and sat in on my intro to cog sci class this morning--a big lecture hall class with >100 students. AND SHE RAISED HER HAND AND ASKED A QUESTION
I'm dying
Is it weird to anyone that in all their travels the Austronesians never established any permanent settlements on the Australian continent? Were they avoiding it? Was it just an effect of currents or natural barriers?
@LinaJan02491152
I don't think this is true. I've cooked recreations of ancient Roman and Mesopotamian recipes, and I think you can do a lot with herbs, salt, and fermentation. It wasn't all tasteless mush!
"western civ" bros are deeply unserious. "Have you read the Iliad?" Bro, the Iliad is FICTION
Do you know the history of the Hittite empire? Of Arzawa, of Wilusa? Do you know the exploits of the great kings? Hattusili, Muwatalli, Suppiluliuma?
Classical Latin was complex compared to modern languages. (More noun cases, more verb forms…) But its predecessors were even more complex, back to Proto-Indo-European. Yet there must have been a time when languages evolved to *increasing* complexity. When was the peak, and why?
The syntactician with nonstandard grammatical intuitions.
The phonologist who can't pronounce the department chair's name.
The typologist who only speaks English.
The phonetician who probably should have been a physicist.
They are... the linguistics department
If imaginary numbers just give an additional axis for the vowel chart, arguably languages are already using it!
Vowels can vary in lots of dimensions: height, backness, rounding, tone, phonation, and length.
Why this bothers me: when people say "basically every language other than English", it's likely they're thinking of a particular set of languages: the languages of Europe
This is probably the best example image of recursion that exists.
Chomsky and Halle holding a picture of Chomsky and Halle holding a picture of Chomsky and Halle
It's good to expand our linguistic scope beyond English, but let's not replace it just with popular European languages we might have studied in high school! There are a lot of languages and a lot of diversity!
Sperm whales have equivalents to human vowels.
We uncovered spectral properties in whales’ clicks that are recurrent across whales, independent of traditional types, and compositional.
We got clues to look into spectral properties from our AI interpretability technique CDEV.
The ship of thesaurus problem: if you gradually replace every word of an essay with a more erudite and obscurantist synonym, will it be the same essay?
Identifying AI generated images is a crucial skill for navigating an online ecosystem that is becoming inundated with AI generated content. Here are a few tips to recognize AI images 👇
I disagree, but the inverse ("LLMs are good cognitive models") is also wrong.
Hot take: what LLMs demonstrate is that language (specifically, the corpus) has way more information in it that is accessible via linear regression than we thought.
Hot take: the unreasonable effectiveness of large language models is an illusion and the real scientific discovery is that human language is massively less complex than we thought it was.
Graduating highschool: I don't know anything
1 year of college: I know everything
4 years of college: I don't know everything
1 year of grad school: I don't know anything
6 years of grad school: no one knows anything
5. Many linguists (including me) are convinced that all languages are equivalently expressive - every language can express the same set of human thoughts.
There is no such thing as "untranslatable" words - any word can be translated. We might just have to use more than one word!
On Monday, I had my students debate the function of language. Is language primarily for thought, or primarily for communication? I split the class in half and had each group argue for one side. Left is how they felt before the debate. Right is how they felt after.
This idea that language has supernatural power, or that magic is mediated through language (eg spells, curses, incantations)--is this idea universal? Do all languages have this power for their speakers?
1. Yes, languages very often encode unique cultural knowledge. But the knowledge precedes the language - you can't gain the knowledge simply by being exposed to the right word forms.